Computer Communications 75 (2016) 1ndash25

Contents lists available at ScienceDirect

Computer Communications

journal homepage wwwelseviercomlocatecomcom

Next generation IEEE 80211 Wireless Local Area Networks

Current status future directions and open challenges

Boris Bellalta alowast Luciano Bononi b Raffaele Bruno c Andreas Kassler d

a Department of Information and Communication Technologies Universitat Pompeu Fabra Barcelona 08018 Spainb Department of Computer Science and Engineering University of Bologna 40127 Bologna Italyc Institute for Informatics and Telematics (IIT) Italian National Research Council (CNR) Pisa 56124 Italyd Computer Science Department Karlstad University 65188 Karlstad Sweden

a r t i c l e i n f o

Article history

Received 23 April 2015

Revised 8 October 2015

Accepted 18 October 2015

Available online 10 November 2015

Keywords

WLANs

IEEE 80211

Video streaming

Cognitive radio

Internet of Things

a b s t r a c t

A new generation of Wireless Local Area Networks (WLANs) will make its appearance in the market in the

forthcoming years based on the amendments to the IEEE 80211 standards that have recently been approved

or are under development Examples of the most expected ones are IEEE 80211aa (Robust Audio Video Trans-

port Streaming) IEEE 80211ac (Very-high throughput at lt 6 GHz) IEEE 80211af (TV White Spaces) and IEEE

80211ah (Machine-to-Machine communications) specifications The aim of this survey is to provide a com-

prehensive overview of these novel technical features and the related open technical challenges that will

drive the future WLAN evolution In contrast to other IEEE 80211 surveys this is a use case oriented study

Specifically we first describe the three key scenarios in which next-generation WLANs will have to oper-

ate We then review the most relevant amendments for each of these use cases focusing on the additional

functionalities and the new technologies they include such as multi-user MIMO techniques groupcast com-

munications dynamic channel bonding spectrum databases and channel sensing enhanced power saving

mechanisms and efficient small data transmissions We also discuss the related work to highlight the key

issues that must still be addressed Finally we review emerging trends that can influence the design of future

WLANs with special focus on software-defined MACs and the internet-working with cellular systems

copy 2015 Elsevier BV All rights reserved

1

(

m

v

l

t

t

t

h

m

a

t

a

h

t

f

a

t

t

m

b

a

i

t

t

i

h

0

Introduction

The IEEE 80211 standard for Wireless Local Area Networks

WLANs) commonly known as WiFi is a mature technology with

ore than 15 years of development and standardisation The earliest

ersion of the IEEE 80211 standard was realised in 1997 as a wire-

ess alternative or extension to existing wired LANs using Ethernet

echnology However since its appearance the IEEE 80211 specifica-

ion has continuously evolved to include new technologies and func-

ionalities and several amendments to the basic IEEE 80211 standard

ave been developed WLANs are currently not only the most com-

on Internet access technology but they have also expanded across

wide variety of markets including consumer mobile and automo-

ive [1] WLANs are thus widely available everywhere (homes public

lowast Corresponding author Tel +34 93 542 14 98

E-mail addresses bbellaltaieeeorg borisbellaltaupfedu

(B Bellalta) lucianobononiuniboit (L Bononi) rbrunoiitcnrit (R Bruno)

ndreaskasslerkause (A Kassler)

e

a

b

a

v

ttpdxdoiorg101016jcomcom201510007

140-3664copy 2015 Elsevier BV All rights reserved

otspots enterprise environments) and IEEE 80211-based radio in-

erfaces are found in many types of devices1

Several factors have contributed to the success of the IEEE 80211

amily of standards interoperability ease of use and flexibility being

mong the most important First the IEEE 80211 standards were ini-

ially designed to be used within unlicensed spectrum bands referred

o as Industrial Scientific and Medical (ISM) bands More precisely

ost IEEE 80211 standards work in 24 GHz and 5 GHz frequency

ands which are globally available although local restrictions may

pply for some aspects of their use Thus anyone can deploy a WLAN

n those bands given that a few basic constraints such as a maximum

ransmission power are satisfied On the downside this also means

hat most WLANs are deployed in an uncontrolled fashion with lim-

ted or no consideration of interference issues This has made it

specially challenging to guarantee performance bounds and reason-

ble Quality of Service (QoS) levels This problem is further exacer-

ated by network densification ie the emerging trend of deploying

large number of base stations in hotspot areas to cope with the

1 According to ABI Research in 2013 more than two billion IEEE 80211-enabled de-

ices were shipped

2 B Bellalta et al Computer Communications 75 (2016) 1ndash25

w

[

W

a

p

m

(

(

a

n

fi

s

8

i

w

c

a

c

t

o

b

l

T

i

F

i

a

v

b

c

a

l

o

A

a

p

s

o

v

a

W

i

i

w

I

t

a

s

W

2

fi

d

w

c

t

t

2 The association between the IEEE 80211 amendments and the different use cases

is specified in Section 2

increase in traffic demands [2] A second fundamental characteristic

of the IEEE 80211 standards is the adoption of a media access control

(MAC) protocol called Carrier Sense Multiple Access with Collision

Avoidance (CSMACA) The main reason is that IEEE 80211-based sys-

tems are half-duplex ie a station cannot carrier-sensereceive while

it is sending and it is hence impossible to detect a collision as in the

case of transmissions over twisted copper wires (eg using Ethernet)

A major advantage of the CSMACA method is that channel access

procedures are simple and cheap to implement as they do not im-

pose stringent timing requirements on the radio interface Further-

more CSMACA protocols are scalable and they provide easy support

for mobility and decentralised network architectures from classical

ad hoc networks to emerging people-centric networks [34] On the

negative side CSMACA protocols can only provide a best effort trans-

mission service and major efforts have been dedicated to the design

of mechanisms for supporting better QoS such as in the IEEE 80211e

amendment [5]

The perceived shortcomings of the first WLAN products have

driven the evolution of the IEEE 80211 standards [6] In particu-

lar throughput enhancements have been a key priority in the IEEE

80211 technology development The key enabler for high-throughput

WLANs was the adoption of new physical-layer techniques The first

of these techniques was the orthogonal frequency-division multi-

plexing (OFDM) which allowed achieving maximum data rates up to

54 Mbs However it is only with the adoption of the IEEE 80211n

amendment in 2009 that the throughput performance of WLANs

came close to that of a wired Ethernet network as a result of the

introduction of multiple-input multiple-output (MIMO) technologies

[7] At the same time new amendments to the original standard

were proposed to foster a more diversified use of WLAN products in

various application domains For instance the IEEE 80211p amend-

ment was approved in 2010 This defines enhancements to the IEEE

80211 standards to support vehicle-to-vehicle (V2V) and vehicle-

to-infrastructure (V2I) communication (together referred to as V2X)

in the 59 GHz band which is licensed for Intelligent Transporta-

tion Systems (ITS) [8] Following the same diversification strategy

the IEEE 80211s amendment was approved in 2011 this described

how wireless mesh networks should operate on top of the existing

IEEE 80211 MAC protocol [9] This includes the specification of new

infrastructure-based elements needed for mesh networking and the

routing protocol to establish mesh paths between these elements

In an attempt to consolidate and systematise all the adopted IEEE

80211 enhancements the last IEEE 80211 standard (identified as

IEEE 80211-2012) was finally released to incorporate in an unique

specification all the amendments published from 2008 to 2011 [10]

As pointed out above the technological development of the WLAN

specifications is a continuously evolving process Thus while the IEEE

80211ndash2012 major revision of the IEEE 80211 standard was finalised

the IEEE 80211 working group was also rapidly moving its focus

towards next-generation WLANs [11] Three key drivers were fore-

casted (i) Machine-to-Machine communications (ii) High-Definition

Multimedia Communications and (iii) ldquoSpectrum Sharingrdquo in licensed

bands by using cognitive radio technology Specifically with the

emergence of the Internet of Things (IoT) vision ie a world were

all sorts of smart objects (ranging from home appliances to small

battery powered devices) are connected to the Internet [12] a low-

power WLAN technology is required [1314] At the same time the

widespread diffusion of mobile devices with diverse networking and

multimedia capabilities as well as the wide adoption of advanced

multimedia applications is fuelling the growth of mobile video traf-

fic which was already more than half of the global mobile data traf-

fic by the end of 2013 [15] Thus WLANs need specific functions to

cope with various multimedia applications including real-time inter-

active audio and video or streaming livestored audio and video [16]

Finally new regulations for the unlicensed usage of TV white spaces

are offering new opportunities for additional spectrum utilisation

hich can be particularly useful to improve rural coverage of WLANs

17] However cognitive radio mechanisms are required for enabling

LAN communications in TV white spaces A new generation of

mendments is consequently under development or has been com-

leted since 2012 to address these new application requirements The

ost relevant are the IEEE 80211aa (approved in 2012) IEEE 80211ac

approved in 2013) IEEE 80211ad (approved in 2012) IEEE 80211af

approved in 2013) IEEE 80211ah (in progress expected for 2016)

nd IEEE 80211ax (in progress expected in 2019) among others2

In this survey we discuss the most compelling challenges of the

ew usage models and applications for WLANs that we have identi-

ed above Then based on those scenarios we classify and review a

elected group of IEEE 80211 amendments ie IEEE 80211ac IEEE

0211ax IEEE 80211aa IEEE 80211ah and IEEE 80211af by describ-

ng the new technologies and functionalities they introduce to cope

ith these challenges such as multi-user MIMO techniques group-

ast communications dynamic channel bonding spectrum databases

nd channel sensing enhanced power saving mechanisms and effi-

ient small data transmissions A summary of the main features of

hese amendments in provided in Table 1 It is important to point

ut that the IEEE 80211 specifications do not define all mechanisms

ut they typically provide the building blocks and interfaces to al-

ow different manufacturers to implement compatible procedures

hus we also provide a detailed review of the main research activ-

ties in the various areas and we identify open technical challenges

inally we look at emerging new trends for WLANs with a special

nterest in Programmable WLANs and LTE-WiFi interworking Over-

ll this survey provides a comprehensive overview of the most rele-

ant features in next-generation WLANs which may be of interest to

oth researchers and engineers working in the field For the sake of

ompleteness in Table 2 we also list the other on-going IEEE 80211

mendments that have not been analysed in this survey

Given the importance of WLANs other surveys have been pub-

ished on the IEEE 80211 standards Earlier surveys primarily focused

n presenting the different classes of proposed MAC protocols [18]

complete overview of the wealth of amendments that have been

ccepted or were in the process of being standardised before 2010 is

rovided in [1] More recently other surveys have given detailed con-

ideration to specific amendments (eg IEEE 80211s [19]) or classes

f similar amendments [111620] However none of the existing sur-

eys follows our use-case oriented approach and covers in such detail

ll the amendments that we believe will be relevant in coming years

e also include some of the latest advances and related research

The structure of this survey is illustrated in Fig 1 and explained

n the following In Section 2 we introduce the four key scenar-

os for WLAN technologies that are considered here In Section 3

e focus on high-throughput WLANs presenting the IEEE 80211ac

EEE 80211ax and IEEE 80211aa amendments Section 4 discusses

he IEEE 80211ah amendment to support M2M communications

nd we review the IEEE 80211af for WLANs operating in TV white

paces Finally Section 6 presents some emerging trends for future

LANs

Future scenarios and new challenges

WLANs can be found everywhere They are common in homes of-

ces public parks in cities shops airports and hotels among many

ifferent places Todayrsquos WLANs are able to provide a fast and reliable

ireless access to Internet for browsing the web exchanging files

hatting receiving and answering e-mails and for low-quality real-

ime audiovideo streams as just a few representative examples of

heir current usage This situation is changing rapidly however The

B Bellalta et al Computer Communications 75 (2016) 1ndash25 3

Table 1

Summary of the IEEE 80211 amendments that are reviewed in this survey

Amendment Release Band Goal New features

80211aa 2012 24 5 GHz Robust streaming of audiovideo

streams

bull Groupcast communication

mechanisms

bull Intra-access category

prioritisation

bull Stream classification service

bull Overlapping BSS management

80211ac 2014 5 GHz Very high-throughput WLAN in

lt 6 GHz band

bull Channel bonding

bull Multi-user Downlink MIMO

bull Packet aggregation

80211af 2014 470ndash790 MHz (EU) WLAN in the TV White Space bull Geolocation-based spectrum

databases

bull Channel sensing

bull Non-contiguous channel

bonding

54ndash72 76ndash88 174ndash216

470ndash698 698ndash806 MHz (US)

80211ah 2016 902ndash928 MHz (US) WLAN in the sub 1 GHz band bull Enhanced power saving

mechanisms

bull Hierarchical station organisation

bull Efficient small data

transmissions

863ndash868 MHz (EU)

755ndash787 (China)

9165ndash9275 MHz (JP)

80211ax 2019 24 5 GHz High efficiency WLANs (HEW) bull Dynamic channel bonding

bull Multi-user Uplink MIMO

bull Full-duplex wireless channel

Table 2

List of other on-going and upcoming IEEE 80211 amendments

Amendment Release Band Goal

80211ae-2012 2012 24 5 GHz Prioritisation of management frames

80211ad-2012 2012 5705ndash64 GHz (US) Very high-throughput WLAN in the 60 GHz band

57ndash66 GHz (EU)

59ndash6290 GHz (China)

57ndash66 GHz (JP)

80211ai 2016 ndash Fast initial link setup

80211aj 2016 45 59ndash64 GHz WLAN in the Chinese Milli-Meter Wave frequency bands

80211aq 2016 ndash Pre-association discovery (PAD)

80211ak 2017 ndash Enhancements for transit links within bridged networks

Fig 1 Survey organisation

n

c

n

c

s

m

u

a

m

v

T

t

umber of persons that use Internet applications and objects that are

onnected to the Internet is growing every day proportionally to the

umber of new applications and services that constantly appear This

learly results in a steady increase of the Internet traffic Two repre-

entative examples of the change in Internet use are (i) the high de-

and for mobile-rich multi-media content mainly motivated by the

se of smart-phones tablets and other multimedia portable devices

nd (ii) the increasing interest in IoT applications driven by the al-

ost ubiquitous existence of devices able to collect data from the en-

ironment ranging from low-power sensor nodes to connected cars

herefore WLANs must also evolve to provide effective solutions

o these new upcoming scenarios and the challenges they pose to

4 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Table 3

Performance requirements for different HD streaming applications

Type Max data rate Max latency

Uncompressed raw video 149 Gbits 100 ms

uncompressed HDTV 150 Mbits 150 ms

Blue-ray Disc 54 Mbits 200 ms

MPEG2 HDTV 192 Mbits 300 ms

MPEG4 HDTV 8ndash10 Mbits 500 ms

f

c

p

B

n

t

p

t

c

f

a

b

a

S

2

n

l

o

s

d

c

i

c

T

n

i

C

c

p

p

i

a

r

s

b

c

c

i

p

l

a

t

t

3

8

s

f

p

a

a

t

3

t

t

l

satisfy their requirements Four of the key use cases for next-

generation WLANs are discussed in the following subsections

21 High-quality multimedia content delivery

Our new mobile and portable devices are designed to handle rich

multimedia contents including high-definition video and images

Table 3 reports the requirements in terms of maximum data rate and

latency for some of the most common real-time video applications

[21] Key scenarios in which the support of real-time video trans-

mission is required of course include Internet TV and video stream-

ing Similarly scenarios in which multiple users connect to the same

wireless network to request different multi-media content at the

same time are increasing every day However not all multi-media

content is real time Stored video and image files can also be ex-

changed between different devices Those files can have sizes rang-

ing from a few Megabits to several Gigabits hence requiring a high

network transport capacity in order to provide a good Quality-of-

Experience to end users Although video encoding schemes exist that

offer substantial video compression efficiency such as H264MPEG-

4 AVC [22] WLANs must be able to achieve very high transmission

rates and have content-aware mechanisms that are specifically de-

signed for multi-media applications to ensure a satisfactory service

for multimedia delivery The mechanisms that are considered by var-

ious IEEE 80211 standardisation groups to satisfy those requirements

are described in Section 3 such as group-cast communication proto-

cols single and multi-user spatial multiplexing and channel bond-

ing among others to make the communication more efficient and of-

fer higher transmission rates The reference IEEE 80211 amendments

for high-quality multimedia content delivery are IEEE 80211aa IEEE

80211ac and IEEE 80211ax

22 Machine-to-Machine (M2M) communications

The almost ubiquitous presence of sensoractuator devices that

are able to interact with the environment has fostered the creation

of new services and applications Concepts such as smart cities and

smart grids are being developed on the basis of the existence of those

sensoractuator networks to achieve a more sustainable use of the

environmental resources and provide citizens with a higher quality

of life [2324]

In a classic sense Wireless Sensor Network (WSN) technologies

are used to collect data from spatially distributed sensor nodes and to

transmit the data over a multi-hop wireless network to a central sink

[25] The M2M paradigm is broadening the scope of the WSN con-

cept because it enables networked devices wireless andor wired as

well as services to exchange information or control data seamlessly

without explicit human intervention Clearly M2M communications

face most of the technical challenges of WSNs One of the main lim-

itations of WSNs and M2M systems is that the network nodes are

usually battery powered or have limited access to power sources De-

signing mechanisms and protocols to reduce their power consump-

tion with the goal of extending the network lifetime is therefore cru-

cial for the successful commercial take-up of these kinds of networks

Fortunately devices in M2M systems typically generate or consume a

limited amount of data per unit of time Thus they can spend a large

raction of their time sleeping This facilitates energy saving at the

ost of additional complexity for the channel access and networking

rotocols

Popular wireless protocol standards for M2M communications are

luetooth ZigBee and BT-LE [12] An alternative promoted by mobile

etworks is to connect devices in M2M systems directly to the In-

ernet by using the cellular network infrastructure for which specific

rotocols are being developed [26] WLANs are envisioned as an al-

ernative to both multi-hop WSNs and cellular networks However

urrent WLANs are not able to satisfy the minimum requirements

or M2M communications [13] Novel specific power-saving mech-

nisms are required to support the long periods of inactivity needed

y the sensoractuator devices and to manage the thousands of nodes

ssociated with a single AP These challenges will be discussed in

ection 4 when presenting the IEEE 80211ah amendment

3 Efficient use of the spectrum

The ISM bands are used by several wireless communication tech-

ologies including IEEE 80211 IEEE 802154 and Long Term Evo-

ution (LTE)-Unlicensed networks This results in a high spectrum

ccupancy Unfortunately wireless networks operating in the same

pectrum region can suffer from mutual interference which might

egrade the performance of all of them This is exacerbated by the un-

ontrolled deployment of wireless networks in the ISM band which

s typically very common in urban environments For example let us

onsider a building with several apartments and a WLAN in each one

here would easily be several WLANs operating in overlapping chan-

els and suffering mutual interference [27] To deal with this issue it

s expected that new APs will increasingly incorporate DCA (Dynamic

hannel Allocation) mechanisms to select and update their operating

hannel at run-time

An alternative approach to alleviate the spectrum occupancy

roblem is to move to a different part of the spectrum even if the new

art of the spectrum is occupied by communication systems operat-

ng under a license In that case WLANs would be the secondary users

nd therefore must avoid causing interference to the primary users In

ecent years the change from analogue to digital TV broadcast emis-

ions has resulted in a reorganisation of the spectrum at VHFUHF

ands This reorganisation has shown that there are many empty TV

hannels called TV white spaces that can be used for data communi-

ation especially in rural regions [28] Furthermore WLANs operat-

ng in those TV white spaces can take advantage of radio propagation

roperties in the UHF band to provide large coverage areas The chal-

enges to be addressed to use CSMACA protocols in VHFUHF bands

s well as how to obtain higher transmission rates when the spec-

rum is fragmented will be discussed in Section 5 when presenting

he IEEE 80211af amendment

High performance WLANs for multimedia applications

This section reviews the IEEE 80211ac IEEE 80211ax and IEEE

0211aa amendments These three amendments target multimedia

cenarios by introducing new physical-layer technologies and MAC

unctionalities to improve the WLAN capacity and QoS provision Ap-

lication examples include home scenarios in which a WLAN AP can

ct as an Internet gateway and wireless media server for home appli-

nces (eg IPTV set-top boxes projectors game consoles) and con-

ent storage devices A possible use case is illustrated in Fig 2

1 The IEEE 80211ac amendment

IEEE 80211ac [29] aims to provide users with a throughput close

o 1 Gbps which represents a roughly four-fold increase with respect

o IEEE 80211n [7] Compared to IEEE 80211n IEEE 80211ac supports

arger channel widths (up to 160 MHz) introduced a new modulation

B Bellalta et al Computer Communications 75 (2016) 1ndash25 5

Fig 2 High-throughput demanding multimedia devices associated to an IEEE 80211acax AP

s

(

3

s

w

b

i

t

s

t

c

o

f

C

t

A

p

c

f

u

c

i

t

u

i

t

t

I

s

m

[

a

t

u

s

T

e

S

Fig 3 A DL-MU-MIMO transmission in IEEE 80211ac Note that the AP transmits two

spatial streams in SU-MIMO mode to STA B and a single spatial stream to STA A Packet

aggregation is used in both transmissions The PHY header is transmitted ommnidirec-

tionally to inform the selected STAs about the next MU-MIMO transmission

n

C

c

F

D

t

t

s

h

d

d

b

m

k

c

M

A

D

d

3

n

(

e

i

i

I

cheme ie a 256-QAM modulation and downlink multiuser MIMO

DL-MU-MIMO)

11 Novel features

The most relevant new features included in IEEE 80211ac are de-

cribed in the following

Channel bonding IEEE 80211ac enables the use of channel band-

idths of 20 40 80 (mandatory) and 160 MHz (optional) Channel

andwidths larger than 20 MHz are created by ldquobondingrdquo (ie group-

ng) a group of consecutive 20 MHz channels and aim to offer higher

ransmission rates

Two extensions have been proposed in IEEE 80211ac for the ba-

ic DCF (Distributed Coordination Function) access method in order

o support channel bonding (i) the Static Bandwidth Channel Ac-

ess Protocol (SBCA) which always transmits over the same group

f 20 MHz channels and requires that all sub-channels are idle be-

ore starting a packet transmission and (ii) the Dynamic Bandwidth

hannel Access scheme (DBCA) which is able to dynamically adapt

he channel width to the instantaneous spectrum availability [3031]

s expected in dense scenarios the use of DBCA offer a much better

erformance than SBCA due to adaptability [32]

To avoid hidden terminals operating in any of the 20 MHz bonded

hannels the IEEE 80211ac amendment includes extended RTSCTS

rames in order to signal the maximum channel width that can be

sed at both the transmitter and the receiver In case the CTS in-

ludes a lower channel width than the RTS the transmitter will adopt

t Similarly to the ACK frames when the RTS and CTS frames are

ransmitted they are duplicated over all the 20 MHz sub-channels

sed The operation and performance of channel bonding in WLANs

s thoroughly analysed in [27] showing the new interactions be-

ween neighbouring WLANs that may appear and their impact in the

hroughput of each one

Downlink multiuser MIMO The main novelty introduced by the

EEE 80211ac amendment compared with the IEEE 80211n one is the

upport of MU-MIMO transmissions in the downlink hence allowing

ultiple simultaneous transmissions from the AP to different STAs

58] In the IEEE 80211ac amendment the AP can be equipped with

maximum of eight antennas and send up to four spatial streams

o two different users or up to two spatial streams to four different

sers at the same time

When an IEEE 80211ac AP performs a multi-user transmission it

pecifies the group of STAs to which that transmission is directed

his information is contained in the new IEEE 80211ac PHY head-

rs which are broadcast omni-directionally to all STAs The way

TAs are grouped is decided by the AP after obtaining the chan-

el state information (CSI) feedback from all STAs To gather the

SI information by the AP IEEE 80211ac considers only an explicit

hannel sounding feedback mechanism called Explicit Compressed

eedBack (ECFB) The channel access is governed by EDCA (Enhanced

istributed Channel Access) At each transmission attempt the mul-

iple access categories (AC) managed by the AP should contend for

he channel medium as only one AC can be served for each transmis-

ion attempt In the case that the queue associated with the AC that

as won the internal contention does not contain packets to enough

ifferent destinations to fill all the available spatial streams it can

ecide to share the remaining ones with the other ACs

Packet aggregation To increase the efficiency of each transmission

y reducing unnecessary overheads IEEE 80211ac allows the trans-

ission of several MPDUs aggregated in a single A-MPDU Then to ac-

nowledge each MPDU individually a Block ACK packet is used which

ontains a bitmap to indicate the correct reception of all included

PDUs Thus leveraging on the information contained in the Block

CK the transmitter is able to selectively retransmit only those MP-

Us that have failed instead of the whole A-MPDU Fig 3 illustrates a

owlink MU-MIMO transmission in which packet aggregation is used

12 Open challenges

Since the IEEE 80211ac amendment has recently been fi-

alised current research around it should cover two main aspects

a) understanding the performance bounds of IEEE 80211ac which

ntails the development of new models simulation tools and exper-

mental platforms of IEEE 80211ac-based WLANs and (b) propos-

ng specific solutions for those aspects that are not defined by the

EEE 80211ac amendment on purpose such as the mechanism for

6 B Bellalta et al Computer Communications 75 (2016) 1ndash25

g

i

l

a

b

b

f

o

w

o

H

t

e

n

a

f

t

m

a

w

s

3

a

8

2

c

s

m

W

r

a

[

o

m

2

n

c

p

r

a

creating the groups of STAs for DL-MU-MIMO transmissions smart

packet schedullers able to decide when the use of DL-MU-MIMO out-

performs SU-MIMO transmissions and the implementation of the

TXOP sharing feature between several ACs The results and conclu-

sions obtained in both cases will be very valuable in the development

of IEEE 80211ac technologies as well as in the conception of the fu-

ture amendments that will substitute IEEE 80211ac in four to five

years such as the recently initiated IEEE 80211ax

Following the first mentioned research direction there are several

efforts that have focused on understanding both theoretical and exper-

imental performance bounds of IEEE 80211ac The maximum downlink

throughput that an IEEE 80211ac AP can achieve when packet aggre-

gation channel bonding and different spatial stream configurations

are considered is presented in [33] In [34] the authors evaluate the

IEEE 80211ac performance experimentally using commodity devices

focusing on the effects that the use of wider channels the 256-QAM

modulation and the number of SU-MIMO spatial streams have in

terms of throughput and energy consumption It is worth mentioning

that DL-MU-MIMO was not yet implemented in the equipment they

were using and that feature was therefore not considered The evalu-

ation of a DL-MU-MIMO implementation for WLANs using the WARP

platform is presented in [35] where a deep evaluation of the potential

benefits of DL-MU-MIMO transmissions is done in terms of the loca-

tion of the receivers number of users and user mobility among other

aspects A solution that combines both packet aggregation and DL-

MU-MIMO transmissions is presented in [36] Results show the need

of properly dimensioning the buffer space to achieve the full potential

of such a combination In [37] the authors compare the throughput

achieved by IEEE 80211n and IEEE 80211ac when packet aggrega-

tion is used with and without channel errors They show that in most

cases the packet aggregation mechanism introduced in IEEE 80211ac

outperforms the one in IEEE 80211n An analytical model to evalu-

ate the performance of the IEEE 80211ac TXOP sharing mechanism in

DL-MU-MIMO communications is developed in [38] The main goal of

this study is to identify how the TXOP sharing mechanism could im-

prove the system efficiency while achieving channel access fairness

among the different ACs

How to optimally exploit the new DL-MU-MIMO capabilities pro-

vided by IEEE 80211ac is still an open challenge First due to the need

of frequent CSI exchanges between STAs and the AP it is not yet clear

in which conditions DL-MU-MIMO outperforms SU-MIMO [39ndash42]

or even whether MU-MIMO does or does not outperform multi-user

packet aggregation when the amount of data directed to each des-

tination is not balanced [43] Packet aggregation can be a solution

to balance the duration of the multi-user spatial streams as shown

in [36] although it will always depend on the amount of traffic di-

rected to each destination and the buffer capacity at the AP In [44]

the authors compare different strategies to assign the spatial streams

between the available destinations at each transmission in a fully

connected mesh network showing in ideal channel conditions the

theoretical benefits of MU-MIMO vs SU-MIMO

Closely related to the previous point a second open challenge is

the design of efficient schedulers that consider traffic priorities the

buffer state the different MIMO strategies TXOP sharing policies

grouping of STAs and the availability of fresh CSI feedbacks to max-

imise the throughput and guarantee the required QoS for each ac-

tive traffic flow It is important to consider that the availability of up-

dated CSI estimates from all STAs allows the AP to reduce the mutual

interference between the transmitted spatial streams which means

lower packet error probabilities and higher transmission rates How-

ever the overheads for obtaining the CSI from all STAs is large and

increases linearly with the channel sounding rate and the number

of STAs Proposals for reducing the CSI overhead are under develop-

ment For example in [40] the CSI overhead is reduced by inhibiting

the channel sounding whenever possible based on the estimation of

the channel stability for all users Another related problem is how to

roup the STAs as the goal is to find groups of STAs with compat-

ble (ie orthogonal) channels In [45] the authors show the chal-

enges inherent to the group assignment problem and they propose

n heuristic method to solve them TXOP sharing is considered in [46]

y presenting two alternative approaches to enhance the considered

ack-off procedure for the purpose of improving both throughput and

airness

A third key challenge for IEEE 80211ac is to achieve an efficient use

f the spectrum when several channel widths are used in scenarios

ith multiple overlapping WLANs Increasing the channel width the-

retically allows individual WLANs to achieve a higher throughput

owever the presence of other WLANs in the vicinity also increases

he chances of frequency overlapping which may cause the opposite

ffect as there appears inter-WLAN contention [27] Adaptive mecha-

isms to select the channel centre frequency and the channel width

nd MAC protocols to choose the instantaneous channel width used

or each transmission are thus required For instance in [47] the au-

hors focus on the channel selection problem when WLANs can use

ultiple channel widths using a game-theoretic framework In [48]

scheme is proposed to enable the communication between nodes

ith partially overlapping channels which may provide stronger re-

ilience to channel interferences

2 The IEEE 80211ax amendment

In 2014 the High Efficiency WLANs (HEW) Task Group [49] initi-

ted the development of a new IEEE 80211 amendment called IEEE

0211ax The IEEE 80211ax amendment is expected to be released in

019 and to some extent it will be the IEEE 80211 response to the

hallenges of future dense and high-bandwidth demanding WLAN

cenarios [5051]

The challenges in the development of the IEEE 80211ax amend-

ent are to

(i) Improve the WLANs performance by providing at least a four-

fold capacity increase compared to IEEE 80211ac

(ii) Provide support for dense networks considering both the ex-

istence of multiple overlapping WLANs and many STAs in each

of them Spatial reuse of the transmission resources is a must

(iii) Achieve an efficient use of the transmission resources by min-

imising the exchange of management and control packets re-

visiting the structure of the packets and improving channel

access and retransmission mechanisms among others aspects

(iv) Provide backward compatibility with previous amendments

This is achieved by the mandatory transmission of the legacy

PHY preamble in all frames and by keeping EDCA as the basic

channel access scheme

(v) Introduce effective energy saving mechanisms to minimise the

energy consumption

(vi) Support multi-user transmission strategies by further devel-

oping MU-MIMO and Orthogonal Frequency Division Multiple

Access (OFDMA) capabilities in both downlink and uplink

In addition to the aforementioned challenges next-generation

LANs will have to implement some other functionalities beyond the

aw packet transmission and reception Examples are a fast efficient

nd robust handoff between APs in the same administration domain

52] device-to-device communication (D2D) [53] and coordination

f multi-AP networks [54] In the first case the IEEE 80211ai amend-

ent called Fast Initial Link Setup is in progress and expected for

016 Its target is to complete a handoff in less than 100 ms including

ew AP discovery user authentication and configuration Using D2D

ommunication we can avoid the use of the AP as a relay hence im-

roving the overall efficiency as the number of packet transmissions

equired is reduced Finally the virtualisation of network functions

dds a new dimension in the management of multiple APs which in

B Bellalta et al Computer Communications 75 (2016) 1ndash25 7

d

e

I

w

f

t

s

3

P

s

G

t

8

p

C

c

m

c

i

t

a

s

w

f

t

t

e

l

p

e

a

H

d

t

c

o

s

p

n

t

b

a

w

c

t

w

n

8

a

c

w

d

H

n

a

h

p

w

i

t

fi

e

c

u

t

m

p

p

b

p

i

t

8

d

I

m

M

u

t

m

i

t

D

W

I

u

M

i

w

p

h

t

n

o

p

t

s

u

s

r

b

A

t

O

t

t

w

s

p

o

i

W

t

t

a

[

m

ense scenarios can contribute to notably improving the user experi-

nce We further discuss this last topic in Section 6

Different from the other amendments covered in this survey the

EEE 80211ax amendment is just in its initial stages of development

ith only very few technical aspects consolidated at this stage There-

ore in the following subsection we will overview both the new fea-

ures and open challenges of the IEEE 80211ax amendment at the

ame time

21 Novel features and open challenges

The IEEE 80111ax Task Group is currently working in four areas

HY MAC Multi-user and Spatial Reuse [49] Next we will overview

ome of the topics currently under discussion in the IEEE 80211 Task

roup in each category

PHY layer The IEEE 80211ax PHY layer will be an evolution of

he IEEE 80211ac one The challenges in the design of the IEEE

0211ax PHY layer are related with the extensions required to sup-

ort multi-user MU-MIMO and OFDMA transmissions and Dynamic

CA Also improvements in the supported modulation and channel

oding techniques will be likely considered to allow for higher trans-

ission rates at lower SNR values For example IEEE 80211ax may

onsider LDPC (Low-Density Parity Check) coding which is optional

n IEEE 80211ac instead of the traditional convolutional codes as

hey provide a coding gain of 1ndash2 dB Moreover the PHY layer may

lso include some functionalities to support the use of Hybrid ARQ

chemes to improve the efficiency of packet retransmissions

Medium acess control In order to keep backward compatibility

ith previous IEEE 80211 amendments besides a common PHY

rame preamble compatible MAC protocols are required This means

hat it is likely that EDCA will be kept as the main channel access

echnique in the IEEE 80211ax amendment Therefore the most rel-

vant open challenges are related to EDCA extensions to support a

arge number of STAs improve traffic differentiation capabilities im-

rove the energy consumption and provide mechanisms to fairly co-

xist with neighboring wireless networks

To support a large number of contenders with a low collision prob-

bility a simple solution is to use large backoff contention windows

owever it would also increase the time a STA is in backoff so re-

ucing the number of packets it can transmit A solution to reduce

he backoff duration when using large contention windows is to de-

rease the duration of the backoff slots Latest technological advances

n electronics should require less time to perform a CCA check to

witch between reception and transmission modes and to process a

acket thus enabling such a possibility

Another approach to enhance the underlying CSMACA mecha-

ism in EDCA and achieve a higher efficiency is to consider decen-

ralised collision-free MAC strategies Those MAC protocols are able to

uild collision-free schedules thus improving the network efficiency

s collisions are reduced while preserving backward compatibility

ith the default EDCA implementation An overview of several de-

entralised collision-free MAC protocols can be found in [55] Among

hem CSMAECA [56] is especially relevant since it is fully compatible

ith EDCA and latest upgrades made it adaptive to the instantaneous

umber of contenders in a single WLAN

In addition to EDCA IEEE 80211ax WLANs can rely on the IEEE

0211aa amendment to further improve the WLAN traffic differenti-

tion capabilities with intra-AC traffic differentiation and groupcast

ommunication mechanisms among other features In Section 33

e will provide further details about the IEEE 80211aa amendment

IEEE 80211ax will likely keep the same channel widths that were

efined in the IEEE 80211ac amendment ie 20 40 80 and 160 MHz

owever it is expected that IEEE 80211ax will extend current chan-

el bonding mechanisms to further improve the spectrum utilisation

nd the coexistence between neighbouring WLANs For example it

as been shown in [32] that the use of dynamic channel bonding

rovides significant throughput gains in dense scenarios compared

ith the static approach while minimising the inter-WLAN negative

nteractions [27] Furthermore additional mechanisms are required

o fully exploit the use of channel bonding such as the use of ef-

cient algorithms to select the position of the primary channel or

ven to consider the existence of multiple primary channels to in-

rease the number of bonded channel combinations that a node can

se for transmission

The MAC layer in IEEE 80211ax may work with the PHY layer

o implement an efficient Hybrid ARQ mechanism able to retrans-

it only incremental redundancy bits in short packets Opportunistic

iggy backing of data packets in ACKs and vice versa may further im-

rove the efficiency of IEEE 80211ax WLANs by reducing the num-

er of transmissions in a bidirectional data exchange [57] Finally

acket headers can be reduced if shorter STA identificators are used

nstead of MAC addresses and unnecessary fields are avoided in all

ransmissions

Finally in terms of power saving mechanisms it is expected IEEE

0211ax will follow the same trends as in previous IEEE 80211 stan-

ards including the extensions currently under development for the

EEE 80211ah amendment that are discussed in Section 4

Multi-user Multi-user communications will likely be one of the

ain characteristics of IEEE 80211ax as both uplink and downlink

U-MUMO and OFDMA are under consideration The use of multi-

ser communication techniques does not necessarily represent a sys-

em capacity increase because the available transmission resources

ay be the same as in the single-user communication case However

n WLANs the simultaneous transmission from different users is able

o parallelise the large temporal overheads of each transmission (ie

IFS SIFS ACKs packet headers etc) which can notably improve the

LAN efficiency

IEEE 80211ax will further develop the MU-MIMO capabilities of

EEE 80211ac by allowing multiple simultaneous transmissions in the

plink which is known as uplink (UL-) MU-MIMO [58] Similar to DL-

U-MIMO transmissions an open challenge to enable UL-MU-MIMO

s to design a mechanism able to efficiently schedule the users that

ill transmit at the same time In one hand a pure decentralised ap-

roach would be easy to implement with minimal signalling over-

eads However since it requires that all STAs finish their backoff at

he same time it may be very inefficient besides that those STAs may

ot be compatible in terms of their respective spatial channels In the

ther hand a pure centralised approach requires that the AP has com-

lete CSI and buffer occupancy information from all STAs to select

he most suitable group to perform a multi-user transmission Once a

uitable group of STAs is selected by the AP a ldquoTriggerrdquo frame may be

sed to notify the group of selected users that can initiate a transmis-

ion This approach guarantees efficient multi-user transmissions but

equires some extra overheads to collect all the required information

y the AP and signal the selected STAs In both cases new multi-user

CKs will be likely introduced by IEEE 80211ax to acknowledge all

ransmissions with a single control packet

Multi-user OFDMA is also in the agenda for IEEE 80211ax Using

FDMA a channel can be split in several sub-channels and assigned

o different users Likely OFDMA will be implemented in combina-

ion with channel bonding where each of the 20 MHz subchannels

ill be assigned to a different user in both downlink and uplink Be-

ides that a similar operation as in the multi-user MIMO case is ex-

ected as there are almost the same challenges to solve A survey

f current OFDMA proposals for WLANs is presented in [59] show-

ng also how the use of OFDMA is able to significantly improve the

LAN efficiency Fig 4 illustrates an uplink OFDMA and MU-MIMO

ransmission

In addition to Multi-user MIMO and OFDMA the use of Simul-

aneous Transmit and Receive (STR) techniques commonly known

s full-duplex transmission have been suggested for IEEE 80211ax

5051] Using STR a pair of nodes is able to transmit and receive si-

ultaneously [60] which theoretically doubles the channel capacity

8 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 4 Multiuser uplink transmission using two different OFDM subchannels Trans-

missions from STA B and C over the same subchannel represent an uplink MU-MIMO

transmission All transmissions are acknowledged with a single Multiuser ACK The

trigger frame is used to signal the selected STAs

t

t

t

a

s

t

f

f

c

3

s

s

t

n

n

A

d

w

t

fi

a

t

8

i

d

t

r

w

d

R

v

f

s

d

p

B

c

e

k

r

G

a

c

m

t

b

d

3 The concealment address protects legacy stations ie GCR-incapable stations

from receiving duplicated group-addressed frames

The challenge is that both the AP and the STA involved in a full-duplex

transmission have to start to transmit at the same time To achieve

that information about full duplex transmission capabilities can be

included in RTSCTS control packets to set-up a full-duplex transmis-

sion between the transmission initiator and its destination

Spatial reuse Dense WLAN deployments are necessary to offer a

continuous coverage with high transmission rates To improve both

the co-existence with those neighboring networks and the spatial

reuse of the spectrum a WLAN has two options (i) minimise its

area of influence by reducing its transmit power and (ii) accept

higher interference levels by increasing the Clear Channel Assess-

ment (CCA) level Use of both techniques may increase the number of

concurrent transmissions between neighbouring WLANs and there-

fore their capacity although it may also result in the opposite effect

since the achievable transmission rates may be negatively affected by

the higher interference levels observed which is the main challenge

to solve

Due the high WLAN dynamics the use of adaptive systems is cru-

cial but challenging as adaptivity requires extra complexity in terms

of computing and memory resources and there are no guarantees

that the implemented solution converges due to the decentralised

operation of each WLAN The use of DSC (Dynamic Sensitivity Con-

trol) to dynamically adjust the CCA level is one of the aspects cur-

rently under discussion in the IEEE 80211ax Task Group First studies

evaluating the performance of DSC for IEEE 80211ax WLANs show

a clear improvement on the spatial reuse and the area throughput

[61] Another example of the achievable throughput gains obtained

by adapting the CCA level can be found in [62] where the authors

show that gains of up to 100 can be achieved Moreover trans-

mit Power Control (TPC) to mitigate interference between WLANs in

dense scenarios is studied in [63] showing the need of jointly opti-

mising both TPC and CCA to maximise the network performance

Finally sectorisation by using beamforming is also under consid-

eration for the development of the IEEE 80211ax amendment as a

potential solution to improve spatial reuse [64] Using sectorisation

only the nodes of a given area are allowed to receive or transmit data

hence reducing the contention between different networks whenever

they activate non-overlapping sectors A challenge here is to coor-

dinate the different neighboring APs when they belong to different

administration domains Decentralised learning approaches may be

implemented to find feasible temporal patterns of non-overlapping

sectors

33 The IEEE 80211aa amendment

As discussed above legacy IEEE 80211 standards do not provide

robust and efficient delivery of audiovideo streaming services Thus

he IEEE 80211aa amendment was developed to include new fea-

ures and additional mechanisms to improve the performance of real-

ime multi-media content delivery [65] Specifically IEEE 80211aa

ddresses the following five shortcomings of previous 80211

tandards [1666]

(i) the lack of reliable and efficient support for multicast and

group communications

(ii) the incapacity of applying traffic prioritisation to different

multimedia streams or different types of frames from the same

stream

(iii) the absence of methods for cooperative resource sharing

among neighbouring APs

(iv) the lack of mechanisms for graceful degradation of audiovideo

streaming quality

(v) the non-interoperability with existing IEEE 8021 standards for

Audio Video Bridging (AVB)

In the following sections we present in detail the solutions to

hose problems introduced in the IEEE 80211aa amendment We

urther discuss the research studies that have provided the basis

or the IEEE 80211aa design and we identify the remaining open

hallenges

31 Novel features

Groupcast communication mechanisms In most audiovideo

treaming applications a group of clients must receive the same

tream simultaneously A multicast protocol is necessary to avoid that

he same content is replicated throughout the network In wireless

etworks multicast transmission can exploit the intrinsic broadcast

ature of the wireless channel ie broadcast transmissions from an

P are physically received by all other stations in the same collision

omain However multicast and broadcast frames in IEEE 80211 net-

orks are not protected by an acknowledgement mechanism as in

he case of unicast frames Thus layer-2 multicast transmissions de-

ned by legacy IEEE 80211 standards are unreliable and not suit-

ble for streaming applications To partially address this limitation

he Direct Multicast Service (DMS) was first specified in the IEEE

0211v amendment [67] Basically DMS converts multicast streams

nto unicast streams In this way frames destined to a multicast ad-

ress are individually transmitted as unicast frames to the stations

hat joined that multicast group Obviously DMS provides the same

eliability as unicast transmission services but the consumed band-

idth increases linearly with the number of group members To ad-

ress this scalability issue IEEE 80211aa includes the Groupcast with

etries (GCR) service in addition to DMS Specifically the GCR ser-

ice defines new mechanisms and the related management frames

or group formation which allows a set of stations to agree on a

hared (non-multicast) address called the groupcast concealment ad-

ress3 Furthermore the GCR service specifies two retransmission

olicies GCR Unsolicited Retry (GCR-UR) and GCR Block Ack (GCR-

A) When using GCR-UR the AP can proactively retransmit all group-

ast frames a number of times to mitigate the impact of channel

rrors (see Fig 5a)) while receivers are not requested to send ac-

nowledgements Intuitively this approach improves transmission

eliability but it still suffers from scalability issues In contrast when

CR-BA is used the AP sends a burst of consecutive groupcast frames

nd it requests the receivers to reply with a Block ACK frame which

ontains a bitmap to positively or negatively acknowledge trans-

itted frames (see Fig 5b)) The Block ACK mechanism defined for

he GCR-BA service is quite flexible because Block ACK frames can

e requested immediately after a transmission burst or after a ran-

omised back-off delay Furthermore the AP can request the Block

B Bellalta et al Computer Communications 75 (2016) 1ndash25 9

Fig 5 GCR service with different retransmission schemes

A

r

o

l

o

e

v

e

v

d

f

d

s

v

c

s

a

a

f

d

t

F

t

d

v

a

f

c

t

t

o

t

i

c

r

v

i

r

s

o

t

a

(

p

i

f

a

o

i

t

t

c

s

fi

o

c

I

n

i

a

t

a

t

f

i

B

T

f

t

a

g

a

a

i

s

a

d

s

p

m

m

i

a

t

b

t

i

(

r

a

S

i

c

s

c

T

Q

p

a

3

g

t

4 TSPEC is a traffic specification sent from a QoS capable wireless client that requests

a certain amount of network traffic from the AP for the traffic stream it represents

CK frame to all groupcast recipients or only to a subset of them to

educe overheads and delays The advantages of the GCR methods

ver broadcast and DMS have been extensively demonstrated in the

iterature [1668]

Intra-access category prioritisation The IEEE 80211e amendment

nly allows traffic differentiation between four different access cat-

gories (ACs) that are broadly mapped to four application classes

oice (VO) video (VD) best-effort (BE) and background (BK) How-

ver there is a variety of streaming services ranging from simple

ideoconferencing to HD streaming over IPTV systems which have

ifferent QoS requirements (see Table 3) To provide the ability to dif-

erentiate among individual streams IEEE 80211aa includes an ad-

itional scheduling layer with respect to IEEE 80211e IEEE 80211aa

plits each one of the transmission queues associated with voice and

ideo ACs into a primary and an alternate queue In this way spe-

ialised scheduling rules can be applied to decide which queue to

erve when the EDCA function for inter-AC collision resolution grants

n access opportunity to voice or video ACs To facilitate the man-

gement of service level agreements IEEE 80211aa follows the de-

ault mappings between user priority values and traffic types that are

efined in the IEEE 8021D standard [69] It is then straightforward

o further map traffic types onto transmission queues and ACs (see

ig 6) Finally it is important to point out that the intra-AC differen-

iation functionality can be used to provide more sophisticated traffic

ifferentiation than simple stream prioritisation For instance most

ideo applications use Scalable Video Coding (SVC) schemes that en-

ble the partitioning of a video sequence into multiple layers with dif-

erent qualities and rates [70] Typically an SVC-based video stream

ontains a base layer which provides a basic level of quality and mul-

iple enhancement layers which can only be decoded together with

he base layer to improve the video quality Thus the different layers

f the same encoded video steam can be easily mapped to different

ransmission queues to receive differentiated QoS [71]

Stream classification service The stream classification service (SCS)

s an optional service that can be provided by an AP to the asso-

iated stations to classify multimedia streams based on arbitrary

ules that are established directly by the stations instead of the con-

entional 8021D user priorities To this end the station request-

ng the use of SCS must initiate an SCS session by sending an SCS

equest frame to the AP which contains an identifier for the SCS

tream and the descriptor of the classification rule The AP may accept

r reject the requirements specified by the station Once accepted

he AP must assign all frames that match the classification rule to

specific AC When intra-access category prioritisation is enabled

see Section 331) calternate transmit queues so that finer grained

rioritisation can be applied Finally there is also a Drop Eligibil-

ty Indicator (DEI) bit in the SCS descriptor that indicates whether

rames from this traffic stream can be dropped in the case that there

re insufficient resources Specifically frames with the DEI bit set to

ne have a higher probability of being discarded because their max-

mum number of allowed retries is smaller than the default Note

hat how to combine intra-AC queues and frame dropping settings

o achieve graceful degradation of the audiovideo stream quality in

ase of bandwidth shortage is beyond the scope of the IEEE 80211aa

pecification

Overlapping Basic Service Set (OBSS) management Network densi-

cation ie a denser deployment of wireless infrastructure nodes is

ne of the key strategies that is used nowadays to easily increase the

apacity of wireless systems even for indoor WLANs [72] However

EEE 80211 networks have a limited number of orthogonal chan-

els available and even if optimised frequency planning is applied

t might happen that neighbouring APs are mutually interfering and

station may affect multiple overlapping BSSs In this case conges-

ion not only increases but it is also likely to observe an unfair us-

ge of wireless capacity with the channel retained by one AP for long

ime intervals This is mainly due to the neighbourhood capture ef-

ect ie hidden terminal phenomena among APs [73] To address this

ssue IEEE 80211aa specifies a new functionality called Overlapping

SS (OBSS) management which is based on two new mechanisms

he first defines a set of parameters to quantify the load and inter-

erence among neighbouring BSSs such as medium occupancy frac-

ion number of admitted audiovideo streams data traffic volumes

nd the number of BSSs that are using the same channel as the tar-

et one Note that the traffic load consists of two components the

llocated traffic which is derived on the basis of the TSPEC values of

dmitted streams4 and predicted traffic which is evaluated by track-

ng the maximum value of the allocated EDCA and HCCA traffic over

even-day periods Once load measurement reports are exchanged

mong the APs a second OBSS component is responsible for coor-

inated admission control procedures on the basis of two suggested

haring schemes proportional sharing and on-demand sharing The

urpose of both schemes is to keep the total allocated traffic below a

aximum value in order to provide some QoS protection to admitted

ultimedia streams Finally IEEE 80211aa recommends implement-

ng additional OBSS management procedures for channel selection

nd cooperatively creating HCCA schedules that do not collide

Interworking with IEEE 8021AVB Audio Video Bridging (AVB) is a

erm commonly used to denote a set of technical standards developed

y IEEE to support real-time streaming services with bounded latency

hrough IEEE 802 networks [74] This objective is achieved by spec-

fying mechanisms to allow the synchronisation of multiple streams

IEEE 8021AS [75]) and traffic shaping (IEEE 8021Qav [76]) and to

eserve network resources for specific audiovideo streams traversing

bridged local area network by using a signalling protocol called the

tream Reservation Protocol (SRP) (IEEE 8021Qat [77]) IEEE 80211aa

ntegrates the SRP operations with the EDCA admission control pro-

edures Specifically the SRP RequestResponse messages are encap-

ulated in the management frames that are used to carry the traffic

haracteristics and the QoS requirements during admission control

his enables the end-to-end management of resource reservation for

oS guaranteed streams even when one or more IEEE 80211 links are

art of a path from the stream producers (called IEEE 8021Q talkers)

nd the stream consumers (called IEEE 8021Q listeners)

32 Open challenges

In recent years several MAC enhancements have been investi-

ated to improve QoS guarantees for real-time multimedia applica-

ions in IEEE 80211 networks [20] and the IEEE 80211aa standard

10 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 6 Stream classification and inter-AC traffic prioritisation

a

l

t

r

n

a

t

n

i

t

i

t

r

d

p

t

t

e

m

b

t

a

i

s

o

s

a

e

t

i

f

e

m

8

t

I

i

fi

which was finalised in 2012 included several of these proposed im-

provements Significant research efforts have focused on improving

the transmission reliability of multicasting by integrating ARQ mech-

anisms in IEEE 80211-based multicast transmissions Modifications

to the legacy MAC protocol were proposed in [78] to enable the

RTSCTS option in multicast mode and to select one or more multi-

cast receivers (called leaders) for acknowledging multicast data pack-

ets However these enhancements require changes to the standard

specifications The main problems of leader-based ARQ schemes are

leader election and the trade-off between scalability and reliability

The authors in [79] propose selecting the multicast recipient operat-

ing in the worst channel conditions as the unique leader but this ap-

proach may perform poorly in lossy environments In the Batch mode

multicast MAC (BMMM) [80] all multicast recipients are polled by

the multicast originator to send individual ACKs but this scheme is

not suitable for large multicast groups The Enhanced Leader Based

Protocol (ELBP) is proposed in [81] on the basis of multiple ACK-

leaders and block acknowledgement techniques Analytical models

are then developed to help select optimal ACK-leaders to meet ap-

plication QoS requirements However the models apply only to sat-

urated traffic while multimedia streams are typically bursty Another

class of reliable multicast protocols relies on busy tones to reduce

packet losses due to collisions [82] but the additional radio inter-

face needed for the busy tone limits the practicality of such solu-

tions An alternative approach to avoid collisions of multicast pack-

ets is the multicast collision prevention (MCP) scheme [83] which

is based on the use of a shorter waiting time for transmitting mul-

ticast packets An interesting approach is also proposed in [84] to

retransmit lost packets using an online linear XOR coding algorithm

However a modification to the standard MAC protocol is required to

enable simultaneous ACK transmissions In summary several differ-

ent methods have been proposed to improve multicast transmission

reliability by integrating ARQ schemes into the protocol architecture

but there are not conclusive results on which is the best solution The

choice of the most efficient mechanism depends on a variety of in-

terdependent factors such as loss ratios channel congestion multi-

cast group size and QoS requirements of multimedia streams A com-

prehensive analytical framework is needed to optimise the setting of

the parameters for each scheme and to dynamically select the best

one

As discussed above one main difference between unicast services

nd multicast services in the legacy IEEE 80211 standard was the

ack of acknowledgements Another critical difference is that mul-

icast frames must be transmitted using a fixed rate in the basic

ate set while the transmission rate of unicast frames can be dy-

amically adapted to the channel and traffic conditions [85] Thus

group of research papers has investigated the use of rate adap-

ation to improve the throughput of multicast services in IEEE 80211

etworks [7186ndash89] For instance the authors in [86] propose us-

ng RTS frames to allow group members to estimate channel condi-

ions Each member will then send a dummy CTS frame with a length

nversely proportional to channel quality In this way the multicast

ransmitter can use the collision duration to predict the lowest data

ate that can be used for group transmissions The overhead intro-

uced by this mechanism is quite high however The solution pro-

osed in [88] called ARSM also relies on feedback messages sent by

he multicast receivers called multicast response frames to identify

he group member exhibiting the poorest channel conditions How-

ver in this case a different back off timer is associated with each

ulticast receiver depending on the SNR of previously received feed-

ack messages in order to prevent collision An approach similar to

he one employed in the Auto Rate Fallback (ARF) protocol a rate

daptation scheme originally proposed in [90] is used in [87] Specif-

cally the number of successful consecutive transmissions and con-

ecutive transmission failures are used to decide when to increase

r decrease the transmission data rate respectively A modified ARF

cheme is also proposed in [71] which can be applied to videos that

re encoded into two layers namely the base and enhancement lay-

rs However how to integrate rate adaptation with the different re-

ransmission policies that are defined in IEEE 80211aa is still an open

ssue

One research area that is expected to be crucial in the success-

ul development of IEEE 80211aa-based products is the design of

fficient scheduling algorithms for supporting voicevideo traffic Al-

ost all research work in this field has been triggered by the IEEE

0211e amendment that enhanced the original IEEE 80211 MAC with

wo new QoS-aware access mechanisms ie EDCA and HCCA [91]

n principle with a well-designed admission control and schedul-

ng scheme HCCA is able to provide hard QoS guarantees to traf-

c flows [9293] However HCCA is rarely implemented in IEEE

B Bellalta et al Computer Communications 75 (2016) 1ndash25 11

8

c

c

a

s

b

w

o

j

t

f

a

Q

t

t

A

p

t

s

a

f

s

H

a

i

c

i

t

s

f

s

s

m

i

t

s

l

t

t

[

t

i

o

b

r

W

v

m

8

s

i

o

p

q

4

c

e

t

d

a

t

m

Fig 7 WLANs for M2M communications STAs represent sensor and actuator devices

4

t

w

s

a

c

g

e

(

p

o

t

p

9

r

o

a

p

c

o

i

p

b

f

t

s

o

i

4

8

f

t

i

s

m

d

0211e-based WLANs owing to its higher complexity and cost con-

erns Instead EDCA is widely adopted Most papers have thus fo-

used on improving EDCA performance Many papers have proposed

nalytical models for various subsets of EDCA functionalities For in-

tance a saturation-based performance analysis is conducted in [94]

y differentiating the minimum back-off window size the back-off

indow-increasing factor and the retransmission limit The authors

f [9596] also model AIFS differentiation while the model in [97]

ointly captures all the four EDCA parameters for traffic differen-

iation More recent papers have analysed the EDCA performance

or non-saturated conditions and for arbitrary buffer sizes [98] The

uthors in [99] have developed an analytical model to predict the

oS levels that can be achieved once a new voicevideo flow is in-

roduced in the WLAN A Kalman filter is proposed in [100] to ob-

ain estimates on the number of active transmission queues of each

ccess Category in EDCA These analytical models can then be ex-

loited to derive the optimal configuration of the EDCA parameters

o achieve given performance criteria or to design admission control

chemes that preserve QoS constraints For instance a scheme that

ssigns contention-window values to achieve pre-defined weighted-

airness goals is proposed in [101] A control-theoretic scheme is de-

igned in [102] with the goal of minimising the video traffic delay

owever most of these solutions rely on non-realistic assumptions

bout video traffic dynamics An alternative class of solutions dynam-

cally updates the EDCA parameters based on the observed network

onditions In [103] the EDCA parameters are optimised consider-

ng a WLAN with rigid and elastic traffic simultaneously analysing

he interactions between both types of traffic The authors in [104]

pecify several bandwidth-sharing mechanisms with guaranteed QoS

or voice and video traffic Measurement-based admission control

chemes are proposed in [105] A TXOP adaptation method is de-

cribed in [106] that takes into account video frame sizes and trans-

it queue lengths However the main drawback of these solutions

s that they are based on heuristics and hence do not ensure op-

imal and guaranteed performance Finally a third category of re-

earch papers tries to improve video performance by designing cross-

ayer scheduling approaches Specifically these works take advan-

age of multi-layer video encoding to classify the frames according

o their importance and assign them to different access categories

107] For instance the authors in [108] define classifiers and waiting

ime priority schedulers that dynamically change the packet prior-

ties according to end-to-end delay measurements A disadvantage

f this approach however is that an additional adaptation layer may

e needed to implement the complex interactions that are typically

equired between the video coding applications and the MAC layer

e conclude this section by pointing out that existing studies pro-

ide the basic design principles and techniques for improving multi-

edia streaming performance in IEEE 80211 networks Still the IEEE

0211aa standard poses new research challenges that have not been

ufficiently explored and that will require innovative solutions For

nstance scheduling between primary and alternate queues is still an

pen research area as the mapping of individual frames to multi-

le queues in order to achieve graceful degradation of voicevideo

uality [16]

Sensor Networks and machine-type communications

As discussed in Section 2 M2M communications refer to any

ommunication technology that enables sensoractuator devices to

xchange information and perform actions without the manual assis-

ance of humans This section reviews the main features currently un-

er consideration in the development of the upcoming IEEE 80211ah

mendment which targets the main challenges of those networks as

he IoT in general such as the energy consumption or the manage-

ent of many devices

1 The IEEE 80211ah amendment

The IEEE 80211ah amendment [109] aims to provide WLANs with

he ability to both manage a large number of heterogeneous STAs

ithin a single BSS and minimise the energy consumption of the

ensor-type battery-powered STAs

The initial design requirements of the IEEE 80211ah amendment

re detailed in [110] these entail the support of up to 8192 STAs asso-

iated with a single AP the adoption of efficient power saving strate-

ies a minimum data rate of 100 kbps the operation in the license-

xempt sub 1 GHz band and a coverage up to 1 km in outdoor areas

see Fig 7 for an illustrative example) A preliminary assessment of

erformance of the IEEE 80211ah technology in terms of the number

f STAs that can be effectively supported in a single WLAN as well as

heir energy consumption is presented in [111]

IEEE 80211ah operates over different sub-1 GHz ISM bands de-

ending on country regulations 863ndash868 MHz in Europe 902ndash

28 MHz in the US and 9165ndash9275 MHz in Japan China South Ko-

ea and Singapore also have specific channelisations Channel widths

f 1 MHz and 2 MHz have been adopted although 4 8 and 16 MHz

re also supported in some countries IEEE 80211ah furthermore pro-

oses new PHY and MAC layers The IEEE 80211ah PHY layer can be

onsidered to some extent a sub-1 GHz version of the IEEE 80211ac

ne At the physical layer OFDM is the chosen modulation method us-

ng 32 or 64 tonessub-carriers that are spaced by 3125 kHz The sup-

orted modulations include BPSK QPSK and from 16 to 256-QAM A

road range of antenna technologies ranging from single-user beam-

orming to MIMO and DL-MU-MIMO which was first introduced in

he IEEE 80211ac amendment are also included in the IEEE 80211ah

pecification Similarly the IEEE 80211ah MAC protocol include most

f IEEE 80211 main characteristics further extending its power sav-

ng (PS) mechanisms

11 Novel features

This section overviews the extensions introduced by the IEEE

0211ah amendment to the IEEE 80211 PS mechanisms to account

or the specific characteristics of resource-constrained sensor and ac-

uator devices A more detailed review can be found in [110] includ-

ng a performance assessment of IEEE 80211ah in several of the key

cenarios for M2M communications such as agriculture and animal

onitoring smart metering and industrial automation plants In ad-

ition a detailed survey of the IEEE 80211ah is reported in [112]

12 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 8 IEEE 80211ah PS mode for TIM-STAs non-TIM STAs and Unscheduled STAs

a

d

w

i

a

a

c

m

i

a

e

t

s

m

t

t

u

b

i

I

s

d

f

a

b

s

a

b

S

T

c

t

w

I

4

a

a

which addresses aspects not considered in [110] such as the use of

sectorisation to avoid overlapping between multiple IEEE 80211ah

WLANs and the use of relays to further extend the IEEE 80211ah cov-

erage among other aspects Finally in [113] both IEEE 80211ah PHY

and MAC characteristics are introduced with emphasis on the ben-

efits of using the sub-1 GHz ISM band in terms of link budget the

location of the OFDM pilots for outdoors operation the use of bidi-

rectional transmission opportunities and the new packet formats to

minimise overheads among others

Enhanced power saving mechanism PS mechanisms for WLANs

were already considered in the development of the first IEEE 80211

standard with the goal of improving the lifetime of battery equipped

devices [114] In PS mode STAs keep the transceiver in sleeping mode

as much time as possible They periodically wake up to listen to the

beacons transmitted by the AP Those beacons indicate whether an

STA has packets waiting for it at the AP In the positive case that STA

remains awake and requests the delivery of those packets Otherwise

given it has nothing to receive it returns to sleep mode until the next

beacon is expected

In the IEEE 80211ah amendment time is divided into pages DTIM

(Delivery Traffic Indication Map) periods TIM (Traffic Indication Map)

periods and slots DTIM and TIM periods begin with the correspond-

ing DTIM and TIM beacons sent by the AP The functions of DTIM and

TIM beacons are described below

1 DTIM beacons They inform as to which TIM Groups (ie the

group of STAs assigned to the same TIM period) have pending

packets at the AP

2 TIM beacons Each TIM message informs a TIM Group about

which specific STA has pending data in the AP Between two

consecutive DTIMs there are as many TIM beacons as TIM

Groups

Using this DTIMTIM-based approach any STA can enter into a

power saving state if it does not have packets pending for transmis-

sion and one of two conditions is met (1) it observes in the DTIM

beacon that there is no downlink traffic addressed to its TIM Group or

(2) it observes in the DTIM beacon that there is some downlink traffic

addressed to its TIM Group but that STA does not explicitly appear in

the TIM beacon Compared to the preliminary IEEE 80211 PS mecha-

nism this approach reduces the size of the Traffic Indication Map in

each TIM beacon thus reducing the overhead and the time STAs need

to listen and process them In addition TIM periods can be organised

in pages which further increases the number of TIM groups between

two DTIM beacons

The temporal organisation of pages DTIM and TIM periods is re-

ported in Fig 8 which shows the division of TIM periods in RAW (Re-

stricted Access Window) and PRAW (Periodic RAW) The RAW is a

time interval in each TIM period where TIM stations can transmit and

receive data (see below the different types of STAs defined in IEEE

80211ah) In addition it can be divided into several downlink and

uplink slots for further granularity In the downlink the slots are as-

signed to a single STA or a group of STAs while the uplink slots are

randomly selected by the STAs with packets ready for transmission

The PRAW is the period of time in each TIM where non-TIM stations

can transmit and receive data

Types of STAs IEEE 80211ah supports three types of STAs TIM

non-TIM and Unscheduled STAs A TIM station is assigned to a TIM

Group Their data transmissions must be performed within a RAW

(Restricted Access Window) period Non-TIM stations do not have to

listen to beacons to transmit data During the association process

non-TIM devices directly negotiate with the AP to obtain a trans-

mission time allocated in a PRAW The following channel access can

either be renegotiated or occurs periodically depending on the re-

quirements set by the STA Unscheduled STAs do not need to listen to

any beacons similar to non-TIM stations Even inside any restricted

access window unscheduled STAs can send a poll frame to the AP

sking for immediate access to the channel The response frame in-

icates an interval (outside both restricted access windows) during

hich unscheduled stations can access the channel This procedure

s meant for STAs that transmit data very sporadically

Hierarchical station organisation To support a large number of STAs

nd their organisation in pages DTIM and TIM periods IEEE 80211ah

ssigns to each associated STA a unique identifier of 13 bits which is

alled the Association Identifier (AID) Using this new AID the maxi-

um number of supported STAs is increased from the original 2007

n IEEE 80211 to 8191 (= 213 minus 1) in IEEE 80211ah However it also

llows categorising STAs according to the type of application they are

xecuting their power level or even their desired QoS by assigning

hem to different TIM groups

Long sleeping periods IEEE 80211ah offers TIM Non-TIM and Un-

cheduled STAs the possibility to set very long doze times (up to

onths) The corresponding clock drift produced by such long doze

imes must be taken into consideration however as the higher the

ime an STA has been asleep the further in advance it should wake

p to avoid possible synchronisation problems with the network

Efficient small data transmission Three new enhancements have

een proposed to reduce the overhead when the data packet size

s small First while IEEE 80211 contains a 28-byte MAC header

EEE 80211ah proposes a short 18-byte version by using AIDs in-

tead of MAC addresses Second IEEE 80211 has defined several null

ata packet (NDP) frames which consist only of a PHY header These

rames can be used to create short ACKs short Block ACKs short CTSs

nd short PS-Polls Finally a Fast Frame Exchange mechanism has

een developed so that if an STA has data to transmit it can notify a

uccessful reception by transmitting its data frame instead of an ACK

Sectorisation Since the PHY layer is based on the IEEE 80211ac

mendment single and multi-user beam-forming are also supported

y IEEE 80211ah This allows the transmission of data to multiple

TAs simultaneously in the downlink increasing the system capacity

he use of the beam-forming capability of IEEE 80211ah APs is also

onsidered as a means to group STAs into different independent an-

enna sectors with the main goal of reducing interference issues This

ould be particularly useful in the case of overlapping with other

EEE 80211ah WLANs or in the presence of hidden nodes

12 Open challenges

A first open challenge is to understand the coverage and achiev-

ble transmission rates in IEEE 80211ah WLANs in both indoor

nd outdoor scenarios Propagation models for WLANs working at

B Bellalta et al Computer Communications 75 (2016) 1ndash25 13

f

c

8

p

R

p

i

m

l

a

m

c

p

p

f

m

t

a

e

p

t

m

a

l

T

m

a

i

n

t

t

t

u

p

c

w

f

a

t

8

a

m

a

s

l

a

m

b

m

d

S

i

a

o

a

R

t

n

G

fi

s

t

p

t

c

u

c

w

i

o

p

g

a

c

i

f

c

(

d

m

l

i

R

t

l

t

L

m

a

m

o

e

f

a

I

i

t

w

i

t

a

c

w

p

f

t

d

5

i

s

e

s

w

5

a

c

a

n

requencies lower than 1 GHz are evaluated in [115] The authors

ompare two path loss propagation models proposed by the IEEE

0211ah Task Group (one for macro and one for picohotzone de-

loyments) [116] with Lee and HatandashOkumura propagation models

esults show that the IEEE 80211ah channel models underestimate

ath loss with respect to Lee and Hata models Moreover in a compar-

son with empirical data it is observed that the IEEE 80211ah channel

odels also underestimate the initial loss and the slope of the path-

oss function A new model parameterisation is thus proposed by the

uthors In [117] the feasibility of an IEEE 80211ah WLAN deploy-

ent is also evaluated in terms of the achievable range and bit rate

omputed on the basis of the link budget using the same path loss

ropagation models as in [115] Results show that the transmission

ower limitations in the uplink can limit the overall network per-

ormance Finally in [118] the authors evaluate the achievable trans-

ission range for the different transmission rates and the achievable

hroughput for different combination of transmission power values

nd transmission rates Further studies are required due to the het-

rogeneity of scenarios in which IEEE 80211ah WLANs can be de-

loyed to characterise the effective coverage ranges with special at-

ention to mixed indoors and outdoors scenarios and in presence of

obile nodes

Due to the wide coverage IEEE 80211ah WLANs may be severely

ffected by the presence of hidden terminals To mitigate such a prob-

em smart solutions to guarantee that the STAs assigned to the same

IM Group are inside each other transmission range or the employ-

ent of RTSCTS-based solutions may contribute to mitigate such

problem An overview of long-range scenarios for IEEE 80211ah

s presented in [119] including a detailed description of the hidden

ode problem in them

Single-hop uplink transmissions from distant STAs require high

ransmission power to reach the AP Since not all STAs may be able

o transmit at the required power a solution could be the introduc-

ion of nodes able to relay transmissions from them However the

se of relays has to be efficiently harmonised with the operation in

ower saving mode allowing temporal periods in which the relays

an gather the data from their associated STAs and then periods in

hich they forward the data to the AP which is still an open challenge

or IEEE 80211ah WLANs In case relays are used data prediction and

ggregation techniques could be implemented to make more efficient

ransmissions [120]

Another challenge is the assessment of the efficiency of IEEE

0211ah PS mechanisms including their optimisation In [111] the

uthors evaluate the capacity of the TIM and page segmentation

echanisms in terms of the maximum number of nodes supported

nd the energy consumed given a certain network traffic profile Re-

ults confirm that a large number of STAs can be supported with

ow energy consumption The impact of the number of nodes oper-

ting in PS mode on the energy consumption and the delay perfor-

ance is analysed in [121] A detailed analysis of the TIM Group-

ased channel access adopted by the IEEE 80211ah task group is

ade in [122] where the STAs assigned to each TIM are uniformly

istributed between the slots of the RAW Since only one group of

TAs is allowed to transmit in each RAW slot the channel contention

s minimised The number of downlink and uplink slots in a RAW

s well as their duration is optimised on the basis of the number

f active STAs and the network traffic profile in [123] In [124] the

uthors propose an algorithm to determine the optimal size of the

AW interval for uplink transmissions based on the estimation of

he number of STAs transmitting and the duration of the RAW Fi-

ally in [125] the authors address the contention problems in Smart

rid communication networks with many nodes and periodic traf-

c when using an IEEE 80211ah WLAN It is still a challenge to con-

ider smart systems able to adapt the IEEE 80211ah parameters to

he instantaneous system state in order to save as much energy as

ossible

Since the PS mechanisms affect only the operation of TIM STAs

he performance of non-TIM and Unscheduled STAs has not yet been

onsidered in the literature To identify the scenarios in which the

se of non-TIM and Unscheduled STAs are of interest is still an open

hallenge as further investigating how non-TIM STAs negotiate over

hich PRAWs they can use to transmit and receive data Co-existence

ssues between the three types of STAs in a single network is also an

pen challenge that requires further work to guarantee a minimum

erformance level for all of them

The design of strategies to distribute the STAs between all the TIM

roups based on their specific traffic profile position battery level

nd application priority is another important challenge that is still

ompletely open EDCA is the default channel access scheme included

n the IEEE 80211ah amendment and provides some basic traffic dif-

erentiation capabilities at the packet level Besides the different ac-

ess categories (ACs) needing to be renamed and their parameters

Arbitration Inter-Frame Spacing or AIFS CWmin TXOP duration) up-

ated to fit the specific traffic profiles of M2M communications other

echanisms can be implemented A mechanism to assign the down-

ink RAW slots when packets from multiple priority levels are wait-

ng for transmission at the AP is required for example Also uplink

AW slots are currently selected randomly by the STAs that want to

ransmit a packet A mechanism to reserve some slots at each up-

ink RAW for high priority STAs may therefore be an option although

his may severely degrade the performance of the low priority STAs

astly different repetition patterns for different priority TIM groups

ay also allow priority STAs assigned to those priority TIM groups to

ccess the channel more often We expect this challenge will receive

uch attention in the upcoming years because of the heterogeneity

f sensors that will be connected though a single IEEE 80211ah AP

specially in urban scenarios

Similar to what has been discussed for 4G cellular networks used

or M2M communications [126] the development of mechanisms to

void and resolve congestion situations is another open challenge for

EEE 80211ah WLANs with many nodes For instance a mechanism

n which STAs have to wait a random number of DTIM periods before

hey can start a transmission may be implemented This approach

ould effectively distribute the traffic over a longer period of time

n overload conditions but otherwise would unnecessarily increase

he access delay Adaptive and load-aware solutions may play again

n important role to keep the optimal WLAN operation in all possible

ircumstances

Finally since IEEE 80211ah will compete with 4G5G cellular net-

orks and WSNs to provide M2M connectivity in many different ap-

lication domains such as smart cities or e-health comparative per-

ormance studies to determine the strong and weak points of each

echnology are required including also aspects such as the cost of the

evices and the system reliability

Cognitive radio technology for TV White Spaces

This section gives an overview of the IEEE 80211af amendment

ntroducing its most relevant features and open challenges Fig 9

hows the basic components and features considered for WLANs op-

rating in the TV White Spaces (TVWS) which include local spectrum

ensing by the AP and STAs and the use of geolocation data bases

ith information on channel availability

1 The IEEE 80211af amendment

Thanks to the transition from analog TV to digital TV several VHF

nd UHF spectrum channels used for decades for analog TV broad-

asting are now unused or are under-utilised in many geographical

reas [127ndash129] This ldquodigital dividendrdquo of spectrum has suggested

ational regulators such as the FCC in the US and Ofcom in the UK

14 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 9 WLAN operating in TVWS basic elements and functionalities

Fig 10 Hidden terminal problem in TVWS when the secondary network uses only in-

stantaneous channel sensing to decide whether or not a TV channel is occupied When

the STA is not able to detect the TV signal due to the presence of obstacles it may

initiate a transmission and create interference to the primary users

s

c

5

a

i

p

t

o

o

t

e

c

u

l

m

b

a

m

t

e

t

t

t

m

D

c

I

n

b

u

8

c

a

t

i

t

w

f

to discuss how to reuse these channels for unlicensed devicesrsquo com-

munications [130131] These potentially vacant channels in the VHF

and UHF bands are referred to as TVWS [132] and include spectrum

portions like the 470ndash790 MHz in Europe and non-contiguous 54ndash72

76ndash88 MHz 174ndash216 MHz 470ndash698 MHz and 698ndash806 MHz in USA

A snapshot of the TV spectrum occupancy in the city of Barcelona in

2012 is shown in [133]

The attractive characteristics of TVWS (not only for WLANs) in-

clude the ability to penetrate through walls and other obstacles much

more effectively than other widely used spectrum bands such as the

24 and 57 GHz ISM bands [134ndash137] This fact along with a progres-

sive spectrum scarcity in ISM bands has suggested the birth of the

IEEE 80211af amendment published in February 2014 [138] which

provides the IEEE 80211 operational characteristics for TVWS access

of unlicensed White Space Devices (WSD) A good summary of IEEE

80211af can be found in [137] The main advantage of operating IEEE

80211 WLANs in the TVWS comes from an increased coverage range

which can reach up to 1 km in rural areas and open fields [139]

and less energy needed to transmit However this comes at the cost

of an increased interference risk to other WSDs which creates co-

existence problems and demands new PHY and MAC layers to ef-

ficiently support channel access and operations preserving licensed

usersrsquo devices

WLANs operated in the TVWS could cover a number of interest-

ing and emerging use cases and scenarios eg Internet access in ru-

ral or sparsely populated areas Smart Grid sensor aggregation me-

tering and control Internet of Things advanced WLAN operations

and TVWS traffic offloading in indoor environments [140141] Nev-

ertheless the concept could be realised to create or extend com-

mercial or municipality WiFi services offered to citizens with cover-

age at the whole city level realised with reasonably limited network

infrastructures [142143]

IEEE 80211af will use the PHY and MAC layers derived from IEEE

80211ac It will adopt concepts such as the OFDM multi-user beam-

forming contiguous and non-contiguous channel bonding and packet

aggregation Among the mandatory and innovative behavioural and

operational parameters the most notable one is the channel acqui-

sition support realised through remote geolocation-based spectrum

allocation databases which maintain the channelsrsquo availability infor-

mation in any given area and time of day providing upon request the

list of free channels available for use

In the following section we highlight and summarise three main

novelties that IEEE 80211af introduces In Section 511 we explain

how the access infrastructure to the remote spectrum database is de-

signed and what the requirements are we discuss the coexistence is-

sues and methodologies adopted in IEEE 80211af and we also discuss

the novel concept of non-contiguous channel bonding The following

ections also contain an illustration of related works Open research

hallenges will be summarised in Section 512

11 Novel features

This section provides an illustration of novel features introduced

nd discussed in the path to IEEE 80211af and the directions provided

n related works and standardisation initiatives to resolve classical

roblems properly characterised in the new framework of TVWS

echnologies

Channel acquisition spectrum database and channel sensing Much

f the attention in operating in the TVWS is given to the protection

f the primary licensed users in the TVWS spectrum band In general

he primary users were considered as the Digital TV (DTV) broadcast-

rs and receivers In fact receivers would suffer the wasteful effect of

ollisions during the TV broadcast reception in the case of secondary

sersrsquo (SU) transmission interference (Fig 10)

It is a well-known problem that the definition of a coherent wire-

ess channel status (idlebusy) from the viewpoint of distributed pri-

ary users (eg DTV receivers) is made difficult by the difference

etween transmitters and receivers interpretation of signals in a vari-

ble time-space collision domain resulting in hiddenexposed ter-

inals concepts Solutions proposed to reduce the hiddenexposed

erminals included the RTSCTS mechanisms and beaconing How-

ver it is difficult to implement solutions that would effectively de-

ermine the channel status of TVWS channels based on localised dis-

ributed sensing activities performed by multiple secondary users in

he same collision domain On the other hand the use of RTSCTS

echanism would not be effective in this case because most of the

TV primary users (and specifically the receivers) were not originally

onceived to transmit signals or implement the RTSCTS handshake

n conclusion the actions to identify free (and busy) TVWS chan-

els should not be based on primary usersrsquo involvement but should

e almost totally implemented by specifically designed secondary

sersrsquo methodologies

There has been much debate on the methodology that IEEE

0211af secondary devices should adopt to get knowledge of free

hannels at a given time on a given target communication area in

n effort to guarantee verified optimised and reliable TVWS spec-

rum use The three main methodologies discussed include sens-

ng solutions geolocation databases (DB) and beaconing Moreover

wo approaches are possible in general distributed and centralised

ith their well known tradeoffs in terms of effectiveness required in-

rastructure deployment and coordination overhead The distributed

B Bellalta et al Computer Communications 75 (2016) 1ndash25 15

m

b

s

c

t

u

a

t

o

l

t

w

t

t

a

o

c

c

a

(

t

o

t

t

D

t

t

o

t

f

a

t

c

i

a

c

d

t

s

t

b

I

l

(

e

a

i

R

p

e

t

c

c

i

b

i

i

t

b

t

t

t

W

a

p

c

i

t

h

p

t

n

a

t

e

t

t

t

p

t

a

r

8

w

s

w

[

i

S

o

a

t

f

i

8

i

r

a

a

i

t

c

(

I

s

a

e

t

t

c

[

i

c

i

g

s

n

s

t

i

o

o

s

c

D

o

ethodology that is receiving more attention in the literature is

ased on distributed channel sensing solutions [127144145] These

olutions include the observation of spectrum use and possible

ooperative aggregation of spectrum sensing information to increase

he accuracy of detection and reduce the vulnerability of primary

sersrsquo transmissions The spectrum sensing capability is required in

ll the secondary devices with a minus114 dBm sensitivity Dynamic

ransmission power control must be provided and the upper limit

n emissions is 100 mW EIRP (20 dBm) for portable devices (further

imited in the case of adjacent channel use to reduce out-of-band in-

erference) However database (DB) coordination of TVWS combined

ith spectrum sensing is considered the most promising and effec-

ive technique compared with spectrum sensing alone The DB spec-

rum information has more chances to be effective (ie identifying

ll free channels opportunities) and reliable (ie not prone to attacks

r erroneous interpretation of channels status) In fact regulators de-

ided to push for the conservative solution of a remotely accessible

entralised spectrum DB that maintains the information on the avail-

bility of all TVWS channels at any given point in time and location

with a target accuracy around 50 m) under their responsibility For

his reason the spectrum databases have become a mandatory part

f many spectrum sharing systems and the dominant technical solu-

ion to support TV white spaces

The IEEE 80211af standard adheres to this vision and also has

he role of a common regulatory framework for different spectrum

B implementations defining a generalised coordination architec-

ure protocols and interfaces for spectrum queries and local spec-

rum information management The protocol explicitly considers out

f scope the communication protocol and technology adopted in

he background leaving the freedom to those players responsible

or the deployment to select any suitable Internet-based and local

ccess network technology Secondary devices that need to access

he TVWS spectrum should get information on which channels they

an effectively use and which parameters to adopt without impact-

ng primary usersrsquo transmissions (and receptions) by querying the

vailable spectrum DB The query must contain the transmission

haracteristics and accurate geographical position of the secondary

evice which can be determined by means of a geo-positioning sys-

em (GPS) regularly updated in the case of movement (or manually

et for static devices) The IEEE 80211af reference system architec-

ure for DB spectrum access is composed of multiple Geolocation-

ased Spectrum information Databases (GDBs) entities connected via

nternet to Registered Location Secure Servers (RLSS) RLSSs work as

ocal proxies of the GDB for a localised group of Basic Service Sets

BSS) RLSS are connected via a secure protocol architecture with the

quivalent of multiple Access Points (AP) in different BSS realising

trusted DB infrastructure APs locally coordinate the exchange of

nformation and channel access management between the GDB (via

LSS) and the secondary usersrsquo end stations (STA)

Mode I devices are those under the control of a device that em-

loys geo-location database access while Mode II devices are those

mploying geo-location database access by themselves The informa-

ion exchange provided between the GDB and a secondary user STA

an be provided in both an open-loop (eg adopted by FCC) and a

losed-loop (eg adopted by ETSI) implementation In an open loop

mplementation daily spectrum availability information is provided

y the GDB and no feedback on the spectrum information received

s provided by STAs thus the system approach for spectrum access

s much more conservative leading to low channelsrsquo utilisation po-

ential In a closed loop implementation the STA can provide feed-

ack to the GDB and there are more communication overheads due

o the high granularity of updates however the system is more effec-

ive and reliable in the exploitation of TVWS on behalf of STAs The

ypical information provided by the GDB includes (1) the updated

hite Space Maps (WSM) of frequencies allowed for secondary use

t the timespace of the querying STA and (2) the device-dependent

ower limitations for transmission (in general conservative and ac-

urate enough to avoid relevant interference effects on primary users

dentified in the area)

The basic IEEE 80211af mechanisms regulating the communica-

ion between STA and GDB can be found in [137138]

In parallel with IEEE 80211af standards such as the IEEE 19006

ave been created that consider interfaces and data structures sup-

orting spectrum sensing information exchange applicable to spec-

rum sensing (and particularly distributed spectrum sensing) sce-

arios In particular in October 2014 the IEEE 19006 WG initiated

project for a new standard called IEEE 19006b [146] concerning

he use of spectrum sensing information to support and optimise the

ffectiveness reliability and robustness of spectrum database solu-

ions The aim is to enhance the performance and capabilities of spec-

rum databases through the use of spectrum sensing information

Co-existence Support for co-existence mechanisms so that mul-

iple technologies can effectively utilise the TVWS spectrum is im-

ortant Self co-existence between network devices of a common

echnology (eg deployed by different operators in the same area)

nd co-existence among different technologies are relevant topic of

esearch for Cognitive Radio (CR) systems and specifically for IEEE

0211af Many solutions appeared on the research scene but no one

as so far finalised as the target solution for IEEE 80211af Specific

tandardisation has been started to regulate coexistence between

ireless standards of unlicensed devices including the IEEE 802191

147] The purpose of the IEEE 802191 standard is to enable the fam-

ly of IEEE 802 Wireless Standards to most effectively use TV White

pace by providing standard coexistence methods among dissimilar

r independently operated TVWS devices Early examples of gener-

lised coexistence mechanisms included Dynamic Frequency Selec-

ion (DFS) Transmission Power Control (TPC) listen before talk (eg

or contention based IEEE 80211 80215) time division multiplex-

ng (also among different techniques such as the IEEE 80216 80220

0222) and Message-based Spectrum Contention (that is beacon-

ng messages that carry coexistence information) Opportune met-

ics must be defined to assess the measurable coexistence achieved

mong different technologies as an example the hidden node prob-

bility for a target scenario or the estimate of percentage variation

n normalised network throughput and latency (before and during

he SU transmissions) On the other hand a centralised coexistence

ontrol mechanism could be effectively realised by a central manager

or coexistence-DB like the GDB) in critical scenarios To this end

EEE 19004 [148] (a standard for heterogeneous networks in dynamic

pectrum context part of IEEE Standards Coordinating Committee 41)

ims to standardise the overall system architecture and information

xchange between the network and mobile devices which will allow

hese elements to optimally choose from available radio resources

There are three possible classification of co-existence archi-

ectures and inter-network coordination channels for CR systems

entralised coordinated and autonomous co-existence mechanisms

149] In centralised co-existence schemes the co-existence is admin-

stered by a central entity (eg like in IEEE 802191) This solution

ould be applied in both homogeneous IEEE 80211af systems (eg

ncarnated by the centralised spectrum DB entity) and in hetero-

eneous inter-networks without requiring modifications to existing

tandards under the assumption that all the involved co-existing

etwork entities would adhere to the same centralised coordination

cheme and inter-network coordination protocol On the other hand

his assumption could be hard to satisfy in some practical scenar-

os The centralised schemes are effective in providing minimisation

f inter-network interference based on the availability and quality

f centralised co-existence information In coordinated co-existence

chemes the centralised entity could be present but not taking de-

isions in general The central entity could provide a co-existence

B with required information which can be used via in-band and

ut-band signalling for co-existence decisions taken by the cluster

16 B Bellalta et al Computer Communications 75 (2016) 1ndash25

c

e

n

c

o

s

i

d

a

S

8

n

a

b

o

b

i

t

a

8

b

[

t

i

n

b

5

t

a

c

b

s

a

c

n

s

p

i

c

l

r

c

f

a

v

t

a

e

c

t

Q

a

b

s

i

c

u

n

c

i

heads of many co-existing inter-networks for implementing a proper

common coordination protocol (eg out-band busy tone signalling)

These are often hybrid solutions realising a compromise between ef-

fectiveness and suitability In autonomous co-existence schemes all the

decisions and coordination are implemented in a distributed way

by the involved network entities In these schemes many policies

can be adopted to realise a sufficient level of co-existence under a

best effort approach Solutions that could be implemented include

the use of busy tones beaconing and other signalling protocols in

dedicated control channels dynamic distributed frequency selection

schemes listen before transmit token-based and dynamic reserva-

tion schemes All the above mentioned solutions have goods and

bads and a general illustration can be found in [149] Other proac-

tive co-existence techniques try to early detect and recovermitigate

co-existence issues with relaxed inter-network coordination eg re-

alised via spectrum sensing and interference avoidancesuppression

techniques

It must be clear that different aims exist regarding the co-

existence for primary TVWS users protection against secondary

users and for secondary devices mutual interference avoidance

Many co-existence problems have been analysed in the literature

for TVWS technologies in particular between the primary (licensed)

and secondary (unlicensed) devices The coexistence management in

TVWS is complex because primary users of DVB transmissions are in-

tended to be the pure receivers rather than the broadcasters In other

words potential protected users could be everywhere and the hid-

den terminal problem could arise for these systems The co-existence

problem is also complicated by many factors that cause asymme-

try and dynamicity in the TVWS such as the mobility variable den-

sity power asymmetry and heterogeneous MACPHY layers TVWS

co-existence problems originate between the IEEE 80211af technolo-

gies and the IEEE 80222 technologies for wireless regional area net-

works (WRAN) Results of analysis of TVWS usage in Europe show

that white spaces are typically present and fragmented They are typ-

ically more abundant in rural areas where larger contiguous blocks

of unused channels are available due to broadcast network planning

giving priorities linked to population density However the exploita-

tion of TVWS in urban areas is possible eg some recent research

was realised under the assumption of exploiting shadowing effects

created in the communication environment (eg by buildings obsta-

cles) in favour of frequency reuse [150]

Since different standards for opportunistic communication in the

TV White Spaces have now been published such as IEEE 80222 [151]

IEEE 802154m [152] Weightless [153] and of course IEEE 80211af

improved methods to guarantee the coexistence of different devices

operating on several protocols in the same bands must be deployed

We highlight here some work eg [150154ndash157] that focuses on

the coexistence between different technologies operating in the same

bands In particular [155] studies the performance degradation of

IEEE 80222 when an IEEE 80211af network is operated in the same

area The problem becomes even worse when the IEEE 80211af net-

work is located near 80222 user equipment causing both networks

to perceive a strong interference from each other One proposed solu-

tion is the Coexistence Beacon Protocol [158] studied for 80222 net-

works which foresees the exchange of a periodic beacon to identify

the neighbouring and possibly interfering networks

For what concerns the co-existence problems research has ad-

dressed other important related aspects of new WLANs on TV White

Spaces For instance the higher coverage range makes them attrac-

tive for applications in the smart grid and also for M2M on TVWS

[134150159] Here the increased range compared to standard tech-

nologies in the ISM bands can have beneficial effects for indoor

mobile devices [150] and M2M [159] Specifically indoor devices

communicating on TV White Spaces offer better propagation char-

acteristics and penetration through obstacles making it easier to

realise scalable home connectivity [135] although interference and

o-existence problems could be exacerbated and must be resolved

g see [139] However studies in the literature show how the sig-

al coming from an indoor transmitter on TV White Spaces remains

onstrained inside the house making it difficult to deploy indoor-to-

utdoor networks On the other hand this shadowing limit can be

een as an opportunity lowering the interference outside the build-

ng in which the TVWS network is locally operated [134150] Un-

er this approach HDTV streaming has recently been considered as

possible use case of WLAN networks operating in the TV White

pace [160]

Non-contiguous channel bonding A relevant novel feature of IEEE

0211af is the potential for contiguous and non-contiguous chan-

el bonding which permits aggregating basic channels (also non-

djacent ones) and leveraging the possible large frequency spread

etween multiple available channels Due to the rather static nature

f primary DTV transmissions where a busy channel is unlikely to

ecome free in the near future and state changes are coarse-grained

n general it is crucial to exploit the time-locality effect and exploit

he maximum physical channel availability that could be aggregated

t any given location With the methodology inherited from IEEE

0211ac IEEE 80211af is capable of bonding together two up to four

asic channels grouped in up to two different non-contiguous chunks

137] The spectrum bandwidth of a DVB-T basic channel can be ei-

her 6 7 or 8 MHz depending on the country in which the service

s operated As an example this creates a 144 to 168 OFDM chan-

elsrsquo bandwidth potential when up to four 6ndash7ndash8 MHz channels are

onded [137]

12 Open challenges

The deployment of WLANs in the TVWS at their maximum poten-

ial still requires the resolution of open problems Generally speaking

s mentioned in previous sections the challenges for such networks

an be divided into three main categories (1) spectrum sharing

etween opportunistic (secondary) and heterogeneous devices (2)

pectrum sensingmanagement and maximum exploitation of avail-

ble spectrum (potentially enabling the spectrum-on-demand con-

ept) and (3) co-existence and interference mitigation to the primary

etwork (primary user protection) and secondary networks [154]

The overall aim of spectrum sharing techniques is to maximise

eparation to avoid overlapping or contiguous channels operations

rovided that sufficient channels are available in the TVWS spectrum

n a given spacetime scenario [154] In general important research

ontributions have to be realised for the design of proper spectrum al-

ocation methodologies and techniques satisfying multi-factorial QoS

equirements at the system-level and the user-level Novel ideas in-

lude the spectrum sharing and spectrum sensing techniques being

urther divided into cooperative and non-cooperative

Non-cooperative solutions attempt to realise spectrum sharing

nd sensing on a local basis without direct cooperation between de-

ices Examples include the ldquolisten-before-transmitrdquo approach and

ransmission power control to limit the interference and spectrum

llocation policies that are based on local node feedback only Coop-

rative solutions rely on the existence of a common communication

hannel through which devices can tentatively agree on the spec-

rum allocation that provides the desired spectrum separation and

oS Major issues include the definition of a methodology to identify

common channel (or multiple channels mutually shared in space

y pairs of heterogeneous and distributed devices) and to efficiently

hare common information For the access to the spectrum DBs the

dentification of proper common access channels is still under dis-

ussion out-of-TVWS-band cellular communications can be widely

sed given their high coverage also indoor

Trying to avoid the use of complex and resource-hungry coordi-

ation schemes and dedicated control channels rendezvous proto-

ols play a significant role both for co-existence and for establish-

ng dynamic communication channels based on spectrum sharing

B Bellalta et al Computer Communications 75 (2016) 1ndash25 17

R

n

c

t

l

o

m

e

n

i

n

w

n

a

u

s

p

b

s

m

s

a

d

t

b

a

a

m

a

t

u

8

n

t

d

d

m

f

t

n

d

n

e

r

a

h

b

f

a

c

t

u

i

f

t

d

t

s

i

f

I

l

i

a

n

c

d

I

n

o

t

i

t

e

w

t

c

m

t

a

i

t

q

b

c

a

i

l

i

a

v

t

a

a

c

t

i

t

b

t

t

a

D

p

c

s

o

q

r

p

a

g

t

e

u

d

t

m

n

t

s

f

i

1

8

f

T

endezvous is considered a fundamental problem in generalised CR

etworks at the basis of many communication processes eg in-

luding neighbour discovery routing and broadcast Rendezvous pro-

ocols attempt to establish a new link for communication on a se-

ected and agreed frequency band (channel) identified within a set

f available resources (channels) thus creating the basis for com-

unication between two or more SUs This problem is often exac-

rbated by the high (and unknown) numbers of contending SUs high

umber of available channels and their ambiguous boundaries and

dentifiers and tough co-existence requirements satisfaction in cog-

itive radio systems [161] Preliminary approaches for rendezvous

ere based on the existence of a dedicated Common Control Chan-

el (CCC) agreed among all the SUs or the existence of a centralised

gent (decision maker) assigning channels to any pairs of requesting

sers This approach could be considered in the case of IEEE 80211af

olutions based on the spectrum DB implementation however the

roblem is complicated by the fact that the common channel must

e identified and made available for all the requesting SUs at the

ame time even when these are spread over multiple collision do-

ains in space On the other hand the scalability and reliability is-

ues which are caused by the congestion of CCCs and the vulner-

bility risks of centralised DBs motivate the research of alternative

istributed solutions called blind rendezvous protocols [162] Dis-

ributed solutions have been mainly based on beaconing andor slot-

ased channel hopping algorithms exploiting randomisation sensing

nd discoverysynchronisation messages spread in slotted time over

set of candidate available channels [161] These protocols aim to

inimise the convergence time to maximise the use of resources

nd to avoid collisions among SUs by identifying and agreeing on

he use of a given common channel among N non-overlapping and

nambiguously labelled channels The main complication in IEEE

0211af systems is given by the wide availability of many different

on-contiguous spectrum bands which cannot be uniformly quan-

ised in single channels and unambiguously labelled in space in or-

er to realise a common reference domain for all the IEEE 80211af

evices In fact the knowledge of the number of channels their com-

on labels the unique IDs of SUs and the number of SUs contending

or the channel assignment are the key factors determining the effec-

iveness and good properties of blind rendezvous protocols Recently

ew classes of rendezvous protocols have been proposed which are

istributed blind and oblivious Oblivious means that available chan-

els may be identified with different labels on behalf of the differ-

nt involved SUs A number of these distributed blind and oblivious

endezvous protocols have been proposed and have been referenced

nd analysed in [162] To conclude solutions for dynamic spectrum

andoff issues (eg in case PUs suddenly appear in a region) must

e resolved in advance to minimise the latency of re-establishing a

unctional communication channel eg by implementing a proactive

lternative rendezvous channel selection in background to ongoing

ommunication processes [163]

Another challenge is to provide cooperative techniques in a dis-

ributed vs centralised implementation by analysing their convergence

nder variable conditions overheads and tradeoffs In some cases as

n vehicular networks the exploitation of scenario characteristics and

actors such as the constrained mobility could help to realise effec-

ive dissemination of sensing information over an extended and pre-

ictable horizon under a cooperative approach [164ndash166] Coopera-

ion could be helpful in the realisation of the cooperative spectrum

ensing in combination with the Geolocation DB approach Another

nteresting direction of research is the investigation of mutual ef-

ects of coexisting cooperation-based vs non-cooperative techniques

n cooperation-based devices the exchange of information could al-

ow a quick convergence to a solution and avoidance of further

nterference [145] Certainly cooperative techniques have desirable

dvantages over non-cooperative ones However tight time synchro-

isation and high coordination overheads are required The common

hannel for cooperation purposes could become a bottleneck and re-

uce the potential advantages of dynamic spectrum allocation The

EEE 80219 foresees the presence of a shared common control chan-

el (CCC) to which devices need to tune in order to gather the state

f the network Solutions such as [141] propose the use of TVWS for

he CCC

A promising direction for research is the adoption of cluster-

ng schemes for secondary devices associated with a hybrid dis-

ributedcooperative vs centralisednon-cooperative approach in an

ffort to reduce overheads while maximising advantages In this

ay secondary nodes could implement cooperation for sensing func-

ions using properly identified common channels and delegate the

entralised decisions and spectrum-DB management to well instru-

ented cluster-leader nodes This concept is incarnated to some ex-

ent by RLSS in IEEE 80211af More recently there have been propos-

ls to build distributed databases that refresh their contents period-

cally by querying the remote spectrum database [167] Here Mas-

er devices (according to Ofcom terminology) periodically cache the

uery replies from the spectrum database in order to reply to and

roadcast the spectrum availability to neighbour Slave devices

Another challenging direction for research is the definition of ac-

urate models and efficient simulation tools enabling dynamic spectrum

nalysis for both rural and urban areas Recent developments of dig-

tal maps simulation tools and propagation models theoretically al-

ow improving the capability of predicting radio propagation effects

n complex scenarios with an acceptable computation time This en-

bling technology could be used in parallel to spectrum DBs to pro-

ide more accurate forecasting of dynamic frequency allocation in

imespace scenarios Another issue to be taken into account is the

ccuracy of the remote spectrum database regarding the channel

vailability estimation Several works have already shown the inac-

uracy of propagation models due to the complexity of parameters

o be considered and differences between modelled and real scenar-

os [132134168169] Better propagation models will make it possible

o more efficiently estimate the interference between devices and

uild Radio Environment Maps (REM) or White Space Maps (WSM)

hat will possibly lead to a more accurate sharing of the radio spec-

rum [144] To this end current research could focus on how to lever-

ge simulation models in order to build more accurate frequency

Bs

In general the adoption of remote spectrum DBs methodology

ushed by national regulators has made the spectrum sensing pro-

ess optional and conservatively realised (eg power control for

ensing-only devices is limited to below 17 dBm EIRP) Thus much

f the research has focused on techniques and solutions to successfully

uery the remote database also in challenging environments such as

ural areas and indoor However since the queries must provide the

osition of the mobile device the accuracy of positioning for mobile

nd handheld devices still needs more appropriate solutions Trian-

ulation can be used to estimate the position of mobile devices with

he accuracy needed by the current regulations outdoor (50 m) How-

ver a low GPS accuracy is possible indoors [170] Certified man-

al positioning can be provided by technicians for static and indoor

evices although limiting mobility Alternatives to costly infrastruc-

ures for positioning (eg based on short range beaconing devices)

ust be deployed in indoor scenarios

Finally regarding spectrum sensing and interference mitigation

ew challenges come from the heterogeneity and different charac-

eristics that networks operating in the TV White Spaces can have

uch as different transmitting power (4 W for static devices 100 mW

or mobile and portable devices and 40 mW for communication

n channels adjacent to occupied ones) different bandwidths (5

0 15 and 20 MHz are currently foreseen for IEEE 80211af IEEE

02154m can have narrower bandwidths or wider ones) and dif-

erent medium access schemes (protocols that use either CSMA or

DMA are required to co-exist in the TV White Spaces) With all this

18 B Bellalta et al Computer Communications 75 (2016) 1ndash25

d

a

T

i

a

c

t

l

b

o

t

a

d

p

a

r

p

n

a

m

t

p

p

d

d

o

t

i

D

w

p

i

a

a

d

b

fi

d

o

t

l

b

t

b

a

i

e

o

o

c

t

t

w

A

c

M

s

c

l

O

w

g

v

t

in place the problem of determining whether another secondary de-

vice is currently transmitting in the same channel is still an open

issue

Spectrum sensing is a very hot topic in the literature with many

proposals that leverage the cooperation between devices to raise the

accuracy of the detection process eg [144171] The current litera-

ture also offers room for proposals that suggest exploiting sensing to

build more accurate spectrum DBs [164172173] Lastly regarding the

interference mitigation to the primary network regulators provide a

conservative approach based on a remote spectrum database to avoid

high risk for the primary users However in particular for the indoor

scenario the concept of secondary usersrsquo transmission in TV Grey

Spaces has been proposed TV Grey Space identifies busy channels

that are formally used by primary users in the area of interest (eg

at the rooftop level of a building for DVB-T) but which could be con-

sidered usable indoors (eg at the basement level) without causing

effective interference to the primary users [134150174] This concept

is based on the generalised spectrum underlay paradigm [175] The

remaining research challenge is to provide TV Grey space access guar-

anteeing a sufficient amount of protection to the primary user while

providing a satisfactory QoS to opportunistic devices eg through

the use of customised and validated propagation models for indoor

TVWS networks and feedback mechanisms to constantly monitor the

interference generated to the primary network

6 Emerging new trends and technologies

In this last section we review three emerging trends that in our

opinion will have a large influence in the conception of future WLANs

as they change the way WLAN protocols and functionalities are de-

veloped implemented tested and integrated with other wireless

networks

61 Programmable Wireless LANs

Especially in the enterprise environment WLAN deployments

need to support a wide range of functionalities and services This is

intrinsically difficult because of the large number of APs that must be

managed which calls for scalable solutions Typical services include

channel assignment load balancing among APs authentication au-

thorisation and accounting (AAA) policy management support for

client mobility and interference coordination Another problem is the

fact that WLAN clients autonomously take several decisions such as

which APs to associate with when to hand-over etc Therefore sup-

porting roaming clients requires the management of a large number

of association states across several APs which is a challenge if sup-

port for real-time hand-over is desired Typically such management

schemes are centralised and most of them are proprietary such as

WLAN controller solutions from Aruba [176] or Cisco [177] although

the IEEE 80211u amendment has been released to allow mobile users

to seamless roam between WiFi networks with automatic authenti-

cation and handoff [178] For example Dyson [179] enables STAs to

send information such as radio channel conditions to a centralised

controller based on a custom API (eg based on Python) As the con-

troller has a centralised view of the network it can enforce a rich

set of policies to control the network also using historical informa-

tion A demo system has been implemented along with applications

such as airtime reservations for specific clients or optimised hand-

offs However Dyson requires STAs to be modified in order to use

those new services offered by the centralised controller TRANTOR

[180] is another example of a centralised management system that

requires changes of clients in order for the infrastructure to gather

information from them in terms of eg interference measurements

In addition control commands enable the infrastructure to exercise

control such as modifying the transmit power or influencing the as-

sociation procedure Clients still use standard CSMA MAC layer for

ata transmission In contrast to DenseAP [181] or DIRAC [182] which

re also based on a centralised controller but do not modify clients

RANTOR exercises much more control on clients by the use of a ded-

cated API which enables a significantly higher gain in coordination

nd thus capacity CENTAUR [183] is another example of a centralised

ontroller that centrally schedules hidden and exposed terminals in

he downlink while using standard CSMA MAC for uplink traffic and

egacy downlink traffic To achieve good performance it uses a fixed

ack-off packet staggering techniques and a hybrid data path that

nly schedules downlink transmissions towards hidden and exposed

erminals centrally All other traffic is sent using standard DCF with

standard back-off procedure In contrast to previous approaches it

oes not require any modifications to the clients but does require data

lane centralisation and it remains questionable how scalable such

solution is for large WLAN deployments with high PHY layer data

ates beyond 100 Mbps However new approaches for fast data plane

rocessing that are currently explored in the virtualisation commu-

ity which move all the packet processing into user space may be

n option to significantly increase speed Designing such manage-

ent systems raises several interesting research questions (a) Cen-

ralised infrastructures may incur high latency but are more simple to

rogram and maintain while designing a completely distributed ap-

roach that operates close to WLAN devices may result in significant

istributed coordination and consistency problems (b) what level of

istributed control is required in order to support the flexibility needs

f future WiFi based networks supporting a high level of mobility

Traditionally WLAN APs are built on proprietary operating sys-

ems that are tightly coupled with the hardware This design makes

t hard to create new applications on top of such networking devices

espite the fact that protocols and mechanisms are available that

ould greatly improve the utility of existing networks those new

rotocols are not deployed because the closed system design makes

t very difficult to extend their functionality An important aspect for

ll such new approaches is therefore how to provide open interfaces

nd open source to speed up innovation As an example Linux-based

evices are completely open source but in order to increase flexi-

ility much more work is needed in the area of open drivers and

rmware In fact the Atheros based ath9k driver has been the main

river of innovation in the open source community because it is the

nly driver that interworks with open firmware In general solutions

hat modify the OS driver focus mainly on programmable network

evel solutions for WLANs such as channel switching or handovers

etween APs While the ath9k driver is an excellent example in how

o speed up innovation it is still very cumbersome to support flexi-

le MAC engine reprogrammability with ath9k In contrast MAClets

nd Wireless MAC processors (WMP) [184] allow a much more flex-

ble reprogramming of MAC functionalities A MAC processor is an

ntity able to execute general MAC commands that specify the MAC

perations through a software-defined state machine The behaviour

f the MAC protocol can therefore be updated at run-time by simply

hanging the sequence in which those commands are executed (ie

he MAClets) As a proof-of-concept the authors of [184] implement

he proposed MAC processor solution in a commodity WLAN hard-

are card extending the basic DCF in three directions piggy-backing

CKs a pseudo TDMA and the use of multiple channels In [185] a

ontrol framework for this WMP system is also proposed to support

AClet code mobility ie for moving loading and activating MAC

oftware programs embedded into ordinary data packets (akin to the

apsule model of traditional active networks)

The difficulty in re-programming networking hardware has also

ed to the concept of Software Defined Networks (SDNs) based on the

penFlow protocol [186] The main idea of SDNs is to extend net-

orking devices with standardised APIs that allow third-party pro-

rammers to flexibly control the data path In addition SDNs pro-

ide higher level abstractions to network designers and programmers

hrough the use of a centralised control plane offered by a network

B Bellalta et al Computer Communications 75 (2016) 1ndash25 19

Fig 11 The CLOUDMAC architecture VAP stands for Virtual Access Point and WT

stands for Wireless Termination

c

a

c

t

m

W

n

t

t

c

p

t

u

h

a

i

e

s

o

c

a

a

f

T

r

s

a

s

p

o

M

C

i

T

t

fl

A

a

a

i

c

t

s

t

e

I

t

w

H

e

b

I

l

d

o

6

r

W

s

l

a

e

e

a

s

b

c

t

(

S

p

c

F

B

t

p

T

e

t

s

u

m

i

e

o

a

A

p

d

t

t

n

p

i

t

t

f

i

ontroller such as NOX [187] which allows reuse of components such

s topology discovery or network access control for different appli-

ations The SDN concept has recently been applied to WLAN archi-

ecture in order to enable fine grained control over mobility manage-

ent and data forwarding focusing on a programmable enterprise

LAN architecture An important part of the SDN architecture is the

etwork controller which provides a centralised view of (parts of)

he SDN enabled network and uses the OpenFlow protocol to install

he forwarding rules on SDN-enabled devices (routers switches ac-

ess points cellular base stations etc)

ODIN [188] is designed to support programmability in enter-

rise WLAN architecture by separating the association state from

he physical AP They implement Light Virtual Access Points (LVAP)

sing a Split-MAC approach where the infrastructure controls the

andover procedure among different WLAN APs By managing

ssociations through SDN controllers ODIN enables proactive mobil-

ty management and load balancing within the SDN enabled WLAN

nterprise network without the need for changes in the client WLAN

tack or IEEE 80211 MAC layer While ODIN requires agents to reside

n the APs that communicate with the Odin Master within the SDN

ontroller CLOUDMAC [189] is based purely on the concept of SDN

nd virtual APs Similar to ODIN CLOUDMAC implements a Split-MAC

pproach but in addition enables the processing of WLAN MAC layer

rames within a co-located Cloud using so-called Virtual APs (VAPs)

he physical APs in CLOUDMAC are lightweight WLAN APs that are

esponsible only for sending out IEEE 80211 based MAC layer ACKs to

tandard WLAN clients and tunnel WLAN MAC layer frames through

n SDN-enabled enterprise WLAN towards the VAPs As association

tates are kept in the VAPs fast mobility is supported using sim-

le OpenFlow forwarding rules Because of the additional processing

f IEEE 80211 WLAN MAC frames in the co-located Cloud CLOUD-

AC has higher latency than a standard WLAN deployment However

LOUDMAC offers a Webservice based API to third party applications

n order to program the enterprise WLAN and enable new services

he architecture of CLOUDMAC is depicted in Fig 11 It has been ex-

ended in [190] to support QoS management and in [191] to support

exible MAC management frame prioritisation based on IEEE 80211

system based on OpenFlow has been recently proposed in [192] to

llow the station to be associated with multiple APs simultaneously

nd to switch between APs with low overhead

An important aspect to consider in order to enable programmabil-

ty is backwards compatibility Several approaches (eg [180]) require

lients to be modified to utilise the features provided This is difficult

o do in practice because it requires changes in the operating system

oftware of all clients If not all clients can utilise those APIs it is ques-

ionable how usable the new architecture will be and how much ben-

fit in terms of aggregate performance such architecture will allow

n contrast the SDN based approaches are interesting in the sense

hat they do not require changes in the client WLAN stack and work

ith the standard IEEE 80211 MAC layer deployed within the clients

owever it remains questionable how scalable such solutions are For

xample an interesting open research topic is to evaluate the scala-

ility of approaches such as CLOUDMAC [189] In addition processing

EEE 80211 MAC frames within co-located private Cloud requires low

atency support from local Cloud solutions (such as OpenStack) in or-

er to reduce the IEEE 80211 MAC processing time which is an area

f active research

2 Prototyping and testing IEEE 80211 enhancements

Most of the new proposals for next-generation WLANs are cur-

ently only evaluated using mathematical analysis and simulation

hile both analysis and simulation are necessary to characterise and

tudy those enhancements in the initial design phase or to consider

arge-scale scenarios it is difficult to consider all practical aspects of

real-world scenario This can sometimes cause significant differ-

nces between what simulations and real experiments show How-

ver real experiments are challenging because of the high complexity

nd costs of building the new hardware and software for each specific

olution to test

To mitigate implementation complexity there are several flexi-

le hardware platforms where low-level MAC mechanisms can be

ompletely implemented in software Examples based on FPGAs are

he USRP (Universal Software Radio Peripheral) [193] and the WARP

Wireless Open-Access Research Platform) [194] As an alternative

ORA (Microsoft Research Software Radio) [195] works on general

urpose computers by taking some advantage from modern multiple

ore systems

A different approach is provided by OpenFWWF [196] Open-

WWF provides an open CSMACA firmware for specific models of

roadcom chipsets so the resulting firmware can be uploaded and

ested in real commercial hardware OpenFWWF implements a sim-

le State Machine (SM) for controlling the hardware in real time

he SM evolution is driven by a main loop that reacts to events by

xecuting specific handlers When a packet originally prepared by

he Linux kernel is ready in the NIC memory handler Packet_Ready

ets up the radio hardware according to the packet metadata sched-

les the transmission and jumps back to the main loop The Trans-

ission Engine (TXE) then takes care of accessing the channel ie

t decrements the back-off counter according to the DCF rules and

ventually starts the actual transmission This triggers the execution

f the TX_frame_now event that prepares the ACK time-out clock

nd finalises the MAC header If the ACK-frame is received or if the

CK time-out expires and the maximum number of attempts for this

acket is reached handler Update_Params resets the contention win-

ow to its minimum value or otherwise doubles it Finally it loads

he back-off counter with a fresh value

In the following we will review a list of selected papers that use

he OpenFWWF firmware and the WARP or USRP platforms to test

ew proposals for WLANs in real scenarios as examples A first im-

lementation of the groupcast mechanisms defined in IEEE 80211aa

s presented in [197] The authors modify the OpenFWWF firmware

o include those functionalities and evaluate them using a 30 STAs

estbed A second example of the use of OpenFWWF is the collision-

ree MAC protocol presented in [198] In [198] the authors design and

mplement in OpenFWWF a MAC protocol that is able to achieve a

20 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 12 LTE to WIFI offloading

a

n

c

t

[

d

a

W

v

t

t

o

b

f

S

g

l

i

I

s

I

b

n

s

m

a

p

d

c

N

o

v

t

t

n

l

i

a

t

a

collision-free operation by waiting a deterministic time after success-

ful transmissions before to start a new transmission This work shows

the experimental performance of a collision-free MAC (CF-MAC) pro-

tocol for WLANs using commercial hardware The results show that

the proposed CF-MAC protocol leads to a better distribution of the

available bandwidth among users a higher throughput and lower

losses than DCF

In [199] the authors use the WARP platform to test a variant of the

DCF ndash called IEEE 80211ec ndash that substitutes control packets such as

the RTS CTS and ACK with short detectable sequences Since those se-

quences are shorter than the control packets and can be detected cor-

rectly even at lower SNRs values a significant gain in performance is

achieved MIDAS (Multiple-Input Distributed Antenna Systems) [200]

implementation using the WARP platform shows the benefits of dis-

tributing the antennas over the area to cover instead of co-locating

them at the AP in terms of capacity when MU-MIMO is employed

The authors also propose a new MAC protocol to benefit from the

spatial reuse that the DAS allows taking as a basis the IEEE 80211ac

Experimental results show that their proposal is able to achieve up

to 200 gains versus the traditional approach where all antennas are

co-located at the AP

The USRP platform has recently been proposed to develop a first

implementation of IEEE 80211ah in order to obtain a preliminary ex-

perimental performance assessment of such a technology because

no commercial off-the-shelf hardware is yet available as the amend-

ment is still in progress [201] The USRP platform is also used in

[202] to evaluate TIMO (Technology Independent Multi-Output) a so-

lution to deal with high-power non-IEEE 80211 interferers in ISM

bands

To conclude this section it is worth mentioning here also to

MAClets and WMP as given such flexibility to implement and dis-

tribute new MAC protocols and other IEEE 80211 functionalities we

believe that if a use-friendly implementation of such a MAC proces-

sors framework is provided it would significantly contribute to the

development and testing of new MAC enhancements by the research

community

63 CellularWLAN interworking

Public hotspots that offer Internet access over a WLAN using IEEE

80211 technology are now nearly ubiquitous It is forecasted that the

cumulative installed base of WiFi hotspots worldwide will amount

to 551 million by 2018 excluding private hotspots (eg WiFi access

points deployed at home) [203] The sharp increase in the availabil-

ity of public WiFi was initially perceived by mobile cellular operators

as a threat due to the additional competition from wireline Internet

service providers or emerging crowdsourced WiFi networks such as

FON5 However as cellular operators are fighting to cope with the ex-

plosion of mobile data traffic created by the rising use of multimedia

content traffic over mobile devices [15] they are also starting to use

WLANs based on the IEEE 80211 technology to offload data from their

core and access networks In general mobile data offloading refers to

the use of complementary network technologies (in licensed or un-

licensed spectrum) for delivering data originally targeted to cellular

networks Intuitively the simplest type of offloading consists of ex-

ploiting connectivity to existing co-located WiFi networks and trans-

ferring data without any delay (Fig 12) Thus this offloading tech-

nique is known as on-the-spot offloading As a consequence of this

new trend the seamless integration of cellular (eg 3GLTE) and WiFi

technologies has attracted significant research interest in recent years

(see [204] for a survey) a few solutions have already been standard-

ised [205206] and roaming between cellular and WiFi is becoming

increasingly transparent to end users CellularWLAN interworking is

5 httpscorpfoncomen

fl

t

f

lso fostered by the support in the evolving 4G standards of heteroge-

eous network deployments (HetNets) in which the existing macro

ells are complemented with a number of small low-power base sta-

ions with the goal of increasing capacity in highly congested areas

207208] It is envisaged that small cells will be based on 4G stan-

ards (eg pico and femto cells) as well as IEEE 80211 technologies

nd multimode base stations that work simultaneously with LTE and

iFi are already entering the market

It is important to point out that LTE standards already support a

ariety of mechanisms that enable data offloading However most of

he existing solutions such as Local IP Access (LIPA) and Selected In-

ernet IP Traffic Offload (SIPTO) focus on data offloading at the core

f the network [205] For instance LIPA allows an IP-enabled mo-

ile device to transfer data to another device in the same pico or

emto cell without passing through the cellular access network while

IPTO enables the routing of selected IP data flows through different

ateways The only LTE offloading mechanism that supports seam-

ess interworking with IEEE 80211-based WLANs is IP Flow Mobil-

ty and Seamless Offload (IFOM) Specifically IFOM relies on Mobile

Pv6 technologies to allow a user terminal to simultaneously route

elected IP flows over different radio access technologies [209210]

n this way a user terminal can offload selected flows to a WLAN

ased on some operator-defined policy while keeping the LTE con-

ection running However to enable such an approach it is neces-

ary to support an entity in the cellular core network that can com-

unicate with the user terminal to exchange information about the

vailability and quality of neighbouring access networks as well as to

rovide the user terminal with predefined rules to manage the han-

over process This entity in current LTE standards is called the Ac-

ess Network Discovery and Selection Function (ANDSF) server [211]

ote that the problem of selecting the best communication technol-

gy in an heterogeneous wireless network has been extensively in-

estigated in the literature both with centralised as well as decen-

ralised approaches (see [212] for a survey) However the use of mul-

iple interfaces in parallel as well as per-flow offloading are relatively

ew concepts Thus the design of scalable and efficient network se-

ection strategies for the ANDSF framework is still an open issue For

nstance three offloading methods suitable for the ANDSF framework

re proposed in [213] which are based on coverage SNR and sys-

em load A reinforcement learning approach is designed in [214] that

llows multimode base stations to autonomously steer their traffic

ows across different access technologies depending on the traffic

ype the usersrsquo QoS requirements the network load and the inter-

erence levels A number of studies have also tried to quantify the

B Bellalta et al Computer Communications 75 (2016) 1ndash25 3

Table 1

Summary of the IEEE 80211 amendments that are reviewed in this survey

Amendment Release Band Goal New features

80211aa 2012 24 5 GHz Robust streaming of audiovideo

streams

bull Groupcast communication

mechanisms

bull Intra-access category

prioritisation

bull Stream classification service

bull Overlapping BSS management

80211ac 2014 5 GHz Very high-throughput WLAN in

lt 6 GHz band

bull Channel bonding

bull Multi-user Downlink MIMO

bull Packet aggregation

80211af 2014 470ndash790 MHz (EU) WLAN in the TV White Space bull Geolocation-based spectrum

databases

bull Channel sensing

bull Non-contiguous channel

bonding

54ndash72 76ndash88 174ndash216

470ndash698 698ndash806 MHz (US)

80211ah 2016 902ndash928 MHz (US) WLAN in the sub 1 GHz band bull Enhanced power saving

mechanisms

bull Hierarchical station organisation

bull Efficient small data

transmissions

863ndash868 MHz (EU)

755ndash787 (China)

9165ndash9275 MHz (JP)

80211ax 2019 24 5 GHz High efficiency WLANs (HEW) bull Dynamic channel bonding

bull Multi-user Uplink MIMO

bull Full-duplex wireless channel

Table 2

List of other on-going and upcoming IEEE 80211 amendments

Amendment Release Band Goal

80211ae-2012 2012 24 5 GHz Prioritisation of management frames

80211ad-2012 2012 5705ndash64 GHz (US) Very high-throughput WLAN in the 60 GHz band

57ndash66 GHz (EU)

59ndash6290 GHz (China)

57ndash66 GHz (JP)

80211ai 2016 ndash Fast initial link setup

80211aj 2016 45 59ndash64 GHz WLAN in the Chinese Milli-Meter Wave frequency bands

80211aq 2016 ndash Pre-association discovery (PAD)

80211ak 2017 ndash Enhancements for transit links within bridged networks

Fig 1 Survey organisation

n

c

n

c

s

m

u

a

m

v

T

t

umber of persons that use Internet applications and objects that are

onnected to the Internet is growing every day proportionally to the

umber of new applications and services that constantly appear This

learly results in a steady increase of the Internet traffic Two repre-

entative examples of the change in Internet use are (i) the high de-

and for mobile-rich multi-media content mainly motivated by the

se of smart-phones tablets and other multimedia portable devices

nd (ii) the increasing interest in IoT applications driven by the al-

ost ubiquitous existence of devices able to collect data from the en-

ironment ranging from low-power sensor nodes to connected cars

herefore WLANs must also evolve to provide effective solutions

o these new upcoming scenarios and the challenges they pose to

4 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Table 3

Performance requirements for different HD streaming applications

Type Max data rate Max latency

Uncompressed raw video 149 Gbits 100 ms

uncompressed HDTV 150 Mbits 150 ms

Blue-ray Disc 54 Mbits 200 ms

MPEG2 HDTV 192 Mbits 300 ms

MPEG4 HDTV 8ndash10 Mbits 500 ms

f

c

p

B

n

t

p

t

c

f

a

b

a

S

2

n

l

o

s

d

c

i

c

T

n

i

C

c

p

p

i

a

r

s

b

c

c

i

p

l

a

t

t

3

8

s

f

p

a

a

t

3

t

t

l

satisfy their requirements Four of the key use cases for next-

generation WLANs are discussed in the following subsections

21 High-quality multimedia content delivery

Our new mobile and portable devices are designed to handle rich

multimedia contents including high-definition video and images

Table 3 reports the requirements in terms of maximum data rate and

latency for some of the most common real-time video applications

[21] Key scenarios in which the support of real-time video trans-

mission is required of course include Internet TV and video stream-

ing Similarly scenarios in which multiple users connect to the same

wireless network to request different multi-media content at the

same time are increasing every day However not all multi-media

content is real time Stored video and image files can also be ex-

changed between different devices Those files can have sizes rang-

ing from a few Megabits to several Gigabits hence requiring a high

network transport capacity in order to provide a good Quality-of-

Experience to end users Although video encoding schemes exist that

offer substantial video compression efficiency such as H264MPEG-

4 AVC [22] WLANs must be able to achieve very high transmission

rates and have content-aware mechanisms that are specifically de-

signed for multi-media applications to ensure a satisfactory service

for multimedia delivery The mechanisms that are considered by var-

ious IEEE 80211 standardisation groups to satisfy those requirements

are described in Section 3 such as group-cast communication proto-

cols single and multi-user spatial multiplexing and channel bond-

ing among others to make the communication more efficient and of-

fer higher transmission rates The reference IEEE 80211 amendments

for high-quality multimedia content delivery are IEEE 80211aa IEEE

80211ac and IEEE 80211ax

22 Machine-to-Machine (M2M) communications

The almost ubiquitous presence of sensoractuator devices that

are able to interact with the environment has fostered the creation

of new services and applications Concepts such as smart cities and

smart grids are being developed on the basis of the existence of those

sensoractuator networks to achieve a more sustainable use of the

environmental resources and provide citizens with a higher quality

of life [2324]

In a classic sense Wireless Sensor Network (WSN) technologies

are used to collect data from spatially distributed sensor nodes and to

transmit the data over a multi-hop wireless network to a central sink

[25] The M2M paradigm is broadening the scope of the WSN con-

cept because it enables networked devices wireless andor wired as

well as services to exchange information or control data seamlessly

without explicit human intervention Clearly M2M communications

face most of the technical challenges of WSNs One of the main lim-

itations of WSNs and M2M systems is that the network nodes are

usually battery powered or have limited access to power sources De-

signing mechanisms and protocols to reduce their power consump-

tion with the goal of extending the network lifetime is therefore cru-

cial for the successful commercial take-up of these kinds of networks

Fortunately devices in M2M systems typically generate or consume a

limited amount of data per unit of time Thus they can spend a large

raction of their time sleeping This facilitates energy saving at the

ost of additional complexity for the channel access and networking

rotocols

Popular wireless protocol standards for M2M communications are

luetooth ZigBee and BT-LE [12] An alternative promoted by mobile

etworks is to connect devices in M2M systems directly to the In-

ernet by using the cellular network infrastructure for which specific

rotocols are being developed [26] WLANs are envisioned as an al-

ernative to both multi-hop WSNs and cellular networks However

urrent WLANs are not able to satisfy the minimum requirements

or M2M communications [13] Novel specific power-saving mech-

nisms are required to support the long periods of inactivity needed

y the sensoractuator devices and to manage the thousands of nodes

ssociated with a single AP These challenges will be discussed in

ection 4 when presenting the IEEE 80211ah amendment

3 Efficient use of the spectrum

The ISM bands are used by several wireless communication tech-

ologies including IEEE 80211 IEEE 802154 and Long Term Evo-

ution (LTE)-Unlicensed networks This results in a high spectrum

ccupancy Unfortunately wireless networks operating in the same

pectrum region can suffer from mutual interference which might

egrade the performance of all of them This is exacerbated by the un-

ontrolled deployment of wireless networks in the ISM band which

s typically very common in urban environments For example let us

onsider a building with several apartments and a WLAN in each one

here would easily be several WLANs operating in overlapping chan-

els and suffering mutual interference [27] To deal with this issue it

s expected that new APs will increasingly incorporate DCA (Dynamic

hannel Allocation) mechanisms to select and update their operating

hannel at run-time

An alternative approach to alleviate the spectrum occupancy

roblem is to move to a different part of the spectrum even if the new

art of the spectrum is occupied by communication systems operat-

ng under a license In that case WLANs would be the secondary users

nd therefore must avoid causing interference to the primary users In

ecent years the change from analogue to digital TV broadcast emis-

ions has resulted in a reorganisation of the spectrum at VHFUHF

ands This reorganisation has shown that there are many empty TV

hannels called TV white spaces that can be used for data communi-

ation especially in rural regions [28] Furthermore WLANs operat-

ng in those TV white spaces can take advantage of radio propagation

roperties in the UHF band to provide large coverage areas The chal-

enges to be addressed to use CSMACA protocols in VHFUHF bands

s well as how to obtain higher transmission rates when the spec-

rum is fragmented will be discussed in Section 5 when presenting

he IEEE 80211af amendment

High performance WLANs for multimedia applications

This section reviews the IEEE 80211ac IEEE 80211ax and IEEE

0211aa amendments These three amendments target multimedia

cenarios by introducing new physical-layer technologies and MAC

unctionalities to improve the WLAN capacity and QoS provision Ap-

lication examples include home scenarios in which a WLAN AP can

ct as an Internet gateway and wireless media server for home appli-

nces (eg IPTV set-top boxes projectors game consoles) and con-

ent storage devices A possible use case is illustrated in Fig 2

1 The IEEE 80211ac amendment

IEEE 80211ac [29] aims to provide users with a throughput close

o 1 Gbps which represents a roughly four-fold increase with respect

o IEEE 80211n [7] Compared to IEEE 80211n IEEE 80211ac supports

arger channel widths (up to 160 MHz) introduced a new modulation

B Bellalta et al Computer Communications 75 (2016) 1ndash25 5

Fig 2 High-throughput demanding multimedia devices associated to an IEEE 80211acax AP

s

(

3

s

w

b

i

t

s

t

c

o

f

C

t

A

p

c

f

u

c

i

t

u

i

t

t

I

s

m

[

a

t

u

s

T

e

S

Fig 3 A DL-MU-MIMO transmission in IEEE 80211ac Note that the AP transmits two

spatial streams in SU-MIMO mode to STA B and a single spatial stream to STA A Packet

aggregation is used in both transmissions The PHY header is transmitted ommnidirec-

tionally to inform the selected STAs about the next MU-MIMO transmission

n

C

c

F

D

t

t

s

h

d

d

b

m

k

c

M

A

D

d

3

n

(

e

i

i

I

cheme ie a 256-QAM modulation and downlink multiuser MIMO

DL-MU-MIMO)

11 Novel features

The most relevant new features included in IEEE 80211ac are de-

cribed in the following

Channel bonding IEEE 80211ac enables the use of channel band-

idths of 20 40 80 (mandatory) and 160 MHz (optional) Channel

andwidths larger than 20 MHz are created by ldquobondingrdquo (ie group-

ng) a group of consecutive 20 MHz channels and aim to offer higher

ransmission rates

Two extensions have been proposed in IEEE 80211ac for the ba-

ic DCF (Distributed Coordination Function) access method in order

o support channel bonding (i) the Static Bandwidth Channel Ac-

ess Protocol (SBCA) which always transmits over the same group

f 20 MHz channels and requires that all sub-channels are idle be-

ore starting a packet transmission and (ii) the Dynamic Bandwidth

hannel Access scheme (DBCA) which is able to dynamically adapt

he channel width to the instantaneous spectrum availability [3031]

s expected in dense scenarios the use of DBCA offer a much better

erformance than SBCA due to adaptability [32]

To avoid hidden terminals operating in any of the 20 MHz bonded

hannels the IEEE 80211ac amendment includes extended RTSCTS

rames in order to signal the maximum channel width that can be

sed at both the transmitter and the receiver In case the CTS in-

ludes a lower channel width than the RTS the transmitter will adopt

t Similarly to the ACK frames when the RTS and CTS frames are

ransmitted they are duplicated over all the 20 MHz sub-channels

sed The operation and performance of channel bonding in WLANs

s thoroughly analysed in [27] showing the new interactions be-

ween neighbouring WLANs that may appear and their impact in the

hroughput of each one

Downlink multiuser MIMO The main novelty introduced by the

EEE 80211ac amendment compared with the IEEE 80211n one is the

upport of MU-MIMO transmissions in the downlink hence allowing

ultiple simultaneous transmissions from the AP to different STAs

58] In the IEEE 80211ac amendment the AP can be equipped with

maximum of eight antennas and send up to four spatial streams

o two different users or up to two spatial streams to four different

sers at the same time

When an IEEE 80211ac AP performs a multi-user transmission it

pecifies the group of STAs to which that transmission is directed

his information is contained in the new IEEE 80211ac PHY head-

rs which are broadcast omni-directionally to all STAs The way

TAs are grouped is decided by the AP after obtaining the chan-

el state information (CSI) feedback from all STAs To gather the

SI information by the AP IEEE 80211ac considers only an explicit

hannel sounding feedback mechanism called Explicit Compressed

eedBack (ECFB) The channel access is governed by EDCA (Enhanced

istributed Channel Access) At each transmission attempt the mul-

iple access categories (AC) managed by the AP should contend for

he channel medium as only one AC can be served for each transmis-

ion attempt In the case that the queue associated with the AC that

as won the internal contention does not contain packets to enough

ifferent destinations to fill all the available spatial streams it can

ecide to share the remaining ones with the other ACs

Packet aggregation To increase the efficiency of each transmission

y reducing unnecessary overheads IEEE 80211ac allows the trans-

ission of several MPDUs aggregated in a single A-MPDU Then to ac-

nowledge each MPDU individually a Block ACK packet is used which

ontains a bitmap to indicate the correct reception of all included

PDUs Thus leveraging on the information contained in the Block

CK the transmitter is able to selectively retransmit only those MP-

Us that have failed instead of the whole A-MPDU Fig 3 illustrates a

owlink MU-MIMO transmission in which packet aggregation is used

12 Open challenges

Since the IEEE 80211ac amendment has recently been fi-

alised current research around it should cover two main aspects

a) understanding the performance bounds of IEEE 80211ac which

ntails the development of new models simulation tools and exper-

mental platforms of IEEE 80211ac-based WLANs and (b) propos-

ng specific solutions for those aspects that are not defined by the

EEE 80211ac amendment on purpose such as the mechanism for

6 B Bellalta et al Computer Communications 75 (2016) 1ndash25

g

i

l

a

b

b

f

o

w

o

H

t

e

n

a

f

t

m

a

w

s

3

a

8

2

c

s

m

W

r

a

[

o

m

2

n

c

p

r

a

creating the groups of STAs for DL-MU-MIMO transmissions smart

packet schedullers able to decide when the use of DL-MU-MIMO out-

performs SU-MIMO transmissions and the implementation of the

TXOP sharing feature between several ACs The results and conclu-

sions obtained in both cases will be very valuable in the development

of IEEE 80211ac technologies as well as in the conception of the fu-

ture amendments that will substitute IEEE 80211ac in four to five

years such as the recently initiated IEEE 80211ax

Following the first mentioned research direction there are several

efforts that have focused on understanding both theoretical and exper-

imental performance bounds of IEEE 80211ac The maximum downlink

throughput that an IEEE 80211ac AP can achieve when packet aggre-

gation channel bonding and different spatial stream configurations

are considered is presented in [33] In [34] the authors evaluate the

IEEE 80211ac performance experimentally using commodity devices

focusing on the effects that the use of wider channels the 256-QAM

modulation and the number of SU-MIMO spatial streams have in

terms of throughput and energy consumption It is worth mentioning

that DL-MU-MIMO was not yet implemented in the equipment they

were using and that feature was therefore not considered The evalu-

ation of a DL-MU-MIMO implementation for WLANs using the WARP

platform is presented in [35] where a deep evaluation of the potential

benefits of DL-MU-MIMO transmissions is done in terms of the loca-

tion of the receivers number of users and user mobility among other

aspects A solution that combines both packet aggregation and DL-

MU-MIMO transmissions is presented in [36] Results show the need

of properly dimensioning the buffer space to achieve the full potential

of such a combination In [37] the authors compare the throughput

achieved by IEEE 80211n and IEEE 80211ac when packet aggrega-

tion is used with and without channel errors They show that in most

cases the packet aggregation mechanism introduced in IEEE 80211ac

outperforms the one in IEEE 80211n An analytical model to evalu-

ate the performance of the IEEE 80211ac TXOP sharing mechanism in

DL-MU-MIMO communications is developed in [38] The main goal of

this study is to identify how the TXOP sharing mechanism could im-

prove the system efficiency while achieving channel access fairness

among the different ACs

How to optimally exploit the new DL-MU-MIMO capabilities pro-

vided by IEEE 80211ac is still an open challenge First due to the need

of frequent CSI exchanges between STAs and the AP it is not yet clear

in which conditions DL-MU-MIMO outperforms SU-MIMO [39ndash42]

or even whether MU-MIMO does or does not outperform multi-user

packet aggregation when the amount of data directed to each des-

tination is not balanced [43] Packet aggregation can be a solution

to balance the duration of the multi-user spatial streams as shown

in [36] although it will always depend on the amount of traffic di-

rected to each destination and the buffer capacity at the AP In [44]

the authors compare different strategies to assign the spatial streams

between the available destinations at each transmission in a fully

connected mesh network showing in ideal channel conditions the

theoretical benefits of MU-MIMO vs SU-MIMO

Closely related to the previous point a second open challenge is

the design of efficient schedulers that consider traffic priorities the

buffer state the different MIMO strategies TXOP sharing policies

grouping of STAs and the availability of fresh CSI feedbacks to max-

imise the throughput and guarantee the required QoS for each ac-

tive traffic flow It is important to consider that the availability of up-

dated CSI estimates from all STAs allows the AP to reduce the mutual

interference between the transmitted spatial streams which means

lower packet error probabilities and higher transmission rates How-

ever the overheads for obtaining the CSI from all STAs is large and

increases linearly with the channel sounding rate and the number

of STAs Proposals for reducing the CSI overhead are under develop-

ment For example in [40] the CSI overhead is reduced by inhibiting

the channel sounding whenever possible based on the estimation of

the channel stability for all users Another related problem is how to

roup the STAs as the goal is to find groups of STAs with compat-

ble (ie orthogonal) channels In [45] the authors show the chal-

enges inherent to the group assignment problem and they propose

n heuristic method to solve them TXOP sharing is considered in [46]

y presenting two alternative approaches to enhance the considered

ack-off procedure for the purpose of improving both throughput and

airness

A third key challenge for IEEE 80211ac is to achieve an efficient use

f the spectrum when several channel widths are used in scenarios

ith multiple overlapping WLANs Increasing the channel width the-

retically allows individual WLANs to achieve a higher throughput

owever the presence of other WLANs in the vicinity also increases

he chances of frequency overlapping which may cause the opposite

ffect as there appears inter-WLAN contention [27] Adaptive mecha-

isms to select the channel centre frequency and the channel width

nd MAC protocols to choose the instantaneous channel width used

or each transmission are thus required For instance in [47] the au-

hors focus on the channel selection problem when WLANs can use

ultiple channel widths using a game-theoretic framework In [48]

scheme is proposed to enable the communication between nodes

ith partially overlapping channels which may provide stronger re-

ilience to channel interferences

2 The IEEE 80211ax amendment

In 2014 the High Efficiency WLANs (HEW) Task Group [49] initi-

ted the development of a new IEEE 80211 amendment called IEEE

0211ax The IEEE 80211ax amendment is expected to be released in

019 and to some extent it will be the IEEE 80211 response to the

hallenges of future dense and high-bandwidth demanding WLAN

cenarios [5051]

The challenges in the development of the IEEE 80211ax amend-

ent are to

(i) Improve the WLANs performance by providing at least a four-

fold capacity increase compared to IEEE 80211ac

(ii) Provide support for dense networks considering both the ex-

istence of multiple overlapping WLANs and many STAs in each

of them Spatial reuse of the transmission resources is a must

(iii) Achieve an efficient use of the transmission resources by min-

imising the exchange of management and control packets re-

visiting the structure of the packets and improving channel

access and retransmission mechanisms among others aspects

(iv) Provide backward compatibility with previous amendments

This is achieved by the mandatory transmission of the legacy

PHY preamble in all frames and by keeping EDCA as the basic

channel access scheme

(v) Introduce effective energy saving mechanisms to minimise the

energy consumption

(vi) Support multi-user transmission strategies by further devel-

oping MU-MIMO and Orthogonal Frequency Division Multiple

Access (OFDMA) capabilities in both downlink and uplink

In addition to the aforementioned challenges next-generation

LANs will have to implement some other functionalities beyond the

aw packet transmission and reception Examples are a fast efficient

nd robust handoff between APs in the same administration domain

52] device-to-device communication (D2D) [53] and coordination

f multi-AP networks [54] In the first case the IEEE 80211ai amend-

ent called Fast Initial Link Setup is in progress and expected for

016 Its target is to complete a handoff in less than 100 ms including

ew AP discovery user authentication and configuration Using D2D

ommunication we can avoid the use of the AP as a relay hence im-

roving the overall efficiency as the number of packet transmissions

equired is reduced Finally the virtualisation of network functions

dds a new dimension in the management of multiple APs which in

B Bellalta et al Computer Communications 75 (2016) 1ndash25 7

d

e

I

w

f

t

s

3

P

s

G

t

8

p

C

c

m

c

i

t

a

s

w

f

t

t

e

l

p

e

a

H

d

t

c

o

s

p

n

t

b

a

w

c

t

w

n

8

a

c

w

d

H

n

a

h

p

w

i

t

fi

e

c

u

t

m

p

p

b

p

i

t

8

d

I

m

M

u

t

m

i

t

D

W

I

u

M

i

w

p

h

t

n

o

p

t

s

u

s

r

b

A

t

O

t

t

w

s

p

o

i

W

t

t

a

[

m

ense scenarios can contribute to notably improving the user experi-

nce We further discuss this last topic in Section 6

Different from the other amendments covered in this survey the

EEE 80211ax amendment is just in its initial stages of development

ith only very few technical aspects consolidated at this stage There-

ore in the following subsection we will overview both the new fea-

ures and open challenges of the IEEE 80211ax amendment at the

ame time

21 Novel features and open challenges

The IEEE 80111ax Task Group is currently working in four areas

HY MAC Multi-user and Spatial Reuse [49] Next we will overview

ome of the topics currently under discussion in the IEEE 80211 Task

roup in each category

PHY layer The IEEE 80211ax PHY layer will be an evolution of

he IEEE 80211ac one The challenges in the design of the IEEE

0211ax PHY layer are related with the extensions required to sup-

ort multi-user MU-MIMO and OFDMA transmissions and Dynamic

CA Also improvements in the supported modulation and channel

oding techniques will be likely considered to allow for higher trans-

ission rates at lower SNR values For example IEEE 80211ax may

onsider LDPC (Low-Density Parity Check) coding which is optional

n IEEE 80211ac instead of the traditional convolutional codes as

hey provide a coding gain of 1ndash2 dB Moreover the PHY layer may

lso include some functionalities to support the use of Hybrid ARQ

chemes to improve the efficiency of packet retransmissions

Medium acess control In order to keep backward compatibility

ith previous IEEE 80211 amendments besides a common PHY

rame preamble compatible MAC protocols are required This means

hat it is likely that EDCA will be kept as the main channel access

echnique in the IEEE 80211ax amendment Therefore the most rel-

vant open challenges are related to EDCA extensions to support a

arge number of STAs improve traffic differentiation capabilities im-

rove the energy consumption and provide mechanisms to fairly co-

xist with neighboring wireless networks

To support a large number of contenders with a low collision prob-

bility a simple solution is to use large backoff contention windows

owever it would also increase the time a STA is in backoff so re-

ucing the number of packets it can transmit A solution to reduce

he backoff duration when using large contention windows is to de-

rease the duration of the backoff slots Latest technological advances

n electronics should require less time to perform a CCA check to

witch between reception and transmission modes and to process a

acket thus enabling such a possibility

Another approach to enhance the underlying CSMACA mecha-

ism in EDCA and achieve a higher efficiency is to consider decen-

ralised collision-free MAC strategies Those MAC protocols are able to

uild collision-free schedules thus improving the network efficiency

s collisions are reduced while preserving backward compatibility

ith the default EDCA implementation An overview of several de-

entralised collision-free MAC protocols can be found in [55] Among

hem CSMAECA [56] is especially relevant since it is fully compatible

ith EDCA and latest upgrades made it adaptive to the instantaneous

umber of contenders in a single WLAN

In addition to EDCA IEEE 80211ax WLANs can rely on the IEEE

0211aa amendment to further improve the WLAN traffic differenti-

tion capabilities with intra-AC traffic differentiation and groupcast

ommunication mechanisms among other features In Section 33

e will provide further details about the IEEE 80211aa amendment

IEEE 80211ax will likely keep the same channel widths that were

efined in the IEEE 80211ac amendment ie 20 40 80 and 160 MHz

owever it is expected that IEEE 80211ax will extend current chan-

el bonding mechanisms to further improve the spectrum utilisation

nd the coexistence between neighbouring WLANs For example it

as been shown in [32] that the use of dynamic channel bonding

rovides significant throughput gains in dense scenarios compared

ith the static approach while minimising the inter-WLAN negative

nteractions [27] Furthermore additional mechanisms are required

o fully exploit the use of channel bonding such as the use of ef-

cient algorithms to select the position of the primary channel or

ven to consider the existence of multiple primary channels to in-

rease the number of bonded channel combinations that a node can

se for transmission

The MAC layer in IEEE 80211ax may work with the PHY layer

o implement an efficient Hybrid ARQ mechanism able to retrans-

it only incremental redundancy bits in short packets Opportunistic

iggy backing of data packets in ACKs and vice versa may further im-

rove the efficiency of IEEE 80211ax WLANs by reducing the num-

er of transmissions in a bidirectional data exchange [57] Finally

acket headers can be reduced if shorter STA identificators are used

nstead of MAC addresses and unnecessary fields are avoided in all

ransmissions

Finally in terms of power saving mechanisms it is expected IEEE

0211ax will follow the same trends as in previous IEEE 80211 stan-

ards including the extensions currently under development for the

EEE 80211ah amendment that are discussed in Section 4

Multi-user Multi-user communications will likely be one of the

ain characteristics of IEEE 80211ax as both uplink and downlink

U-MUMO and OFDMA are under consideration The use of multi-

ser communication techniques does not necessarily represent a sys-

em capacity increase because the available transmission resources

ay be the same as in the single-user communication case However

n WLANs the simultaneous transmission from different users is able

o parallelise the large temporal overheads of each transmission (ie

IFS SIFS ACKs packet headers etc) which can notably improve the

LAN efficiency

IEEE 80211ax will further develop the MU-MIMO capabilities of

EEE 80211ac by allowing multiple simultaneous transmissions in the

plink which is known as uplink (UL-) MU-MIMO [58] Similar to DL-

U-MIMO transmissions an open challenge to enable UL-MU-MIMO

s to design a mechanism able to efficiently schedule the users that

ill transmit at the same time In one hand a pure decentralised ap-

roach would be easy to implement with minimal signalling over-

eads However since it requires that all STAs finish their backoff at

he same time it may be very inefficient besides that those STAs may

ot be compatible in terms of their respective spatial channels In the

ther hand a pure centralised approach requires that the AP has com-

lete CSI and buffer occupancy information from all STAs to select

he most suitable group to perform a multi-user transmission Once a

uitable group of STAs is selected by the AP a ldquoTriggerrdquo frame may be

sed to notify the group of selected users that can initiate a transmis-

ion This approach guarantees efficient multi-user transmissions but

equires some extra overheads to collect all the required information

y the AP and signal the selected STAs In both cases new multi-user

CKs will be likely introduced by IEEE 80211ax to acknowledge all

ransmissions with a single control packet

Multi-user OFDMA is also in the agenda for IEEE 80211ax Using

FDMA a channel can be split in several sub-channels and assigned

o different users Likely OFDMA will be implemented in combina-

ion with channel bonding where each of the 20 MHz subchannels

ill be assigned to a different user in both downlink and uplink Be-

ides that a similar operation as in the multi-user MIMO case is ex-

ected as there are almost the same challenges to solve A survey

f current OFDMA proposals for WLANs is presented in [59] show-

ng also how the use of OFDMA is able to significantly improve the

LAN efficiency Fig 4 illustrates an uplink OFDMA and MU-MIMO

ransmission

In addition to Multi-user MIMO and OFDMA the use of Simul-

aneous Transmit and Receive (STR) techniques commonly known

s full-duplex transmission have been suggested for IEEE 80211ax

5051] Using STR a pair of nodes is able to transmit and receive si-

ultaneously [60] which theoretically doubles the channel capacity

8 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 4 Multiuser uplink transmission using two different OFDM subchannels Trans-

missions from STA B and C over the same subchannel represent an uplink MU-MIMO

transmission All transmissions are acknowledged with a single Multiuser ACK The

trigger frame is used to signal the selected STAs

t

t

t

a

s

t

f

f

c

3

s

s

t

n

n

A

d

w

t

fi

a

t

8

i

d

t

r

w

d

R

v

f

s

d

p

B

c

e

k

r

G

a

c

m

t

b

d

3 The concealment address protects legacy stations ie GCR-incapable stations

from receiving duplicated group-addressed frames

The challenge is that both the AP and the STA involved in a full-duplex

transmission have to start to transmit at the same time To achieve

that information about full duplex transmission capabilities can be

included in RTSCTS control packets to set-up a full-duplex transmis-

sion between the transmission initiator and its destination

Spatial reuse Dense WLAN deployments are necessary to offer a

continuous coverage with high transmission rates To improve both

the co-existence with those neighboring networks and the spatial

reuse of the spectrum a WLAN has two options (i) minimise its

area of influence by reducing its transmit power and (ii) accept

higher interference levels by increasing the Clear Channel Assess-

ment (CCA) level Use of both techniques may increase the number of

concurrent transmissions between neighbouring WLANs and there-

fore their capacity although it may also result in the opposite effect

since the achievable transmission rates may be negatively affected by

the higher interference levels observed which is the main challenge

to solve

Due the high WLAN dynamics the use of adaptive systems is cru-

cial but challenging as adaptivity requires extra complexity in terms

of computing and memory resources and there are no guarantees

that the implemented solution converges due to the decentralised

operation of each WLAN The use of DSC (Dynamic Sensitivity Con-

trol) to dynamically adjust the CCA level is one of the aspects cur-

rently under discussion in the IEEE 80211ax Task Group First studies

evaluating the performance of DSC for IEEE 80211ax WLANs show

a clear improvement on the spatial reuse and the area throughput

[61] Another example of the achievable throughput gains obtained

by adapting the CCA level can be found in [62] where the authors

show that gains of up to 100 can be achieved Moreover trans-

mit Power Control (TPC) to mitigate interference between WLANs in

dense scenarios is studied in [63] showing the need of jointly opti-

mising both TPC and CCA to maximise the network performance

Finally sectorisation by using beamforming is also under consid-

eration for the development of the IEEE 80211ax amendment as a

potential solution to improve spatial reuse [64] Using sectorisation

only the nodes of a given area are allowed to receive or transmit data

hence reducing the contention between different networks whenever

they activate non-overlapping sectors A challenge here is to coor-

dinate the different neighboring APs when they belong to different

administration domains Decentralised learning approaches may be

implemented to find feasible temporal patterns of non-overlapping

sectors

33 The IEEE 80211aa amendment

As discussed above legacy IEEE 80211 standards do not provide

robust and efficient delivery of audiovideo streaming services Thus

he IEEE 80211aa amendment was developed to include new fea-

ures and additional mechanisms to improve the performance of real-

ime multi-media content delivery [65] Specifically IEEE 80211aa

ddresses the following five shortcomings of previous 80211

tandards [1666]

(i) the lack of reliable and efficient support for multicast and

group communications

(ii) the incapacity of applying traffic prioritisation to different

multimedia streams or different types of frames from the same

stream

(iii) the absence of methods for cooperative resource sharing

among neighbouring APs

(iv) the lack of mechanisms for graceful degradation of audiovideo

streaming quality

(v) the non-interoperability with existing IEEE 8021 standards for

Audio Video Bridging (AVB)

In the following sections we present in detail the solutions to

hose problems introduced in the IEEE 80211aa amendment We

urther discuss the research studies that have provided the basis

or the IEEE 80211aa design and we identify the remaining open

hallenges

31 Novel features

Groupcast communication mechanisms In most audiovideo

treaming applications a group of clients must receive the same

tream simultaneously A multicast protocol is necessary to avoid that

he same content is replicated throughout the network In wireless

etworks multicast transmission can exploit the intrinsic broadcast

ature of the wireless channel ie broadcast transmissions from an

P are physically received by all other stations in the same collision

omain However multicast and broadcast frames in IEEE 80211 net-

orks are not protected by an acknowledgement mechanism as in

he case of unicast frames Thus layer-2 multicast transmissions de-

ned by legacy IEEE 80211 standards are unreliable and not suit-

ble for streaming applications To partially address this limitation

he Direct Multicast Service (DMS) was first specified in the IEEE

0211v amendment [67] Basically DMS converts multicast streams

nto unicast streams In this way frames destined to a multicast ad-

ress are individually transmitted as unicast frames to the stations

hat joined that multicast group Obviously DMS provides the same

eliability as unicast transmission services but the consumed band-

idth increases linearly with the number of group members To ad-

ress this scalability issue IEEE 80211aa includes the Groupcast with

etries (GCR) service in addition to DMS Specifically the GCR ser-

ice defines new mechanisms and the related management frames

or group formation which allows a set of stations to agree on a

hared (non-multicast) address called the groupcast concealment ad-

ress3 Furthermore the GCR service specifies two retransmission

olicies GCR Unsolicited Retry (GCR-UR) and GCR Block Ack (GCR-

A) When using GCR-UR the AP can proactively retransmit all group-

ast frames a number of times to mitigate the impact of channel

rrors (see Fig 5a)) while receivers are not requested to send ac-

nowledgements Intuitively this approach improves transmission

eliability but it still suffers from scalability issues In contrast when

CR-BA is used the AP sends a burst of consecutive groupcast frames

nd it requests the receivers to reply with a Block ACK frame which

ontains a bitmap to positively or negatively acknowledge trans-

itted frames (see Fig 5b)) The Block ACK mechanism defined for

he GCR-BA service is quite flexible because Block ACK frames can

e requested immediately after a transmission burst or after a ran-

omised back-off delay Furthermore the AP can request the Block

B Bellalta et al Computer Communications 75 (2016) 1ndash25 9

Fig 5 GCR service with different retransmission schemes

A

r

o

l

o

e

v

e

v

d

f

d

s

v

c

s

a

a

f

d

t

F

t

d

v

a

f

c

t

t

o

t

i

c

r

v

i

r

s

o

t

a

(

p

i

f

a

o

i

t

t

c

s

fi

o

c

I

n

i

a

t

a

t

f

i

B

T

f

t

a

g

a

a

i

s

a

d

s

p

m

m

i

a

t

b

t

i

(

r

a

S

i

c

s

c

T

Q

p

a

3

g

t

4 TSPEC is a traffic specification sent from a QoS capable wireless client that requests

a certain amount of network traffic from the AP for the traffic stream it represents

CK frame to all groupcast recipients or only to a subset of them to

educe overheads and delays The advantages of the GCR methods

ver broadcast and DMS have been extensively demonstrated in the

iterature [1668]

Intra-access category prioritisation The IEEE 80211e amendment

nly allows traffic differentiation between four different access cat-

gories (ACs) that are broadly mapped to four application classes

oice (VO) video (VD) best-effort (BE) and background (BK) How-

ver there is a variety of streaming services ranging from simple

ideoconferencing to HD streaming over IPTV systems which have

ifferent QoS requirements (see Table 3) To provide the ability to dif-

erentiate among individual streams IEEE 80211aa includes an ad-

itional scheduling layer with respect to IEEE 80211e IEEE 80211aa

plits each one of the transmission queues associated with voice and

ideo ACs into a primary and an alternate queue In this way spe-

ialised scheduling rules can be applied to decide which queue to

erve when the EDCA function for inter-AC collision resolution grants

n access opportunity to voice or video ACs To facilitate the man-

gement of service level agreements IEEE 80211aa follows the de-

ault mappings between user priority values and traffic types that are

efined in the IEEE 8021D standard [69] It is then straightforward

o further map traffic types onto transmission queues and ACs (see

ig 6) Finally it is important to point out that the intra-AC differen-

iation functionality can be used to provide more sophisticated traffic

ifferentiation than simple stream prioritisation For instance most

ideo applications use Scalable Video Coding (SVC) schemes that en-

ble the partitioning of a video sequence into multiple layers with dif-

erent qualities and rates [70] Typically an SVC-based video stream

ontains a base layer which provides a basic level of quality and mul-

iple enhancement layers which can only be decoded together with

he base layer to improve the video quality Thus the different layers

f the same encoded video steam can be easily mapped to different

ransmission queues to receive differentiated QoS [71]

Stream classification service The stream classification service (SCS)

s an optional service that can be provided by an AP to the asso-

iated stations to classify multimedia streams based on arbitrary

ules that are established directly by the stations instead of the con-

entional 8021D user priorities To this end the station request-

ng the use of SCS must initiate an SCS session by sending an SCS

equest frame to the AP which contains an identifier for the SCS

tream and the descriptor of the classification rule The AP may accept

r reject the requirements specified by the station Once accepted

he AP must assign all frames that match the classification rule to

specific AC When intra-access category prioritisation is enabled

see Section 331) calternate transmit queues so that finer grained

rioritisation can be applied Finally there is also a Drop Eligibil-

ty Indicator (DEI) bit in the SCS descriptor that indicates whether

rames from this traffic stream can be dropped in the case that there

re insufficient resources Specifically frames with the DEI bit set to

ne have a higher probability of being discarded because their max-

mum number of allowed retries is smaller than the default Note

hat how to combine intra-AC queues and frame dropping settings

o achieve graceful degradation of the audiovideo stream quality in

ase of bandwidth shortage is beyond the scope of the IEEE 80211aa

pecification

Overlapping Basic Service Set (OBSS) management Network densi-

cation ie a denser deployment of wireless infrastructure nodes is

ne of the key strategies that is used nowadays to easily increase the

apacity of wireless systems even for indoor WLANs [72] However

EEE 80211 networks have a limited number of orthogonal chan-

els available and even if optimised frequency planning is applied

t might happen that neighbouring APs are mutually interfering and

station may affect multiple overlapping BSSs In this case conges-

ion not only increases but it is also likely to observe an unfair us-

ge of wireless capacity with the channel retained by one AP for long

ime intervals This is mainly due to the neighbourhood capture ef-

ect ie hidden terminal phenomena among APs [73] To address this

ssue IEEE 80211aa specifies a new functionality called Overlapping

SS (OBSS) management which is based on two new mechanisms

he first defines a set of parameters to quantify the load and inter-

erence among neighbouring BSSs such as medium occupancy frac-

ion number of admitted audiovideo streams data traffic volumes

nd the number of BSSs that are using the same channel as the tar-

et one Note that the traffic load consists of two components the

llocated traffic which is derived on the basis of the TSPEC values of

dmitted streams4 and predicted traffic which is evaluated by track-

ng the maximum value of the allocated EDCA and HCCA traffic over

even-day periods Once load measurement reports are exchanged

mong the APs a second OBSS component is responsible for coor-

inated admission control procedures on the basis of two suggested

haring schemes proportional sharing and on-demand sharing The

urpose of both schemes is to keep the total allocated traffic below a

aximum value in order to provide some QoS protection to admitted

ultimedia streams Finally IEEE 80211aa recommends implement-

ng additional OBSS management procedures for channel selection

nd cooperatively creating HCCA schedules that do not collide

Interworking with IEEE 8021AVB Audio Video Bridging (AVB) is a

erm commonly used to denote a set of technical standards developed

y IEEE to support real-time streaming services with bounded latency

hrough IEEE 802 networks [74] This objective is achieved by spec-

fying mechanisms to allow the synchronisation of multiple streams

IEEE 8021AS [75]) and traffic shaping (IEEE 8021Qav [76]) and to

eserve network resources for specific audiovideo streams traversing

bridged local area network by using a signalling protocol called the

tream Reservation Protocol (SRP) (IEEE 8021Qat [77]) IEEE 80211aa

ntegrates the SRP operations with the EDCA admission control pro-

edures Specifically the SRP RequestResponse messages are encap-

ulated in the management frames that are used to carry the traffic

haracteristics and the QoS requirements during admission control

his enables the end-to-end management of resource reservation for

oS guaranteed streams even when one or more IEEE 80211 links are

art of a path from the stream producers (called IEEE 8021Q talkers)

nd the stream consumers (called IEEE 8021Q listeners)

32 Open challenges

In recent years several MAC enhancements have been investi-

ated to improve QoS guarantees for real-time multimedia applica-

ions in IEEE 80211 networks [20] and the IEEE 80211aa standard

10 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 6 Stream classification and inter-AC traffic prioritisation

a

l

t

r

n

a

t

n

i

t

i

t

r

d

p

t

t

e

m

b

t

a

i

s

o

s

a

e

t

i

f

e

m

8

t

I

i

fi

which was finalised in 2012 included several of these proposed im-

provements Significant research efforts have focused on improving

the transmission reliability of multicasting by integrating ARQ mech-

anisms in IEEE 80211-based multicast transmissions Modifications

to the legacy MAC protocol were proposed in [78] to enable the

RTSCTS option in multicast mode and to select one or more multi-

cast receivers (called leaders) for acknowledging multicast data pack-

ets However these enhancements require changes to the standard

specifications The main problems of leader-based ARQ schemes are

leader election and the trade-off between scalability and reliability

The authors in [79] propose selecting the multicast recipient operat-

ing in the worst channel conditions as the unique leader but this ap-

proach may perform poorly in lossy environments In the Batch mode

multicast MAC (BMMM) [80] all multicast recipients are polled by

the multicast originator to send individual ACKs but this scheme is

not suitable for large multicast groups The Enhanced Leader Based

Protocol (ELBP) is proposed in [81] on the basis of multiple ACK-

leaders and block acknowledgement techniques Analytical models

are then developed to help select optimal ACK-leaders to meet ap-

plication QoS requirements However the models apply only to sat-

urated traffic while multimedia streams are typically bursty Another

class of reliable multicast protocols relies on busy tones to reduce

packet losses due to collisions [82] but the additional radio inter-

face needed for the busy tone limits the practicality of such solu-

tions An alternative approach to avoid collisions of multicast pack-

ets is the multicast collision prevention (MCP) scheme [83] which

is based on the use of a shorter waiting time for transmitting mul-

ticast packets An interesting approach is also proposed in [84] to

retransmit lost packets using an online linear XOR coding algorithm

However a modification to the standard MAC protocol is required to

enable simultaneous ACK transmissions In summary several differ-

ent methods have been proposed to improve multicast transmission

reliability by integrating ARQ schemes into the protocol architecture

but there are not conclusive results on which is the best solution The

choice of the most efficient mechanism depends on a variety of in-

terdependent factors such as loss ratios channel congestion multi-

cast group size and QoS requirements of multimedia streams A com-

prehensive analytical framework is needed to optimise the setting of

the parameters for each scheme and to dynamically select the best

one

As discussed above one main difference between unicast services

nd multicast services in the legacy IEEE 80211 standard was the

ack of acknowledgements Another critical difference is that mul-

icast frames must be transmitted using a fixed rate in the basic

ate set while the transmission rate of unicast frames can be dy-

amically adapted to the channel and traffic conditions [85] Thus

group of research papers has investigated the use of rate adap-

ation to improve the throughput of multicast services in IEEE 80211

etworks [7186ndash89] For instance the authors in [86] propose us-

ng RTS frames to allow group members to estimate channel condi-

ions Each member will then send a dummy CTS frame with a length

nversely proportional to channel quality In this way the multicast

ransmitter can use the collision duration to predict the lowest data

ate that can be used for group transmissions The overhead intro-

uced by this mechanism is quite high however The solution pro-

osed in [88] called ARSM also relies on feedback messages sent by

he multicast receivers called multicast response frames to identify

he group member exhibiting the poorest channel conditions How-

ver in this case a different back off timer is associated with each

ulticast receiver depending on the SNR of previously received feed-

ack messages in order to prevent collision An approach similar to

he one employed in the Auto Rate Fallback (ARF) protocol a rate

daptation scheme originally proposed in [90] is used in [87] Specif-

cally the number of successful consecutive transmissions and con-

ecutive transmission failures are used to decide when to increase

r decrease the transmission data rate respectively A modified ARF

cheme is also proposed in [71] which can be applied to videos that

re encoded into two layers namely the base and enhancement lay-

rs However how to integrate rate adaptation with the different re-

ransmission policies that are defined in IEEE 80211aa is still an open

ssue

One research area that is expected to be crucial in the success-

ul development of IEEE 80211aa-based products is the design of

fficient scheduling algorithms for supporting voicevideo traffic Al-

ost all research work in this field has been triggered by the IEEE

0211e amendment that enhanced the original IEEE 80211 MAC with

wo new QoS-aware access mechanisms ie EDCA and HCCA [91]

n principle with a well-designed admission control and schedul-

ng scheme HCCA is able to provide hard QoS guarantees to traf-

c flows [9293] However HCCA is rarely implemented in IEEE

B Bellalta et al Computer Communications 75 (2016) 1ndash25 11

8

c

c

a

s

b

w

o

j

t

f

a

Q

t

t

A

p

t

s

a

f

s

H

a

i

c

i

t

s

f

s

s

m

i

t

s

l

t

t

[

t

i

o

b

r

W

v

m

8

s

i

o

p

q

4

c

e

t

d

a

t

m

Fig 7 WLANs for M2M communications STAs represent sensor and actuator devices

4

t

w

s

a

c

g

e

(

p

o

t

p

9

r

o

a

p

c

o

i

p

b

f

t

s

o

i

4

8

f

t

i

s

m

d

0211e-based WLANs owing to its higher complexity and cost con-

erns Instead EDCA is widely adopted Most papers have thus fo-

used on improving EDCA performance Many papers have proposed

nalytical models for various subsets of EDCA functionalities For in-

tance a saturation-based performance analysis is conducted in [94]

y differentiating the minimum back-off window size the back-off

indow-increasing factor and the retransmission limit The authors

f [9596] also model AIFS differentiation while the model in [97]

ointly captures all the four EDCA parameters for traffic differen-

iation More recent papers have analysed the EDCA performance

or non-saturated conditions and for arbitrary buffer sizes [98] The

uthors in [99] have developed an analytical model to predict the

oS levels that can be achieved once a new voicevideo flow is in-

roduced in the WLAN A Kalman filter is proposed in [100] to ob-

ain estimates on the number of active transmission queues of each

ccess Category in EDCA These analytical models can then be ex-

loited to derive the optimal configuration of the EDCA parameters

o achieve given performance criteria or to design admission control

chemes that preserve QoS constraints For instance a scheme that

ssigns contention-window values to achieve pre-defined weighted-

airness goals is proposed in [101] A control-theoretic scheme is de-

igned in [102] with the goal of minimising the video traffic delay

owever most of these solutions rely on non-realistic assumptions

bout video traffic dynamics An alternative class of solutions dynam-

cally updates the EDCA parameters based on the observed network

onditions In [103] the EDCA parameters are optimised consider-

ng a WLAN with rigid and elastic traffic simultaneously analysing

he interactions between both types of traffic The authors in [104]

pecify several bandwidth-sharing mechanisms with guaranteed QoS

or voice and video traffic Measurement-based admission control

chemes are proposed in [105] A TXOP adaptation method is de-

cribed in [106] that takes into account video frame sizes and trans-

it queue lengths However the main drawback of these solutions

s that they are based on heuristics and hence do not ensure op-

imal and guaranteed performance Finally a third category of re-

earch papers tries to improve video performance by designing cross-

ayer scheduling approaches Specifically these works take advan-

age of multi-layer video encoding to classify the frames according

o their importance and assign them to different access categories

107] For instance the authors in [108] define classifiers and waiting

ime priority schedulers that dynamically change the packet prior-

ties according to end-to-end delay measurements A disadvantage

f this approach however is that an additional adaptation layer may

e needed to implement the complex interactions that are typically

equired between the video coding applications and the MAC layer

e conclude this section by pointing out that existing studies pro-

ide the basic design principles and techniques for improving multi-

edia streaming performance in IEEE 80211 networks Still the IEEE

0211aa standard poses new research challenges that have not been

ufficiently explored and that will require innovative solutions For

nstance scheduling between primary and alternate queues is still an

pen research area as the mapping of individual frames to multi-

le queues in order to achieve graceful degradation of voicevideo

uality [16]

Sensor Networks and machine-type communications

As discussed in Section 2 M2M communications refer to any

ommunication technology that enables sensoractuator devices to

xchange information and perform actions without the manual assis-

ance of humans This section reviews the main features currently un-

er consideration in the development of the upcoming IEEE 80211ah

mendment which targets the main challenges of those networks as

he IoT in general such as the energy consumption or the manage-

ent of many devices

1 The IEEE 80211ah amendment

The IEEE 80211ah amendment [109] aims to provide WLANs with

he ability to both manage a large number of heterogeneous STAs

ithin a single BSS and minimise the energy consumption of the

ensor-type battery-powered STAs

The initial design requirements of the IEEE 80211ah amendment

re detailed in [110] these entail the support of up to 8192 STAs asso-

iated with a single AP the adoption of efficient power saving strate-

ies a minimum data rate of 100 kbps the operation in the license-

xempt sub 1 GHz band and a coverage up to 1 km in outdoor areas

see Fig 7 for an illustrative example) A preliminary assessment of

erformance of the IEEE 80211ah technology in terms of the number

f STAs that can be effectively supported in a single WLAN as well as

heir energy consumption is presented in [111]

IEEE 80211ah operates over different sub-1 GHz ISM bands de-

ending on country regulations 863ndash868 MHz in Europe 902ndash

28 MHz in the US and 9165ndash9275 MHz in Japan China South Ko-

ea and Singapore also have specific channelisations Channel widths

f 1 MHz and 2 MHz have been adopted although 4 8 and 16 MHz

re also supported in some countries IEEE 80211ah furthermore pro-

oses new PHY and MAC layers The IEEE 80211ah PHY layer can be

onsidered to some extent a sub-1 GHz version of the IEEE 80211ac

ne At the physical layer OFDM is the chosen modulation method us-

ng 32 or 64 tonessub-carriers that are spaced by 3125 kHz The sup-

orted modulations include BPSK QPSK and from 16 to 256-QAM A

road range of antenna technologies ranging from single-user beam-

orming to MIMO and DL-MU-MIMO which was first introduced in

he IEEE 80211ac amendment are also included in the IEEE 80211ah

pecification Similarly the IEEE 80211ah MAC protocol include most

f IEEE 80211 main characteristics further extending its power sav-

ng (PS) mechanisms

11 Novel features

This section overviews the extensions introduced by the IEEE

0211ah amendment to the IEEE 80211 PS mechanisms to account

or the specific characteristics of resource-constrained sensor and ac-

uator devices A more detailed review can be found in [110] includ-

ng a performance assessment of IEEE 80211ah in several of the key

cenarios for M2M communications such as agriculture and animal

onitoring smart metering and industrial automation plants In ad-

ition a detailed survey of the IEEE 80211ah is reported in [112]

12 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 8 IEEE 80211ah PS mode for TIM-STAs non-TIM STAs and Unscheduled STAs

a

d

w

i

a

a

c

m

i

a

e

t

s

m

t

t

u

b

i

I

s

d

f

a

b

s

a

b

S

T

c

t

w

I

4

a

a

which addresses aspects not considered in [110] such as the use of

sectorisation to avoid overlapping between multiple IEEE 80211ah

WLANs and the use of relays to further extend the IEEE 80211ah cov-

erage among other aspects Finally in [113] both IEEE 80211ah PHY

and MAC characteristics are introduced with emphasis on the ben-

efits of using the sub-1 GHz ISM band in terms of link budget the

location of the OFDM pilots for outdoors operation the use of bidi-

rectional transmission opportunities and the new packet formats to

minimise overheads among others

Enhanced power saving mechanism PS mechanisms for WLANs

were already considered in the development of the first IEEE 80211

standard with the goal of improving the lifetime of battery equipped

devices [114] In PS mode STAs keep the transceiver in sleeping mode

as much time as possible They periodically wake up to listen to the

beacons transmitted by the AP Those beacons indicate whether an

STA has packets waiting for it at the AP In the positive case that STA

remains awake and requests the delivery of those packets Otherwise

given it has nothing to receive it returns to sleep mode until the next

beacon is expected

In the IEEE 80211ah amendment time is divided into pages DTIM

(Delivery Traffic Indication Map) periods TIM (Traffic Indication Map)

periods and slots DTIM and TIM periods begin with the correspond-

ing DTIM and TIM beacons sent by the AP The functions of DTIM and

TIM beacons are described below

1 DTIM beacons They inform as to which TIM Groups (ie the

group of STAs assigned to the same TIM period) have pending

packets at the AP

2 TIM beacons Each TIM message informs a TIM Group about

which specific STA has pending data in the AP Between two

consecutive DTIMs there are as many TIM beacons as TIM

Groups

Using this DTIMTIM-based approach any STA can enter into a

power saving state if it does not have packets pending for transmis-

sion and one of two conditions is met (1) it observes in the DTIM

beacon that there is no downlink traffic addressed to its TIM Group or

(2) it observes in the DTIM beacon that there is some downlink traffic

addressed to its TIM Group but that STA does not explicitly appear in

the TIM beacon Compared to the preliminary IEEE 80211 PS mecha-

nism this approach reduces the size of the Traffic Indication Map in

each TIM beacon thus reducing the overhead and the time STAs need

to listen and process them In addition TIM periods can be organised

in pages which further increases the number of TIM groups between

two DTIM beacons

The temporal organisation of pages DTIM and TIM periods is re-

ported in Fig 8 which shows the division of TIM periods in RAW (Re-

stricted Access Window) and PRAW (Periodic RAW) The RAW is a

time interval in each TIM period where TIM stations can transmit and

receive data (see below the different types of STAs defined in IEEE

80211ah) In addition it can be divided into several downlink and

uplink slots for further granularity In the downlink the slots are as-

signed to a single STA or a group of STAs while the uplink slots are

randomly selected by the STAs with packets ready for transmission

The PRAW is the period of time in each TIM where non-TIM stations

can transmit and receive data

Types of STAs IEEE 80211ah supports three types of STAs TIM

non-TIM and Unscheduled STAs A TIM station is assigned to a TIM

Group Their data transmissions must be performed within a RAW

(Restricted Access Window) period Non-TIM stations do not have to

listen to beacons to transmit data During the association process

non-TIM devices directly negotiate with the AP to obtain a trans-

mission time allocated in a PRAW The following channel access can

either be renegotiated or occurs periodically depending on the re-

quirements set by the STA Unscheduled STAs do not need to listen to

any beacons similar to non-TIM stations Even inside any restricted

access window unscheduled STAs can send a poll frame to the AP

sking for immediate access to the channel The response frame in-

icates an interval (outside both restricted access windows) during

hich unscheduled stations can access the channel This procedure

s meant for STAs that transmit data very sporadically

Hierarchical station organisation To support a large number of STAs

nd their organisation in pages DTIM and TIM periods IEEE 80211ah

ssigns to each associated STA a unique identifier of 13 bits which is

alled the Association Identifier (AID) Using this new AID the maxi-

um number of supported STAs is increased from the original 2007

n IEEE 80211 to 8191 (= 213 minus 1) in IEEE 80211ah However it also

llows categorising STAs according to the type of application they are

xecuting their power level or even their desired QoS by assigning

hem to different TIM groups

Long sleeping periods IEEE 80211ah offers TIM Non-TIM and Un-

cheduled STAs the possibility to set very long doze times (up to

onths) The corresponding clock drift produced by such long doze

imes must be taken into consideration however as the higher the

ime an STA has been asleep the further in advance it should wake

p to avoid possible synchronisation problems with the network

Efficient small data transmission Three new enhancements have

een proposed to reduce the overhead when the data packet size

s small First while IEEE 80211 contains a 28-byte MAC header

EEE 80211ah proposes a short 18-byte version by using AIDs in-

tead of MAC addresses Second IEEE 80211 has defined several null

ata packet (NDP) frames which consist only of a PHY header These

rames can be used to create short ACKs short Block ACKs short CTSs

nd short PS-Polls Finally a Fast Frame Exchange mechanism has

een developed so that if an STA has data to transmit it can notify a

uccessful reception by transmitting its data frame instead of an ACK

Sectorisation Since the PHY layer is based on the IEEE 80211ac

mendment single and multi-user beam-forming are also supported

y IEEE 80211ah This allows the transmission of data to multiple

TAs simultaneously in the downlink increasing the system capacity

he use of the beam-forming capability of IEEE 80211ah APs is also

onsidered as a means to group STAs into different independent an-

enna sectors with the main goal of reducing interference issues This

ould be particularly useful in the case of overlapping with other

EEE 80211ah WLANs or in the presence of hidden nodes

12 Open challenges

A first open challenge is to understand the coverage and achiev-

ble transmission rates in IEEE 80211ah WLANs in both indoor

nd outdoor scenarios Propagation models for WLANs working at

B Bellalta et al Computer Communications 75 (2016) 1ndash25 13

f

c

8

p

R

p

i

m

l

a

m

c

p

p

f

m

t

a

e

p

t

m

a

l

T

m

a

i

n

t

t

t

u

p

c

w

f

a

t

8

a

m

a

s

l

a

m

b

m

d

S

i

a

o

a

R

t

n

G

fi

s

t

p

t

c

u

c

w

i

o

p

g

a

c

i

f

c

(

d

m

l

i

R

t

l

t

L

m

a

m

o

e

f

a

I

i

t

w

i

t

a

c

w

p

f

t

d

5

i

s

e

s

w

5

a

c

a

n

requencies lower than 1 GHz are evaluated in [115] The authors

ompare two path loss propagation models proposed by the IEEE

0211ah Task Group (one for macro and one for picohotzone de-

loyments) [116] with Lee and HatandashOkumura propagation models

esults show that the IEEE 80211ah channel models underestimate

ath loss with respect to Lee and Hata models Moreover in a compar-

son with empirical data it is observed that the IEEE 80211ah channel

odels also underestimate the initial loss and the slope of the path-

oss function A new model parameterisation is thus proposed by the

uthors In [117] the feasibility of an IEEE 80211ah WLAN deploy-

ent is also evaluated in terms of the achievable range and bit rate

omputed on the basis of the link budget using the same path loss

ropagation models as in [115] Results show that the transmission

ower limitations in the uplink can limit the overall network per-

ormance Finally in [118] the authors evaluate the achievable trans-

ission range for the different transmission rates and the achievable

hroughput for different combination of transmission power values

nd transmission rates Further studies are required due to the het-

rogeneity of scenarios in which IEEE 80211ah WLANs can be de-

loyed to characterise the effective coverage ranges with special at-

ention to mixed indoors and outdoors scenarios and in presence of

obile nodes

Due to the wide coverage IEEE 80211ah WLANs may be severely

ffected by the presence of hidden terminals To mitigate such a prob-

em smart solutions to guarantee that the STAs assigned to the same

IM Group are inside each other transmission range or the employ-

ent of RTSCTS-based solutions may contribute to mitigate such

problem An overview of long-range scenarios for IEEE 80211ah

s presented in [119] including a detailed description of the hidden

ode problem in them

Single-hop uplink transmissions from distant STAs require high

ransmission power to reach the AP Since not all STAs may be able

o transmit at the required power a solution could be the introduc-

ion of nodes able to relay transmissions from them However the

se of relays has to be efficiently harmonised with the operation in

ower saving mode allowing temporal periods in which the relays

an gather the data from their associated STAs and then periods in

hich they forward the data to the AP which is still an open challenge

or IEEE 80211ah WLANs In case relays are used data prediction and

ggregation techniques could be implemented to make more efficient

ransmissions [120]

Another challenge is the assessment of the efficiency of IEEE

0211ah PS mechanisms including their optimisation In [111] the

uthors evaluate the capacity of the TIM and page segmentation

echanisms in terms of the maximum number of nodes supported

nd the energy consumed given a certain network traffic profile Re-

ults confirm that a large number of STAs can be supported with

ow energy consumption The impact of the number of nodes oper-

ting in PS mode on the energy consumption and the delay perfor-

ance is analysed in [121] A detailed analysis of the TIM Group-

ased channel access adopted by the IEEE 80211ah task group is

ade in [122] where the STAs assigned to each TIM are uniformly

istributed between the slots of the RAW Since only one group of

TAs is allowed to transmit in each RAW slot the channel contention

s minimised The number of downlink and uplink slots in a RAW

s well as their duration is optimised on the basis of the number

f active STAs and the network traffic profile in [123] In [124] the

uthors propose an algorithm to determine the optimal size of the

AW interval for uplink transmissions based on the estimation of

he number of STAs transmitting and the duration of the RAW Fi-

ally in [125] the authors address the contention problems in Smart

rid communication networks with many nodes and periodic traf-

c when using an IEEE 80211ah WLAN It is still a challenge to con-

ider smart systems able to adapt the IEEE 80211ah parameters to

he instantaneous system state in order to save as much energy as

ossible

Since the PS mechanisms affect only the operation of TIM STAs

he performance of non-TIM and Unscheduled STAs has not yet been

onsidered in the literature To identify the scenarios in which the

se of non-TIM and Unscheduled STAs are of interest is still an open

hallenge as further investigating how non-TIM STAs negotiate over

hich PRAWs they can use to transmit and receive data Co-existence

ssues between the three types of STAs in a single network is also an

pen challenge that requires further work to guarantee a minimum

erformance level for all of them

The design of strategies to distribute the STAs between all the TIM

roups based on their specific traffic profile position battery level

nd application priority is another important challenge that is still

ompletely open EDCA is the default channel access scheme included

n the IEEE 80211ah amendment and provides some basic traffic dif-

erentiation capabilities at the packet level Besides the different ac-

ess categories (ACs) needing to be renamed and their parameters

Arbitration Inter-Frame Spacing or AIFS CWmin TXOP duration) up-

ated to fit the specific traffic profiles of M2M communications other

echanisms can be implemented A mechanism to assign the down-

ink RAW slots when packets from multiple priority levels are wait-

ng for transmission at the AP is required for example Also uplink

AW slots are currently selected randomly by the STAs that want to

ransmit a packet A mechanism to reserve some slots at each up-

ink RAW for high priority STAs may therefore be an option although

his may severely degrade the performance of the low priority STAs

astly different repetition patterns for different priority TIM groups

ay also allow priority STAs assigned to those priority TIM groups to

ccess the channel more often We expect this challenge will receive

uch attention in the upcoming years because of the heterogeneity

f sensors that will be connected though a single IEEE 80211ah AP

specially in urban scenarios

Similar to what has been discussed for 4G cellular networks used

or M2M communications [126] the development of mechanisms to

void and resolve congestion situations is another open challenge for

EEE 80211ah WLANs with many nodes For instance a mechanism

n which STAs have to wait a random number of DTIM periods before

hey can start a transmission may be implemented This approach

ould effectively distribute the traffic over a longer period of time

n overload conditions but otherwise would unnecessarily increase

he access delay Adaptive and load-aware solutions may play again

n important role to keep the optimal WLAN operation in all possible

ircumstances

Finally since IEEE 80211ah will compete with 4G5G cellular net-

orks and WSNs to provide M2M connectivity in many different ap-

lication domains such as smart cities or e-health comparative per-

ormance studies to determine the strong and weak points of each

echnology are required including also aspects such as the cost of the

evices and the system reliability

Cognitive radio technology for TV White Spaces

This section gives an overview of the IEEE 80211af amendment

ntroducing its most relevant features and open challenges Fig 9

hows the basic components and features considered for WLANs op-

rating in the TV White Spaces (TVWS) which include local spectrum

ensing by the AP and STAs and the use of geolocation data bases

ith information on channel availability

1 The IEEE 80211af amendment

Thanks to the transition from analog TV to digital TV several VHF

nd UHF spectrum channels used for decades for analog TV broad-

asting are now unused or are under-utilised in many geographical

reas [127ndash129] This ldquodigital dividendrdquo of spectrum has suggested

ational regulators such as the FCC in the US and Ofcom in the UK

14 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 9 WLAN operating in TVWS basic elements and functionalities

Fig 10 Hidden terminal problem in TVWS when the secondary network uses only in-

stantaneous channel sensing to decide whether or not a TV channel is occupied When

the STA is not able to detect the TV signal due to the presence of obstacles it may

initiate a transmission and create interference to the primary users

s

c

5

a

i

p

t

o

o

t

e

c

u

l

m

b

a

m

t

e

t

t

t

m

D

c

I

n

b

u

8

c

a

t

i

t

w

f

to discuss how to reuse these channels for unlicensed devicesrsquo com-

munications [130131] These potentially vacant channels in the VHF

and UHF bands are referred to as TVWS [132] and include spectrum

portions like the 470ndash790 MHz in Europe and non-contiguous 54ndash72

76ndash88 MHz 174ndash216 MHz 470ndash698 MHz and 698ndash806 MHz in USA

A snapshot of the TV spectrum occupancy in the city of Barcelona in

2012 is shown in [133]

The attractive characteristics of TVWS (not only for WLANs) in-

clude the ability to penetrate through walls and other obstacles much

more effectively than other widely used spectrum bands such as the

24 and 57 GHz ISM bands [134ndash137] This fact along with a progres-

sive spectrum scarcity in ISM bands has suggested the birth of the

IEEE 80211af amendment published in February 2014 [138] which

provides the IEEE 80211 operational characteristics for TVWS access

of unlicensed White Space Devices (WSD) A good summary of IEEE

80211af can be found in [137] The main advantage of operating IEEE

80211 WLANs in the TVWS comes from an increased coverage range

which can reach up to 1 km in rural areas and open fields [139]

and less energy needed to transmit However this comes at the cost

of an increased interference risk to other WSDs which creates co-

existence problems and demands new PHY and MAC layers to ef-

ficiently support channel access and operations preserving licensed

usersrsquo devices

WLANs operated in the TVWS could cover a number of interest-

ing and emerging use cases and scenarios eg Internet access in ru-

ral or sparsely populated areas Smart Grid sensor aggregation me-

tering and control Internet of Things advanced WLAN operations

and TVWS traffic offloading in indoor environments [140141] Nev-

ertheless the concept could be realised to create or extend com-

mercial or municipality WiFi services offered to citizens with cover-

age at the whole city level realised with reasonably limited network

infrastructures [142143]

IEEE 80211af will use the PHY and MAC layers derived from IEEE

80211ac It will adopt concepts such as the OFDM multi-user beam-

forming contiguous and non-contiguous channel bonding and packet

aggregation Among the mandatory and innovative behavioural and

operational parameters the most notable one is the channel acqui-

sition support realised through remote geolocation-based spectrum

allocation databases which maintain the channelsrsquo availability infor-

mation in any given area and time of day providing upon request the

list of free channels available for use

In the following section we highlight and summarise three main

novelties that IEEE 80211af introduces In Section 511 we explain

how the access infrastructure to the remote spectrum database is de-

signed and what the requirements are we discuss the coexistence is-

sues and methodologies adopted in IEEE 80211af and we also discuss

the novel concept of non-contiguous channel bonding The following

ections also contain an illustration of related works Open research

hallenges will be summarised in Section 512

11 Novel features

This section provides an illustration of novel features introduced

nd discussed in the path to IEEE 80211af and the directions provided

n related works and standardisation initiatives to resolve classical

roblems properly characterised in the new framework of TVWS

echnologies

Channel acquisition spectrum database and channel sensing Much

f the attention in operating in the TVWS is given to the protection

f the primary licensed users in the TVWS spectrum band In general

he primary users were considered as the Digital TV (DTV) broadcast-

rs and receivers In fact receivers would suffer the wasteful effect of

ollisions during the TV broadcast reception in the case of secondary

sersrsquo (SU) transmission interference (Fig 10)

It is a well-known problem that the definition of a coherent wire-

ess channel status (idlebusy) from the viewpoint of distributed pri-

ary users (eg DTV receivers) is made difficult by the difference

etween transmitters and receivers interpretation of signals in a vari-

ble time-space collision domain resulting in hiddenexposed ter-

inals concepts Solutions proposed to reduce the hiddenexposed

erminals included the RTSCTS mechanisms and beaconing How-

ver it is difficult to implement solutions that would effectively de-

ermine the channel status of TVWS channels based on localised dis-

ributed sensing activities performed by multiple secondary users in

he same collision domain On the other hand the use of RTSCTS

echanism would not be effective in this case because most of the

TV primary users (and specifically the receivers) were not originally

onceived to transmit signals or implement the RTSCTS handshake

n conclusion the actions to identify free (and busy) TVWS chan-

els should not be based on primary usersrsquo involvement but should

e almost totally implemented by specifically designed secondary

sersrsquo methodologies

There has been much debate on the methodology that IEEE

0211af secondary devices should adopt to get knowledge of free

hannels at a given time on a given target communication area in

n effort to guarantee verified optimised and reliable TVWS spec-

rum use The three main methodologies discussed include sens-

ng solutions geolocation databases (DB) and beaconing Moreover

wo approaches are possible in general distributed and centralised

ith their well known tradeoffs in terms of effectiveness required in-

rastructure deployment and coordination overhead The distributed

B Bellalta et al Computer Communications 75 (2016) 1ndash25 15

m

b

s

c

t

u

a

t

o

l

t

w

t

t

a

o

c

c

a

(

t

o

t

t

D

t

t

o

t

f

a

t

c

i

a

c

d

t

s

t

b

I

l

(

e

a

i

R

p

e

t

c

c

i

b

i

i

t

b

t

t

t

W

a

p

c

i

t

h

p

t

n

a

t

e

t

t

t

p

t

a

r

8

w

s

w

[

i

S

o

a

t

f

i

8

i

r

a

a

i

t

c

(

I

s

a

e

t

t

c

[

i

c

i

g

s

n

s

t

i

o

o

s

c

D

o

ethodology that is receiving more attention in the literature is

ased on distributed channel sensing solutions [127144145] These

olutions include the observation of spectrum use and possible

ooperative aggregation of spectrum sensing information to increase

he accuracy of detection and reduce the vulnerability of primary

sersrsquo transmissions The spectrum sensing capability is required in

ll the secondary devices with a minus114 dBm sensitivity Dynamic

ransmission power control must be provided and the upper limit

n emissions is 100 mW EIRP (20 dBm) for portable devices (further

imited in the case of adjacent channel use to reduce out-of-band in-

erference) However database (DB) coordination of TVWS combined

ith spectrum sensing is considered the most promising and effec-

ive technique compared with spectrum sensing alone The DB spec-

rum information has more chances to be effective (ie identifying

ll free channels opportunities) and reliable (ie not prone to attacks

r erroneous interpretation of channels status) In fact regulators de-

ided to push for the conservative solution of a remotely accessible

entralised spectrum DB that maintains the information on the avail-

bility of all TVWS channels at any given point in time and location

with a target accuracy around 50 m) under their responsibility For

his reason the spectrum databases have become a mandatory part

f many spectrum sharing systems and the dominant technical solu-

ion to support TV white spaces

The IEEE 80211af standard adheres to this vision and also has

he role of a common regulatory framework for different spectrum

B implementations defining a generalised coordination architec-

ure protocols and interfaces for spectrum queries and local spec-

rum information management The protocol explicitly considers out

f scope the communication protocol and technology adopted in

he background leaving the freedom to those players responsible

or the deployment to select any suitable Internet-based and local

ccess network technology Secondary devices that need to access

he TVWS spectrum should get information on which channels they

an effectively use and which parameters to adopt without impact-

ng primary usersrsquo transmissions (and receptions) by querying the

vailable spectrum DB The query must contain the transmission

haracteristics and accurate geographical position of the secondary

evice which can be determined by means of a geo-positioning sys-

em (GPS) regularly updated in the case of movement (or manually

et for static devices) The IEEE 80211af reference system architec-

ure for DB spectrum access is composed of multiple Geolocation-

ased Spectrum information Databases (GDBs) entities connected via

nternet to Registered Location Secure Servers (RLSS) RLSSs work as

ocal proxies of the GDB for a localised group of Basic Service Sets

BSS) RLSS are connected via a secure protocol architecture with the

quivalent of multiple Access Points (AP) in different BSS realising

trusted DB infrastructure APs locally coordinate the exchange of

nformation and channel access management between the GDB (via

LSS) and the secondary usersrsquo end stations (STA)

Mode I devices are those under the control of a device that em-

loys geo-location database access while Mode II devices are those

mploying geo-location database access by themselves The informa-

ion exchange provided between the GDB and a secondary user STA

an be provided in both an open-loop (eg adopted by FCC) and a

losed-loop (eg adopted by ETSI) implementation In an open loop

mplementation daily spectrum availability information is provided

y the GDB and no feedback on the spectrum information received

s provided by STAs thus the system approach for spectrum access

s much more conservative leading to low channelsrsquo utilisation po-

ential In a closed loop implementation the STA can provide feed-

ack to the GDB and there are more communication overheads due

o the high granularity of updates however the system is more effec-

ive and reliable in the exploitation of TVWS on behalf of STAs The

ypical information provided by the GDB includes (1) the updated

hite Space Maps (WSM) of frequencies allowed for secondary use

t the timespace of the querying STA and (2) the device-dependent

ower limitations for transmission (in general conservative and ac-

urate enough to avoid relevant interference effects on primary users

dentified in the area)

The basic IEEE 80211af mechanisms regulating the communica-

ion between STA and GDB can be found in [137138]

In parallel with IEEE 80211af standards such as the IEEE 19006

ave been created that consider interfaces and data structures sup-

orting spectrum sensing information exchange applicable to spec-

rum sensing (and particularly distributed spectrum sensing) sce-

arios In particular in October 2014 the IEEE 19006 WG initiated

project for a new standard called IEEE 19006b [146] concerning

he use of spectrum sensing information to support and optimise the

ffectiveness reliability and robustness of spectrum database solu-

ions The aim is to enhance the performance and capabilities of spec-

rum databases through the use of spectrum sensing information

Co-existence Support for co-existence mechanisms so that mul-

iple technologies can effectively utilise the TVWS spectrum is im-

ortant Self co-existence between network devices of a common

echnology (eg deployed by different operators in the same area)

nd co-existence among different technologies are relevant topic of

esearch for Cognitive Radio (CR) systems and specifically for IEEE

0211af Many solutions appeared on the research scene but no one

as so far finalised as the target solution for IEEE 80211af Specific

tandardisation has been started to regulate coexistence between

ireless standards of unlicensed devices including the IEEE 802191

147] The purpose of the IEEE 802191 standard is to enable the fam-

ly of IEEE 802 Wireless Standards to most effectively use TV White

pace by providing standard coexistence methods among dissimilar

r independently operated TVWS devices Early examples of gener-

lised coexistence mechanisms included Dynamic Frequency Selec-

ion (DFS) Transmission Power Control (TPC) listen before talk (eg

or contention based IEEE 80211 80215) time division multiplex-

ng (also among different techniques such as the IEEE 80216 80220

0222) and Message-based Spectrum Contention (that is beacon-

ng messages that carry coexistence information) Opportune met-

ics must be defined to assess the measurable coexistence achieved

mong different technologies as an example the hidden node prob-

bility for a target scenario or the estimate of percentage variation

n normalised network throughput and latency (before and during

he SU transmissions) On the other hand a centralised coexistence

ontrol mechanism could be effectively realised by a central manager

or coexistence-DB like the GDB) in critical scenarios To this end

EEE 19004 [148] (a standard for heterogeneous networks in dynamic

pectrum context part of IEEE Standards Coordinating Committee 41)

ims to standardise the overall system architecture and information

xchange between the network and mobile devices which will allow

hese elements to optimally choose from available radio resources

There are three possible classification of co-existence archi-

ectures and inter-network coordination channels for CR systems

entralised coordinated and autonomous co-existence mechanisms

149] In centralised co-existence schemes the co-existence is admin-

stered by a central entity (eg like in IEEE 802191) This solution

ould be applied in both homogeneous IEEE 80211af systems (eg

ncarnated by the centralised spectrum DB entity) and in hetero-

eneous inter-networks without requiring modifications to existing

tandards under the assumption that all the involved co-existing

etwork entities would adhere to the same centralised coordination

cheme and inter-network coordination protocol On the other hand

his assumption could be hard to satisfy in some practical scenar-

os The centralised schemes are effective in providing minimisation

f inter-network interference based on the availability and quality

f centralised co-existence information In coordinated co-existence

chemes the centralised entity could be present but not taking de-

isions in general The central entity could provide a co-existence

B with required information which can be used via in-band and

ut-band signalling for co-existence decisions taken by the cluster

16 B Bellalta et al Computer Communications 75 (2016) 1ndash25

c

e

n

c

o

s

i

d

a

S

8

n

a

b

o

b

i

t

a

8

b

[

t

i

n

b

5

t

a

c

b

s

a

c

n

s

p

i

c

l

r

c

f

a

v

t

a

e

c

t

Q

a

b

s

i

c

u

n

c

i

heads of many co-existing inter-networks for implementing a proper

common coordination protocol (eg out-band busy tone signalling)

These are often hybrid solutions realising a compromise between ef-

fectiveness and suitability In autonomous co-existence schemes all the

decisions and coordination are implemented in a distributed way

by the involved network entities In these schemes many policies

can be adopted to realise a sufficient level of co-existence under a

best effort approach Solutions that could be implemented include

the use of busy tones beaconing and other signalling protocols in

dedicated control channels dynamic distributed frequency selection

schemes listen before transmit token-based and dynamic reserva-

tion schemes All the above mentioned solutions have goods and

bads and a general illustration can be found in [149] Other proac-

tive co-existence techniques try to early detect and recovermitigate

co-existence issues with relaxed inter-network coordination eg re-

alised via spectrum sensing and interference avoidancesuppression

techniques

It must be clear that different aims exist regarding the co-

existence for primary TVWS users protection against secondary

users and for secondary devices mutual interference avoidance

Many co-existence problems have been analysed in the literature

for TVWS technologies in particular between the primary (licensed)

and secondary (unlicensed) devices The coexistence management in

TVWS is complex because primary users of DVB transmissions are in-

tended to be the pure receivers rather than the broadcasters In other

words potential protected users could be everywhere and the hid-

den terminal problem could arise for these systems The co-existence

problem is also complicated by many factors that cause asymme-

try and dynamicity in the TVWS such as the mobility variable den-

sity power asymmetry and heterogeneous MACPHY layers TVWS

co-existence problems originate between the IEEE 80211af technolo-

gies and the IEEE 80222 technologies for wireless regional area net-

works (WRAN) Results of analysis of TVWS usage in Europe show

that white spaces are typically present and fragmented They are typ-

ically more abundant in rural areas where larger contiguous blocks

of unused channels are available due to broadcast network planning

giving priorities linked to population density However the exploita-

tion of TVWS in urban areas is possible eg some recent research

was realised under the assumption of exploiting shadowing effects

created in the communication environment (eg by buildings obsta-

cles) in favour of frequency reuse [150]

Since different standards for opportunistic communication in the

TV White Spaces have now been published such as IEEE 80222 [151]

IEEE 802154m [152] Weightless [153] and of course IEEE 80211af

improved methods to guarantee the coexistence of different devices

operating on several protocols in the same bands must be deployed

We highlight here some work eg [150154ndash157] that focuses on

the coexistence between different technologies operating in the same

bands In particular [155] studies the performance degradation of

IEEE 80222 when an IEEE 80211af network is operated in the same

area The problem becomes even worse when the IEEE 80211af net-

work is located near 80222 user equipment causing both networks

to perceive a strong interference from each other One proposed solu-

tion is the Coexistence Beacon Protocol [158] studied for 80222 net-

works which foresees the exchange of a periodic beacon to identify

the neighbouring and possibly interfering networks

For what concerns the co-existence problems research has ad-

dressed other important related aspects of new WLANs on TV White

Spaces For instance the higher coverage range makes them attrac-

tive for applications in the smart grid and also for M2M on TVWS

[134150159] Here the increased range compared to standard tech-

nologies in the ISM bands can have beneficial effects for indoor

mobile devices [150] and M2M [159] Specifically indoor devices

communicating on TV White Spaces offer better propagation char-

acteristics and penetration through obstacles making it easier to

realise scalable home connectivity [135] although interference and

o-existence problems could be exacerbated and must be resolved

g see [139] However studies in the literature show how the sig-

al coming from an indoor transmitter on TV White Spaces remains

onstrained inside the house making it difficult to deploy indoor-to-

utdoor networks On the other hand this shadowing limit can be

een as an opportunity lowering the interference outside the build-

ng in which the TVWS network is locally operated [134150] Un-

er this approach HDTV streaming has recently been considered as

possible use case of WLAN networks operating in the TV White

pace [160]

Non-contiguous channel bonding A relevant novel feature of IEEE

0211af is the potential for contiguous and non-contiguous chan-

el bonding which permits aggregating basic channels (also non-

djacent ones) and leveraging the possible large frequency spread

etween multiple available channels Due to the rather static nature

f primary DTV transmissions where a busy channel is unlikely to

ecome free in the near future and state changes are coarse-grained

n general it is crucial to exploit the time-locality effect and exploit

he maximum physical channel availability that could be aggregated

t any given location With the methodology inherited from IEEE

0211ac IEEE 80211af is capable of bonding together two up to four

asic channels grouped in up to two different non-contiguous chunks

137] The spectrum bandwidth of a DVB-T basic channel can be ei-

her 6 7 or 8 MHz depending on the country in which the service

s operated As an example this creates a 144 to 168 OFDM chan-

elsrsquo bandwidth potential when up to four 6ndash7ndash8 MHz channels are

onded [137]

12 Open challenges

The deployment of WLANs in the TVWS at their maximum poten-

ial still requires the resolution of open problems Generally speaking

s mentioned in previous sections the challenges for such networks

an be divided into three main categories (1) spectrum sharing

etween opportunistic (secondary) and heterogeneous devices (2)

pectrum sensingmanagement and maximum exploitation of avail-

ble spectrum (potentially enabling the spectrum-on-demand con-

ept) and (3) co-existence and interference mitigation to the primary

etwork (primary user protection) and secondary networks [154]

The overall aim of spectrum sharing techniques is to maximise

eparation to avoid overlapping or contiguous channels operations

rovided that sufficient channels are available in the TVWS spectrum

n a given spacetime scenario [154] In general important research

ontributions have to be realised for the design of proper spectrum al-

ocation methodologies and techniques satisfying multi-factorial QoS

equirements at the system-level and the user-level Novel ideas in-

lude the spectrum sharing and spectrum sensing techniques being

urther divided into cooperative and non-cooperative

Non-cooperative solutions attempt to realise spectrum sharing

nd sensing on a local basis without direct cooperation between de-

ices Examples include the ldquolisten-before-transmitrdquo approach and

ransmission power control to limit the interference and spectrum

llocation policies that are based on local node feedback only Coop-

rative solutions rely on the existence of a common communication

hannel through which devices can tentatively agree on the spec-

rum allocation that provides the desired spectrum separation and

oS Major issues include the definition of a methodology to identify

common channel (or multiple channels mutually shared in space

y pairs of heterogeneous and distributed devices) and to efficiently

hare common information For the access to the spectrum DBs the

dentification of proper common access channels is still under dis-

ussion out-of-TVWS-band cellular communications can be widely

sed given their high coverage also indoor

Trying to avoid the use of complex and resource-hungry coordi-

ation schemes and dedicated control channels rendezvous proto-

ols play a significant role both for co-existence and for establish-

ng dynamic communication channels based on spectrum sharing

B Bellalta et al Computer Communications 75 (2016) 1ndash25 17

R

n

c

t

l

o

m

e

n

i

n

w

n

a

u

s

p

b

s

m

s

a

d

t

b

a

a

m

a

t

u

8

n

t

d

d

m

f

t

n

d

n

e

r

a

h

b

f

a

c

t

u

i

f

t

d

t

s

i

f

I

l

i

a

n

c

d

I

n

o

t

i

t

e

w

t

c

m

t

a

i

t

q

b

c

a

i

l

i

a

v

t

a

a

c

t

i

t

b

t

t

a

D

p

c

s

o

q

r

p

a

g

t

e

u

d

t

m

n

t

s

f

i

1

8

f

T

endezvous is considered a fundamental problem in generalised CR

etworks at the basis of many communication processes eg in-

luding neighbour discovery routing and broadcast Rendezvous pro-

ocols attempt to establish a new link for communication on a se-

ected and agreed frequency band (channel) identified within a set

f available resources (channels) thus creating the basis for com-

unication between two or more SUs This problem is often exac-

rbated by the high (and unknown) numbers of contending SUs high

umber of available channels and their ambiguous boundaries and

dentifiers and tough co-existence requirements satisfaction in cog-

itive radio systems [161] Preliminary approaches for rendezvous

ere based on the existence of a dedicated Common Control Chan-

el (CCC) agreed among all the SUs or the existence of a centralised

gent (decision maker) assigning channels to any pairs of requesting

sers This approach could be considered in the case of IEEE 80211af

olutions based on the spectrum DB implementation however the

roblem is complicated by the fact that the common channel must

e identified and made available for all the requesting SUs at the

ame time even when these are spread over multiple collision do-

ains in space On the other hand the scalability and reliability is-

ues which are caused by the congestion of CCCs and the vulner-

bility risks of centralised DBs motivate the research of alternative

istributed solutions called blind rendezvous protocols [162] Dis-

ributed solutions have been mainly based on beaconing andor slot-

ased channel hopping algorithms exploiting randomisation sensing

nd discoverysynchronisation messages spread in slotted time over

set of candidate available channels [161] These protocols aim to

inimise the convergence time to maximise the use of resources

nd to avoid collisions among SUs by identifying and agreeing on

he use of a given common channel among N non-overlapping and

nambiguously labelled channels The main complication in IEEE

0211af systems is given by the wide availability of many different

on-contiguous spectrum bands which cannot be uniformly quan-

ised in single channels and unambiguously labelled in space in or-

er to realise a common reference domain for all the IEEE 80211af

evices In fact the knowledge of the number of channels their com-

on labels the unique IDs of SUs and the number of SUs contending

or the channel assignment are the key factors determining the effec-

iveness and good properties of blind rendezvous protocols Recently

ew classes of rendezvous protocols have been proposed which are

istributed blind and oblivious Oblivious means that available chan-

els may be identified with different labels on behalf of the differ-

nt involved SUs A number of these distributed blind and oblivious

endezvous protocols have been proposed and have been referenced

nd analysed in [162] To conclude solutions for dynamic spectrum

andoff issues (eg in case PUs suddenly appear in a region) must

e resolved in advance to minimise the latency of re-establishing a

unctional communication channel eg by implementing a proactive

lternative rendezvous channel selection in background to ongoing

ommunication processes [163]

Another challenge is to provide cooperative techniques in a dis-

ributed vs centralised implementation by analysing their convergence

nder variable conditions overheads and tradeoffs In some cases as

n vehicular networks the exploitation of scenario characteristics and

actors such as the constrained mobility could help to realise effec-

ive dissemination of sensing information over an extended and pre-

ictable horizon under a cooperative approach [164ndash166] Coopera-

ion could be helpful in the realisation of the cooperative spectrum

ensing in combination with the Geolocation DB approach Another

nteresting direction of research is the investigation of mutual ef-

ects of coexisting cooperation-based vs non-cooperative techniques

n cooperation-based devices the exchange of information could al-

ow a quick convergence to a solution and avoidance of further

nterference [145] Certainly cooperative techniques have desirable

dvantages over non-cooperative ones However tight time synchro-

isation and high coordination overheads are required The common

hannel for cooperation purposes could become a bottleneck and re-

uce the potential advantages of dynamic spectrum allocation The

EEE 80219 foresees the presence of a shared common control chan-

el (CCC) to which devices need to tune in order to gather the state

f the network Solutions such as [141] propose the use of TVWS for

he CCC

A promising direction for research is the adoption of cluster-

ng schemes for secondary devices associated with a hybrid dis-

ributedcooperative vs centralisednon-cooperative approach in an

ffort to reduce overheads while maximising advantages In this

ay secondary nodes could implement cooperation for sensing func-

ions using properly identified common channels and delegate the

entralised decisions and spectrum-DB management to well instru-

ented cluster-leader nodes This concept is incarnated to some ex-

ent by RLSS in IEEE 80211af More recently there have been propos-

ls to build distributed databases that refresh their contents period-

cally by querying the remote spectrum database [167] Here Mas-

er devices (according to Ofcom terminology) periodically cache the

uery replies from the spectrum database in order to reply to and

roadcast the spectrum availability to neighbour Slave devices

Another challenging direction for research is the definition of ac-

urate models and efficient simulation tools enabling dynamic spectrum

nalysis for both rural and urban areas Recent developments of dig-

tal maps simulation tools and propagation models theoretically al-

ow improving the capability of predicting radio propagation effects

n complex scenarios with an acceptable computation time This en-

bling technology could be used in parallel to spectrum DBs to pro-

ide more accurate forecasting of dynamic frequency allocation in

imespace scenarios Another issue to be taken into account is the

ccuracy of the remote spectrum database regarding the channel

vailability estimation Several works have already shown the inac-

uracy of propagation models due to the complexity of parameters

o be considered and differences between modelled and real scenar-

os [132134168169] Better propagation models will make it possible

o more efficiently estimate the interference between devices and

uild Radio Environment Maps (REM) or White Space Maps (WSM)

hat will possibly lead to a more accurate sharing of the radio spec-

rum [144] To this end current research could focus on how to lever-

ge simulation models in order to build more accurate frequency

Bs

In general the adoption of remote spectrum DBs methodology

ushed by national regulators has made the spectrum sensing pro-

ess optional and conservatively realised (eg power control for

ensing-only devices is limited to below 17 dBm EIRP) Thus much

f the research has focused on techniques and solutions to successfully

uery the remote database also in challenging environments such as

ural areas and indoor However since the queries must provide the

osition of the mobile device the accuracy of positioning for mobile

nd handheld devices still needs more appropriate solutions Trian-

ulation can be used to estimate the position of mobile devices with

he accuracy needed by the current regulations outdoor (50 m) How-

ver a low GPS accuracy is possible indoors [170] Certified man-

al positioning can be provided by technicians for static and indoor

evices although limiting mobility Alternatives to costly infrastruc-

ures for positioning (eg based on short range beaconing devices)

ust be deployed in indoor scenarios

Finally regarding spectrum sensing and interference mitigation

ew challenges come from the heterogeneity and different charac-

eristics that networks operating in the TV White Spaces can have

uch as different transmitting power (4 W for static devices 100 mW

or mobile and portable devices and 40 mW for communication

n channels adjacent to occupied ones) different bandwidths (5

0 15 and 20 MHz are currently foreseen for IEEE 80211af IEEE

02154m can have narrower bandwidths or wider ones) and dif-

erent medium access schemes (protocols that use either CSMA or

DMA are required to co-exist in the TV White Spaces) With all this

18 B Bellalta et al Computer Communications 75 (2016) 1ndash25

d

a

T

i

a

c

t

l

b

o

t

a

d

p

a

r

p

n

a

m

t

p

p

d

d

o

t

i

D

w

p

i

a

a

d

b

fi

d

o

t

l

b

t

b

a

i

e

o

o

c

t

t

w

A

c

M

s

c

l

O

w

g

v

t

in place the problem of determining whether another secondary de-

vice is currently transmitting in the same channel is still an open

issue

Spectrum sensing is a very hot topic in the literature with many

proposals that leverage the cooperation between devices to raise the

accuracy of the detection process eg [144171] The current litera-

ture also offers room for proposals that suggest exploiting sensing to

build more accurate spectrum DBs [164172173] Lastly regarding the

interference mitigation to the primary network regulators provide a

conservative approach based on a remote spectrum database to avoid

high risk for the primary users However in particular for the indoor

scenario the concept of secondary usersrsquo transmission in TV Grey

Spaces has been proposed TV Grey Space identifies busy channels

that are formally used by primary users in the area of interest (eg

at the rooftop level of a building for DVB-T) but which could be con-

sidered usable indoors (eg at the basement level) without causing

effective interference to the primary users [134150174] This concept

is based on the generalised spectrum underlay paradigm [175] The

remaining research challenge is to provide TV Grey space access guar-

anteeing a sufficient amount of protection to the primary user while

providing a satisfactory QoS to opportunistic devices eg through

the use of customised and validated propagation models for indoor

TVWS networks and feedback mechanisms to constantly monitor the

interference generated to the primary network

6 Emerging new trends and technologies

In this last section we review three emerging trends that in our

opinion will have a large influence in the conception of future WLANs

as they change the way WLAN protocols and functionalities are de-

veloped implemented tested and integrated with other wireless

networks

61 Programmable Wireless LANs

Especially in the enterprise environment WLAN deployments

need to support a wide range of functionalities and services This is

intrinsically difficult because of the large number of APs that must be

managed which calls for scalable solutions Typical services include

channel assignment load balancing among APs authentication au-

thorisation and accounting (AAA) policy management support for

client mobility and interference coordination Another problem is the

fact that WLAN clients autonomously take several decisions such as

which APs to associate with when to hand-over etc Therefore sup-

porting roaming clients requires the management of a large number

of association states across several APs which is a challenge if sup-

port for real-time hand-over is desired Typically such management

schemes are centralised and most of them are proprietary such as

WLAN controller solutions from Aruba [176] or Cisco [177] although

the IEEE 80211u amendment has been released to allow mobile users

to seamless roam between WiFi networks with automatic authenti-

cation and handoff [178] For example Dyson [179] enables STAs to

send information such as radio channel conditions to a centralised

controller based on a custom API (eg based on Python) As the con-

troller has a centralised view of the network it can enforce a rich

set of policies to control the network also using historical informa-

tion A demo system has been implemented along with applications

such as airtime reservations for specific clients or optimised hand-

offs However Dyson requires STAs to be modified in order to use

those new services offered by the centralised controller TRANTOR

[180] is another example of a centralised management system that

requires changes of clients in order for the infrastructure to gather

information from them in terms of eg interference measurements

In addition control commands enable the infrastructure to exercise

control such as modifying the transmit power or influencing the as-

sociation procedure Clients still use standard CSMA MAC layer for

ata transmission In contrast to DenseAP [181] or DIRAC [182] which

re also based on a centralised controller but do not modify clients

RANTOR exercises much more control on clients by the use of a ded-

cated API which enables a significantly higher gain in coordination

nd thus capacity CENTAUR [183] is another example of a centralised

ontroller that centrally schedules hidden and exposed terminals in

he downlink while using standard CSMA MAC for uplink traffic and

egacy downlink traffic To achieve good performance it uses a fixed

ack-off packet staggering techniques and a hybrid data path that

nly schedules downlink transmissions towards hidden and exposed

erminals centrally All other traffic is sent using standard DCF with

standard back-off procedure In contrast to previous approaches it

oes not require any modifications to the clients but does require data

lane centralisation and it remains questionable how scalable such

solution is for large WLAN deployments with high PHY layer data

ates beyond 100 Mbps However new approaches for fast data plane

rocessing that are currently explored in the virtualisation commu-

ity which move all the packet processing into user space may be

n option to significantly increase speed Designing such manage-

ent systems raises several interesting research questions (a) Cen-

ralised infrastructures may incur high latency but are more simple to

rogram and maintain while designing a completely distributed ap-

roach that operates close to WLAN devices may result in significant

istributed coordination and consistency problems (b) what level of

istributed control is required in order to support the flexibility needs

f future WiFi based networks supporting a high level of mobility

Traditionally WLAN APs are built on proprietary operating sys-

ems that are tightly coupled with the hardware This design makes

t hard to create new applications on top of such networking devices

espite the fact that protocols and mechanisms are available that

ould greatly improve the utility of existing networks those new

rotocols are not deployed because the closed system design makes

t very difficult to extend their functionality An important aspect for

ll such new approaches is therefore how to provide open interfaces

nd open source to speed up innovation As an example Linux-based

evices are completely open source but in order to increase flexi-

ility much more work is needed in the area of open drivers and

rmware In fact the Atheros based ath9k driver has been the main

river of innovation in the open source community because it is the

nly driver that interworks with open firmware In general solutions

hat modify the OS driver focus mainly on programmable network

evel solutions for WLANs such as channel switching or handovers

etween APs While the ath9k driver is an excellent example in how

o speed up innovation it is still very cumbersome to support flexi-

le MAC engine reprogrammability with ath9k In contrast MAClets

nd Wireless MAC processors (WMP) [184] allow a much more flex-

ble reprogramming of MAC functionalities A MAC processor is an

ntity able to execute general MAC commands that specify the MAC

perations through a software-defined state machine The behaviour

f the MAC protocol can therefore be updated at run-time by simply

hanging the sequence in which those commands are executed (ie

he MAClets) As a proof-of-concept the authors of [184] implement

he proposed MAC processor solution in a commodity WLAN hard-

are card extending the basic DCF in three directions piggy-backing

CKs a pseudo TDMA and the use of multiple channels In [185] a

ontrol framework for this WMP system is also proposed to support

AClet code mobility ie for moving loading and activating MAC

oftware programs embedded into ordinary data packets (akin to the

apsule model of traditional active networks)

The difficulty in re-programming networking hardware has also

ed to the concept of Software Defined Networks (SDNs) based on the

penFlow protocol [186] The main idea of SDNs is to extend net-

orking devices with standardised APIs that allow third-party pro-

rammers to flexibly control the data path In addition SDNs pro-

ide higher level abstractions to network designers and programmers

hrough the use of a centralised control plane offered by a network

B Bellalta et al Computer Communications 75 (2016) 1ndash25 19

Fig 11 The CLOUDMAC architecture VAP stands for Virtual Access Point and WT

stands for Wireless Termination

c

a

c

t

m

W

n

t

t

c

p

t

u

h

a

i

e

s

o

c

a

a

f

T

r

s

a

s

p

o

M

C

i

T

t

fl

A

a

a

i

c

t

s

t

e

I

t

w

H

e

b

I

l

d

o

6

r

W

s

l

a

e

e

a

s

b

c

t

(

S

p

c

F

B

t

p

T

e

t

s

u

m

i

e

o

a

A

p

d

t

t

n

p

i

t

t

f

i

ontroller such as NOX [187] which allows reuse of components such

s topology discovery or network access control for different appli-

ations The SDN concept has recently been applied to WLAN archi-

ecture in order to enable fine grained control over mobility manage-

ent and data forwarding focusing on a programmable enterprise

LAN architecture An important part of the SDN architecture is the

etwork controller which provides a centralised view of (parts of)

he SDN enabled network and uses the OpenFlow protocol to install

he forwarding rules on SDN-enabled devices (routers switches ac-

ess points cellular base stations etc)

ODIN [188] is designed to support programmability in enter-

rise WLAN architecture by separating the association state from

he physical AP They implement Light Virtual Access Points (LVAP)

sing a Split-MAC approach where the infrastructure controls the

andover procedure among different WLAN APs By managing

ssociations through SDN controllers ODIN enables proactive mobil-

ty management and load balancing within the SDN enabled WLAN

nterprise network without the need for changes in the client WLAN

tack or IEEE 80211 MAC layer While ODIN requires agents to reside

n the APs that communicate with the Odin Master within the SDN

ontroller CLOUDMAC [189] is based purely on the concept of SDN

nd virtual APs Similar to ODIN CLOUDMAC implements a Split-MAC

pproach but in addition enables the processing of WLAN MAC layer

rames within a co-located Cloud using so-called Virtual APs (VAPs)

he physical APs in CLOUDMAC are lightweight WLAN APs that are

esponsible only for sending out IEEE 80211 based MAC layer ACKs to

tandard WLAN clients and tunnel WLAN MAC layer frames through

n SDN-enabled enterprise WLAN towards the VAPs As association

tates are kept in the VAPs fast mobility is supported using sim-

le OpenFlow forwarding rules Because of the additional processing

f IEEE 80211 WLAN MAC frames in the co-located Cloud CLOUD-

AC has higher latency than a standard WLAN deployment However

LOUDMAC offers a Webservice based API to third party applications

n order to program the enterprise WLAN and enable new services

he architecture of CLOUDMAC is depicted in Fig 11 It has been ex-

ended in [190] to support QoS management and in [191] to support

exible MAC management frame prioritisation based on IEEE 80211

system based on OpenFlow has been recently proposed in [192] to

llow the station to be associated with multiple APs simultaneously

nd to switch between APs with low overhead

An important aspect to consider in order to enable programmabil-

ty is backwards compatibility Several approaches (eg [180]) require

lients to be modified to utilise the features provided This is difficult

o do in practice because it requires changes in the operating system

oftware of all clients If not all clients can utilise those APIs it is ques-

ionable how usable the new architecture will be and how much ben-

fit in terms of aggregate performance such architecture will allow

n contrast the SDN based approaches are interesting in the sense

hat they do not require changes in the client WLAN stack and work

ith the standard IEEE 80211 MAC layer deployed within the clients

owever it remains questionable how scalable such solutions are For

xample an interesting open research topic is to evaluate the scala-

ility of approaches such as CLOUDMAC [189] In addition processing

EEE 80211 MAC frames within co-located private Cloud requires low

atency support from local Cloud solutions (such as OpenStack) in or-

er to reduce the IEEE 80211 MAC processing time which is an area

f active research

2 Prototyping and testing IEEE 80211 enhancements

Most of the new proposals for next-generation WLANs are cur-

ently only evaluated using mathematical analysis and simulation

hile both analysis and simulation are necessary to characterise and

tudy those enhancements in the initial design phase or to consider

arge-scale scenarios it is difficult to consider all practical aspects of

real-world scenario This can sometimes cause significant differ-

nces between what simulations and real experiments show How-

ver real experiments are challenging because of the high complexity

nd costs of building the new hardware and software for each specific

olution to test

To mitigate implementation complexity there are several flexi-

le hardware platforms where low-level MAC mechanisms can be

ompletely implemented in software Examples based on FPGAs are

he USRP (Universal Software Radio Peripheral) [193] and the WARP

Wireless Open-Access Research Platform) [194] As an alternative

ORA (Microsoft Research Software Radio) [195] works on general

urpose computers by taking some advantage from modern multiple

ore systems

A different approach is provided by OpenFWWF [196] Open-

WWF provides an open CSMACA firmware for specific models of

roadcom chipsets so the resulting firmware can be uploaded and

ested in real commercial hardware OpenFWWF implements a sim-

le State Machine (SM) for controlling the hardware in real time

he SM evolution is driven by a main loop that reacts to events by

xecuting specific handlers When a packet originally prepared by

he Linux kernel is ready in the NIC memory handler Packet_Ready

ets up the radio hardware according to the packet metadata sched-

les the transmission and jumps back to the main loop The Trans-

ission Engine (TXE) then takes care of accessing the channel ie

t decrements the back-off counter according to the DCF rules and

ventually starts the actual transmission This triggers the execution

f the TX_frame_now event that prepares the ACK time-out clock

nd finalises the MAC header If the ACK-frame is received or if the

CK time-out expires and the maximum number of attempts for this

acket is reached handler Update_Params resets the contention win-

ow to its minimum value or otherwise doubles it Finally it loads

he back-off counter with a fresh value

In the following we will review a list of selected papers that use

he OpenFWWF firmware and the WARP or USRP platforms to test

ew proposals for WLANs in real scenarios as examples A first im-

lementation of the groupcast mechanisms defined in IEEE 80211aa

s presented in [197] The authors modify the OpenFWWF firmware

o include those functionalities and evaluate them using a 30 STAs

estbed A second example of the use of OpenFWWF is the collision-

ree MAC protocol presented in [198] In [198] the authors design and

mplement in OpenFWWF a MAC protocol that is able to achieve a

20 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 12 LTE to WIFI offloading

a

n

c

t

[

d

a

W

v

t

t

o

b

f

S

g

l

i

I

s

I

b

n

s

m

a

p

d

c

N

o

v

t

t

n

l

i

a

t

a

collision-free operation by waiting a deterministic time after success-

ful transmissions before to start a new transmission This work shows

the experimental performance of a collision-free MAC (CF-MAC) pro-

tocol for WLANs using commercial hardware The results show that

the proposed CF-MAC protocol leads to a better distribution of the

available bandwidth among users a higher throughput and lower

losses than DCF

In [199] the authors use the WARP platform to test a variant of the

DCF ndash called IEEE 80211ec ndash that substitutes control packets such as

the RTS CTS and ACK with short detectable sequences Since those se-

quences are shorter than the control packets and can be detected cor-

rectly even at lower SNRs values a significant gain in performance is

achieved MIDAS (Multiple-Input Distributed Antenna Systems) [200]

implementation using the WARP platform shows the benefits of dis-

tributing the antennas over the area to cover instead of co-locating

them at the AP in terms of capacity when MU-MIMO is employed

The authors also propose a new MAC protocol to benefit from the

spatial reuse that the DAS allows taking as a basis the IEEE 80211ac

Experimental results show that their proposal is able to achieve up

to 200 gains versus the traditional approach where all antennas are

co-located at the AP

The USRP platform has recently been proposed to develop a first

implementation of IEEE 80211ah in order to obtain a preliminary ex-

perimental performance assessment of such a technology because

no commercial off-the-shelf hardware is yet available as the amend-

ment is still in progress [201] The USRP platform is also used in

[202] to evaluate TIMO (Technology Independent Multi-Output) a so-

lution to deal with high-power non-IEEE 80211 interferers in ISM

bands

To conclude this section it is worth mentioning here also to

MAClets and WMP as given such flexibility to implement and dis-

tribute new MAC protocols and other IEEE 80211 functionalities we

believe that if a use-friendly implementation of such a MAC proces-

sors framework is provided it would significantly contribute to the

development and testing of new MAC enhancements by the research

community

63 CellularWLAN interworking

Public hotspots that offer Internet access over a WLAN using IEEE

80211 technology are now nearly ubiquitous It is forecasted that the

cumulative installed base of WiFi hotspots worldwide will amount

to 551 million by 2018 excluding private hotspots (eg WiFi access

points deployed at home) [203] The sharp increase in the availabil-

ity of public WiFi was initially perceived by mobile cellular operators

as a threat due to the additional competition from wireline Internet

service providers or emerging crowdsourced WiFi networks such as

FON5 However as cellular operators are fighting to cope with the ex-

plosion of mobile data traffic created by the rising use of multimedia

content traffic over mobile devices [15] they are also starting to use

WLANs based on the IEEE 80211 technology to offload data from their

core and access networks In general mobile data offloading refers to

the use of complementary network technologies (in licensed or un-

licensed spectrum) for delivering data originally targeted to cellular

networks Intuitively the simplest type of offloading consists of ex-

ploiting connectivity to existing co-located WiFi networks and trans-

ferring data without any delay (Fig 12) Thus this offloading tech-

nique is known as on-the-spot offloading As a consequence of this

new trend the seamless integration of cellular (eg 3GLTE) and WiFi

technologies has attracted significant research interest in recent years

(see [204] for a survey) a few solutions have already been standard-

ised [205206] and roaming between cellular and WiFi is becoming

increasingly transparent to end users CellularWLAN interworking is

5 httpscorpfoncomen

fl

t

f

lso fostered by the support in the evolving 4G standards of heteroge-

eous network deployments (HetNets) in which the existing macro

ells are complemented with a number of small low-power base sta-

ions with the goal of increasing capacity in highly congested areas

207208] It is envisaged that small cells will be based on 4G stan-

ards (eg pico and femto cells) as well as IEEE 80211 technologies

nd multimode base stations that work simultaneously with LTE and

iFi are already entering the market

It is important to point out that LTE standards already support a

ariety of mechanisms that enable data offloading However most of

he existing solutions such as Local IP Access (LIPA) and Selected In-

ernet IP Traffic Offload (SIPTO) focus on data offloading at the core

f the network [205] For instance LIPA allows an IP-enabled mo-

ile device to transfer data to another device in the same pico or

emto cell without passing through the cellular access network while

IPTO enables the routing of selected IP data flows through different

ateways The only LTE offloading mechanism that supports seam-

ess interworking with IEEE 80211-based WLANs is IP Flow Mobil-

ty and Seamless Offload (IFOM) Specifically IFOM relies on Mobile

Pv6 technologies to allow a user terminal to simultaneously route

elected IP flows over different radio access technologies [209210]

n this way a user terminal can offload selected flows to a WLAN

ased on some operator-defined policy while keeping the LTE con-

ection running However to enable such an approach it is neces-

ary to support an entity in the cellular core network that can com-

unicate with the user terminal to exchange information about the

vailability and quality of neighbouring access networks as well as to

rovide the user terminal with predefined rules to manage the han-

over process This entity in current LTE standards is called the Ac-

ess Network Discovery and Selection Function (ANDSF) server [211]

ote that the problem of selecting the best communication technol-

gy in an heterogeneous wireless network has been extensively in-

estigated in the literature both with centralised as well as decen-

ralised approaches (see [212] for a survey) However the use of mul-

iple interfaces in parallel as well as per-flow offloading are relatively

ew concepts Thus the design of scalable and efficient network se-

ection strategies for the ANDSF framework is still an open issue For

nstance three offloading methods suitable for the ANDSF framework

re proposed in [213] which are based on coverage SNR and sys-

em load A reinforcement learning approach is designed in [214] that

llows multimode base stations to autonomously steer their traffic

ows across different access technologies depending on the traffic

ype the usersrsquo QoS requirements the network load and the inter-

erence levels A number of studies have also tried to quantify the

4 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Table 3

Performance requirements for different HD streaming applications

Type Max data rate Max latency

Uncompressed raw video 149 Gbits 100 ms

uncompressed HDTV 150 Mbits 150 ms

Blue-ray Disc 54 Mbits 200 ms

MPEG2 HDTV 192 Mbits 300 ms

MPEG4 HDTV 8ndash10 Mbits 500 ms

f

c

p

B

n

t

p

t

c

f

a

b

a

S

2

n

l

o

s

d

c

i

c

T

n

i

C

c

p

p

i

a

r

s

b

c

c

i

p

l

a

t

t

3

8

s

f

p

a

a

t

3

t

t

l

satisfy their requirements Four of the key use cases for next-

generation WLANs are discussed in the following subsections

21 High-quality multimedia content delivery

Our new mobile and portable devices are designed to handle rich

multimedia contents including high-definition video and images

Table 3 reports the requirements in terms of maximum data rate and

latency for some of the most common real-time video applications

[21] Key scenarios in which the support of real-time video trans-

mission is required of course include Internet TV and video stream-

ing Similarly scenarios in which multiple users connect to the same

wireless network to request different multi-media content at the

same time are increasing every day However not all multi-media

content is real time Stored video and image files can also be ex-

changed between different devices Those files can have sizes rang-

ing from a few Megabits to several Gigabits hence requiring a high

network transport capacity in order to provide a good Quality-of-

Experience to end users Although video encoding schemes exist that

offer substantial video compression efficiency such as H264MPEG-

4 AVC [22] WLANs must be able to achieve very high transmission

rates and have content-aware mechanisms that are specifically de-

signed for multi-media applications to ensure a satisfactory service

for multimedia delivery The mechanisms that are considered by var-

ious IEEE 80211 standardisation groups to satisfy those requirements

are described in Section 3 such as group-cast communication proto-

cols single and multi-user spatial multiplexing and channel bond-

ing among others to make the communication more efficient and of-

fer higher transmission rates The reference IEEE 80211 amendments

for high-quality multimedia content delivery are IEEE 80211aa IEEE

80211ac and IEEE 80211ax

22 Machine-to-Machine (M2M) communications

The almost ubiquitous presence of sensoractuator devices that

are able to interact with the environment has fostered the creation

of new services and applications Concepts such as smart cities and

smart grids are being developed on the basis of the existence of those

sensoractuator networks to achieve a more sustainable use of the

environmental resources and provide citizens with a higher quality

of life [2324]

In a classic sense Wireless Sensor Network (WSN) technologies

are used to collect data from spatially distributed sensor nodes and to

transmit the data over a multi-hop wireless network to a central sink

[25] The M2M paradigm is broadening the scope of the WSN con-

cept because it enables networked devices wireless andor wired as

well as services to exchange information or control data seamlessly

without explicit human intervention Clearly M2M communications

face most of the technical challenges of WSNs One of the main lim-

itations of WSNs and M2M systems is that the network nodes are

usually battery powered or have limited access to power sources De-

signing mechanisms and protocols to reduce their power consump-

tion with the goal of extending the network lifetime is therefore cru-

cial for the successful commercial take-up of these kinds of networks

Fortunately devices in M2M systems typically generate or consume a

limited amount of data per unit of time Thus they can spend a large

raction of their time sleeping This facilitates energy saving at the

ost of additional complexity for the channel access and networking

rotocols

Popular wireless protocol standards for M2M communications are

luetooth ZigBee and BT-LE [12] An alternative promoted by mobile

etworks is to connect devices in M2M systems directly to the In-

ernet by using the cellular network infrastructure for which specific

rotocols are being developed [26] WLANs are envisioned as an al-

ernative to both multi-hop WSNs and cellular networks However

urrent WLANs are not able to satisfy the minimum requirements

or M2M communications [13] Novel specific power-saving mech-

nisms are required to support the long periods of inactivity needed

y the sensoractuator devices and to manage the thousands of nodes

ssociated with a single AP These challenges will be discussed in

ection 4 when presenting the IEEE 80211ah amendment

3 Efficient use of the spectrum

The ISM bands are used by several wireless communication tech-

ologies including IEEE 80211 IEEE 802154 and Long Term Evo-

ution (LTE)-Unlicensed networks This results in a high spectrum

ccupancy Unfortunately wireless networks operating in the same

pectrum region can suffer from mutual interference which might

egrade the performance of all of them This is exacerbated by the un-

ontrolled deployment of wireless networks in the ISM band which

s typically very common in urban environments For example let us

onsider a building with several apartments and a WLAN in each one

here would easily be several WLANs operating in overlapping chan-

els and suffering mutual interference [27] To deal with this issue it

s expected that new APs will increasingly incorporate DCA (Dynamic

hannel Allocation) mechanisms to select and update their operating

hannel at run-time

An alternative approach to alleviate the spectrum occupancy

roblem is to move to a different part of the spectrum even if the new

art of the spectrum is occupied by communication systems operat-

ng under a license In that case WLANs would be the secondary users

nd therefore must avoid causing interference to the primary users In

ecent years the change from analogue to digital TV broadcast emis-

ions has resulted in a reorganisation of the spectrum at VHFUHF

ands This reorganisation has shown that there are many empty TV

hannels called TV white spaces that can be used for data communi-

ation especially in rural regions [28] Furthermore WLANs operat-

ng in those TV white spaces can take advantage of radio propagation

roperties in the UHF band to provide large coverage areas The chal-

enges to be addressed to use CSMACA protocols in VHFUHF bands

s well as how to obtain higher transmission rates when the spec-

rum is fragmented will be discussed in Section 5 when presenting

he IEEE 80211af amendment

High performance WLANs for multimedia applications

This section reviews the IEEE 80211ac IEEE 80211ax and IEEE

0211aa amendments These three amendments target multimedia

cenarios by introducing new physical-layer technologies and MAC

unctionalities to improve the WLAN capacity and QoS provision Ap-

lication examples include home scenarios in which a WLAN AP can

ct as an Internet gateway and wireless media server for home appli-

nces (eg IPTV set-top boxes projectors game consoles) and con-

ent storage devices A possible use case is illustrated in Fig 2

1 The IEEE 80211ac amendment

IEEE 80211ac [29] aims to provide users with a throughput close

o 1 Gbps which represents a roughly four-fold increase with respect

o IEEE 80211n [7] Compared to IEEE 80211n IEEE 80211ac supports

arger channel widths (up to 160 MHz) introduced a new modulation

B Bellalta et al Computer Communications 75 (2016) 1ndash25 5

Fig 2 High-throughput demanding multimedia devices associated to an IEEE 80211acax AP

s

(

3

s

w

b

i

t

s

t

c

o

f

C

t

A

p

c

f

u

c

i

t

u

i

t

t

I

s

m

[

a

t

u

s

T

e

S

Fig 3 A DL-MU-MIMO transmission in IEEE 80211ac Note that the AP transmits two

spatial streams in SU-MIMO mode to STA B and a single spatial stream to STA A Packet

aggregation is used in both transmissions The PHY header is transmitted ommnidirec-

tionally to inform the selected STAs about the next MU-MIMO transmission

n

C

c

F

D

t

t

s

h

d

d

b

m

k

c

M

A

D

d

3

n

(

e

i

i

I

cheme ie a 256-QAM modulation and downlink multiuser MIMO

DL-MU-MIMO)

11 Novel features

The most relevant new features included in IEEE 80211ac are de-

cribed in the following

Channel bonding IEEE 80211ac enables the use of channel band-

idths of 20 40 80 (mandatory) and 160 MHz (optional) Channel

andwidths larger than 20 MHz are created by ldquobondingrdquo (ie group-

ng) a group of consecutive 20 MHz channels and aim to offer higher

ransmission rates

Two extensions have been proposed in IEEE 80211ac for the ba-

ic DCF (Distributed Coordination Function) access method in order

o support channel bonding (i) the Static Bandwidth Channel Ac-

ess Protocol (SBCA) which always transmits over the same group

f 20 MHz channels and requires that all sub-channels are idle be-

ore starting a packet transmission and (ii) the Dynamic Bandwidth

hannel Access scheme (DBCA) which is able to dynamically adapt

he channel width to the instantaneous spectrum availability [3031]

s expected in dense scenarios the use of DBCA offer a much better

erformance than SBCA due to adaptability [32]

To avoid hidden terminals operating in any of the 20 MHz bonded

hannels the IEEE 80211ac amendment includes extended RTSCTS

rames in order to signal the maximum channel width that can be

sed at both the transmitter and the receiver In case the CTS in-

ludes a lower channel width than the RTS the transmitter will adopt

t Similarly to the ACK frames when the RTS and CTS frames are

ransmitted they are duplicated over all the 20 MHz sub-channels

sed The operation and performance of channel bonding in WLANs

s thoroughly analysed in [27] showing the new interactions be-

ween neighbouring WLANs that may appear and their impact in the

hroughput of each one

Downlink multiuser MIMO The main novelty introduced by the

EEE 80211ac amendment compared with the IEEE 80211n one is the

upport of MU-MIMO transmissions in the downlink hence allowing

ultiple simultaneous transmissions from the AP to different STAs

58] In the IEEE 80211ac amendment the AP can be equipped with

maximum of eight antennas and send up to four spatial streams

o two different users or up to two spatial streams to four different

sers at the same time

When an IEEE 80211ac AP performs a multi-user transmission it

pecifies the group of STAs to which that transmission is directed

his information is contained in the new IEEE 80211ac PHY head-

rs which are broadcast omni-directionally to all STAs The way

TAs are grouped is decided by the AP after obtaining the chan-

el state information (CSI) feedback from all STAs To gather the

SI information by the AP IEEE 80211ac considers only an explicit

hannel sounding feedback mechanism called Explicit Compressed

eedBack (ECFB) The channel access is governed by EDCA (Enhanced

istributed Channel Access) At each transmission attempt the mul-

iple access categories (AC) managed by the AP should contend for

he channel medium as only one AC can be served for each transmis-

ion attempt In the case that the queue associated with the AC that

as won the internal contention does not contain packets to enough

ifferent destinations to fill all the available spatial streams it can

ecide to share the remaining ones with the other ACs

Packet aggregation To increase the efficiency of each transmission

y reducing unnecessary overheads IEEE 80211ac allows the trans-

ission of several MPDUs aggregated in a single A-MPDU Then to ac-

nowledge each MPDU individually a Block ACK packet is used which

ontains a bitmap to indicate the correct reception of all included

PDUs Thus leveraging on the information contained in the Block

CK the transmitter is able to selectively retransmit only those MP-

Us that have failed instead of the whole A-MPDU Fig 3 illustrates a

owlink MU-MIMO transmission in which packet aggregation is used

12 Open challenges

Since the IEEE 80211ac amendment has recently been fi-

alised current research around it should cover two main aspects

a) understanding the performance bounds of IEEE 80211ac which

ntails the development of new models simulation tools and exper-

mental platforms of IEEE 80211ac-based WLANs and (b) propos-

ng specific solutions for those aspects that are not defined by the

EEE 80211ac amendment on purpose such as the mechanism for

6 B Bellalta et al Computer Communications 75 (2016) 1ndash25

g

i

l

a

b

b

f

o

w

o

H

t

e

n

a

f

t

m

a

w

s

3

a

8

2

c

s

m

W

r

a

[

o

m

2

n

c

p

r

a

creating the groups of STAs for DL-MU-MIMO transmissions smart

packet schedullers able to decide when the use of DL-MU-MIMO out-

performs SU-MIMO transmissions and the implementation of the

TXOP sharing feature between several ACs The results and conclu-

sions obtained in both cases will be very valuable in the development

of IEEE 80211ac technologies as well as in the conception of the fu-

ture amendments that will substitute IEEE 80211ac in four to five

years such as the recently initiated IEEE 80211ax

Following the first mentioned research direction there are several

efforts that have focused on understanding both theoretical and exper-

imental performance bounds of IEEE 80211ac The maximum downlink

throughput that an IEEE 80211ac AP can achieve when packet aggre-

gation channel bonding and different spatial stream configurations

are considered is presented in [33] In [34] the authors evaluate the

IEEE 80211ac performance experimentally using commodity devices

focusing on the effects that the use of wider channels the 256-QAM

modulation and the number of SU-MIMO spatial streams have in

terms of throughput and energy consumption It is worth mentioning

that DL-MU-MIMO was not yet implemented in the equipment they

were using and that feature was therefore not considered The evalu-

ation of a DL-MU-MIMO implementation for WLANs using the WARP

platform is presented in [35] where a deep evaluation of the potential

benefits of DL-MU-MIMO transmissions is done in terms of the loca-

tion of the receivers number of users and user mobility among other

aspects A solution that combines both packet aggregation and DL-

MU-MIMO transmissions is presented in [36] Results show the need

of properly dimensioning the buffer space to achieve the full potential

of such a combination In [37] the authors compare the throughput

achieved by IEEE 80211n and IEEE 80211ac when packet aggrega-

tion is used with and without channel errors They show that in most

cases the packet aggregation mechanism introduced in IEEE 80211ac

outperforms the one in IEEE 80211n An analytical model to evalu-

ate the performance of the IEEE 80211ac TXOP sharing mechanism in

DL-MU-MIMO communications is developed in [38] The main goal of

this study is to identify how the TXOP sharing mechanism could im-

prove the system efficiency while achieving channel access fairness

among the different ACs

How to optimally exploit the new DL-MU-MIMO capabilities pro-

vided by IEEE 80211ac is still an open challenge First due to the need

of frequent CSI exchanges between STAs and the AP it is not yet clear

in which conditions DL-MU-MIMO outperforms SU-MIMO [39ndash42]

or even whether MU-MIMO does or does not outperform multi-user

packet aggregation when the amount of data directed to each des-

tination is not balanced [43] Packet aggregation can be a solution

to balance the duration of the multi-user spatial streams as shown

in [36] although it will always depend on the amount of traffic di-

rected to each destination and the buffer capacity at the AP In [44]

the authors compare different strategies to assign the spatial streams

between the available destinations at each transmission in a fully

connected mesh network showing in ideal channel conditions the

theoretical benefits of MU-MIMO vs SU-MIMO

Closely related to the previous point a second open challenge is

the design of efficient schedulers that consider traffic priorities the

buffer state the different MIMO strategies TXOP sharing policies

grouping of STAs and the availability of fresh CSI feedbacks to max-

imise the throughput and guarantee the required QoS for each ac-

tive traffic flow It is important to consider that the availability of up-

dated CSI estimates from all STAs allows the AP to reduce the mutual

interference between the transmitted spatial streams which means

lower packet error probabilities and higher transmission rates How-

ever the overheads for obtaining the CSI from all STAs is large and

increases linearly with the channel sounding rate and the number

of STAs Proposals for reducing the CSI overhead are under develop-

ment For example in [40] the CSI overhead is reduced by inhibiting

the channel sounding whenever possible based on the estimation of

the channel stability for all users Another related problem is how to

roup the STAs as the goal is to find groups of STAs with compat-

ble (ie orthogonal) channels In [45] the authors show the chal-

enges inherent to the group assignment problem and they propose

n heuristic method to solve them TXOP sharing is considered in [46]

y presenting two alternative approaches to enhance the considered

ack-off procedure for the purpose of improving both throughput and

airness

A third key challenge for IEEE 80211ac is to achieve an efficient use

f the spectrum when several channel widths are used in scenarios

ith multiple overlapping WLANs Increasing the channel width the-

retically allows individual WLANs to achieve a higher throughput

owever the presence of other WLANs in the vicinity also increases

he chances of frequency overlapping which may cause the opposite

ffect as there appears inter-WLAN contention [27] Adaptive mecha-

isms to select the channel centre frequency and the channel width

nd MAC protocols to choose the instantaneous channel width used

or each transmission are thus required For instance in [47] the au-

hors focus on the channel selection problem when WLANs can use

ultiple channel widths using a game-theoretic framework In [48]

scheme is proposed to enable the communication between nodes

ith partially overlapping channels which may provide stronger re-

ilience to channel interferences

2 The IEEE 80211ax amendment

In 2014 the High Efficiency WLANs (HEW) Task Group [49] initi-

ted the development of a new IEEE 80211 amendment called IEEE

0211ax The IEEE 80211ax amendment is expected to be released in

019 and to some extent it will be the IEEE 80211 response to the

hallenges of future dense and high-bandwidth demanding WLAN

cenarios [5051]

The challenges in the development of the IEEE 80211ax amend-

ent are to

(i) Improve the WLANs performance by providing at least a four-

fold capacity increase compared to IEEE 80211ac

(ii) Provide support for dense networks considering both the ex-

istence of multiple overlapping WLANs and many STAs in each

of them Spatial reuse of the transmission resources is a must

(iii) Achieve an efficient use of the transmission resources by min-

imising the exchange of management and control packets re-

visiting the structure of the packets and improving channel

access and retransmission mechanisms among others aspects

(iv) Provide backward compatibility with previous amendments

This is achieved by the mandatory transmission of the legacy

PHY preamble in all frames and by keeping EDCA as the basic

channel access scheme

(v) Introduce effective energy saving mechanisms to minimise the

energy consumption

(vi) Support multi-user transmission strategies by further devel-

oping MU-MIMO and Orthogonal Frequency Division Multiple

Access (OFDMA) capabilities in both downlink and uplink

In addition to the aforementioned challenges next-generation

LANs will have to implement some other functionalities beyond the

aw packet transmission and reception Examples are a fast efficient

nd robust handoff between APs in the same administration domain

52] device-to-device communication (D2D) [53] and coordination

f multi-AP networks [54] In the first case the IEEE 80211ai amend-

ent called Fast Initial Link Setup is in progress and expected for

016 Its target is to complete a handoff in less than 100 ms including

ew AP discovery user authentication and configuration Using D2D

ommunication we can avoid the use of the AP as a relay hence im-

roving the overall efficiency as the number of packet transmissions

equired is reduced Finally the virtualisation of network functions

dds a new dimension in the management of multiple APs which in

B Bellalta et al Computer Communications 75 (2016) 1ndash25 7

d

e

I

w

f

t

s

3

P

s

G

t

8

p

C

c

m

c

i

t

a

s

w

f

t

t

e

l

p

e

a

H

d

t

c

o

s

p

n

t

b

a

w

c

t

w

n

8

a

c

w

d

H

n

a

h

p

w

i

t

fi

e

c

u

t

m

p

p

b

p

i

t

8

d

I

m

M

u

t

m

i

t

D

W

I

u

M

i

w

p

h

t

n

o

p

t

s

u

s

r

b

A

t

O

t

t

w

s

p

o

i

W

t

t

a

[

m

ense scenarios can contribute to notably improving the user experi-

nce We further discuss this last topic in Section 6

Different from the other amendments covered in this survey the

EEE 80211ax amendment is just in its initial stages of development

ith only very few technical aspects consolidated at this stage There-

ore in the following subsection we will overview both the new fea-

ures and open challenges of the IEEE 80211ax amendment at the

ame time

21 Novel features and open challenges

The IEEE 80111ax Task Group is currently working in four areas

HY MAC Multi-user and Spatial Reuse [49] Next we will overview

ome of the topics currently under discussion in the IEEE 80211 Task

roup in each category

PHY layer The IEEE 80211ax PHY layer will be an evolution of

he IEEE 80211ac one The challenges in the design of the IEEE

0211ax PHY layer are related with the extensions required to sup-

ort multi-user MU-MIMO and OFDMA transmissions and Dynamic

CA Also improvements in the supported modulation and channel

oding techniques will be likely considered to allow for higher trans-

ission rates at lower SNR values For example IEEE 80211ax may

onsider LDPC (Low-Density Parity Check) coding which is optional

n IEEE 80211ac instead of the traditional convolutional codes as

hey provide a coding gain of 1ndash2 dB Moreover the PHY layer may

lso include some functionalities to support the use of Hybrid ARQ

chemes to improve the efficiency of packet retransmissions

Medium acess control In order to keep backward compatibility

ith previous IEEE 80211 amendments besides a common PHY

rame preamble compatible MAC protocols are required This means

hat it is likely that EDCA will be kept as the main channel access

echnique in the IEEE 80211ax amendment Therefore the most rel-

vant open challenges are related to EDCA extensions to support a

arge number of STAs improve traffic differentiation capabilities im-

rove the energy consumption and provide mechanisms to fairly co-

xist with neighboring wireless networks

To support a large number of contenders with a low collision prob-

bility a simple solution is to use large backoff contention windows

owever it would also increase the time a STA is in backoff so re-

ucing the number of packets it can transmit A solution to reduce

he backoff duration when using large contention windows is to de-

rease the duration of the backoff slots Latest technological advances

n electronics should require less time to perform a CCA check to

witch between reception and transmission modes and to process a

acket thus enabling such a possibility

Another approach to enhance the underlying CSMACA mecha-

ism in EDCA and achieve a higher efficiency is to consider decen-

ralised collision-free MAC strategies Those MAC protocols are able to

uild collision-free schedules thus improving the network efficiency

s collisions are reduced while preserving backward compatibility

ith the default EDCA implementation An overview of several de-

entralised collision-free MAC protocols can be found in [55] Among

hem CSMAECA [56] is especially relevant since it is fully compatible

ith EDCA and latest upgrades made it adaptive to the instantaneous

umber of contenders in a single WLAN

In addition to EDCA IEEE 80211ax WLANs can rely on the IEEE

0211aa amendment to further improve the WLAN traffic differenti-

tion capabilities with intra-AC traffic differentiation and groupcast

ommunication mechanisms among other features In Section 33

e will provide further details about the IEEE 80211aa amendment

IEEE 80211ax will likely keep the same channel widths that were

efined in the IEEE 80211ac amendment ie 20 40 80 and 160 MHz

owever it is expected that IEEE 80211ax will extend current chan-

el bonding mechanisms to further improve the spectrum utilisation

nd the coexistence between neighbouring WLANs For example it

as been shown in [32] that the use of dynamic channel bonding

rovides significant throughput gains in dense scenarios compared

ith the static approach while minimising the inter-WLAN negative

nteractions [27] Furthermore additional mechanisms are required

o fully exploit the use of channel bonding such as the use of ef-

cient algorithms to select the position of the primary channel or

ven to consider the existence of multiple primary channels to in-

rease the number of bonded channel combinations that a node can

se for transmission

The MAC layer in IEEE 80211ax may work with the PHY layer

o implement an efficient Hybrid ARQ mechanism able to retrans-

it only incremental redundancy bits in short packets Opportunistic

iggy backing of data packets in ACKs and vice versa may further im-

rove the efficiency of IEEE 80211ax WLANs by reducing the num-

er of transmissions in a bidirectional data exchange [57] Finally

acket headers can be reduced if shorter STA identificators are used

nstead of MAC addresses and unnecessary fields are avoided in all

ransmissions

Finally in terms of power saving mechanisms it is expected IEEE

0211ax will follow the same trends as in previous IEEE 80211 stan-

ards including the extensions currently under development for the

EEE 80211ah amendment that are discussed in Section 4

Multi-user Multi-user communications will likely be one of the

ain characteristics of IEEE 80211ax as both uplink and downlink

U-MUMO and OFDMA are under consideration The use of multi-

ser communication techniques does not necessarily represent a sys-

em capacity increase because the available transmission resources

ay be the same as in the single-user communication case However

n WLANs the simultaneous transmission from different users is able

o parallelise the large temporal overheads of each transmission (ie

IFS SIFS ACKs packet headers etc) which can notably improve the

LAN efficiency

IEEE 80211ax will further develop the MU-MIMO capabilities of

EEE 80211ac by allowing multiple simultaneous transmissions in the

plink which is known as uplink (UL-) MU-MIMO [58] Similar to DL-

U-MIMO transmissions an open challenge to enable UL-MU-MIMO

s to design a mechanism able to efficiently schedule the users that

ill transmit at the same time In one hand a pure decentralised ap-

roach would be easy to implement with minimal signalling over-

eads However since it requires that all STAs finish their backoff at

he same time it may be very inefficient besides that those STAs may

ot be compatible in terms of their respective spatial channels In the

ther hand a pure centralised approach requires that the AP has com-

lete CSI and buffer occupancy information from all STAs to select

he most suitable group to perform a multi-user transmission Once a

uitable group of STAs is selected by the AP a ldquoTriggerrdquo frame may be

sed to notify the group of selected users that can initiate a transmis-

ion This approach guarantees efficient multi-user transmissions but

equires some extra overheads to collect all the required information

y the AP and signal the selected STAs In both cases new multi-user

CKs will be likely introduced by IEEE 80211ax to acknowledge all

ransmissions with a single control packet

Multi-user OFDMA is also in the agenda for IEEE 80211ax Using

FDMA a channel can be split in several sub-channels and assigned

o different users Likely OFDMA will be implemented in combina-

ion with channel bonding where each of the 20 MHz subchannels

ill be assigned to a different user in both downlink and uplink Be-

ides that a similar operation as in the multi-user MIMO case is ex-

ected as there are almost the same challenges to solve A survey

f current OFDMA proposals for WLANs is presented in [59] show-

ng also how the use of OFDMA is able to significantly improve the

LAN efficiency Fig 4 illustrates an uplink OFDMA and MU-MIMO

ransmission

In addition to Multi-user MIMO and OFDMA the use of Simul-

aneous Transmit and Receive (STR) techniques commonly known

s full-duplex transmission have been suggested for IEEE 80211ax

5051] Using STR a pair of nodes is able to transmit and receive si-

ultaneously [60] which theoretically doubles the channel capacity

8 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 4 Multiuser uplink transmission using two different OFDM subchannels Trans-

missions from STA B and C over the same subchannel represent an uplink MU-MIMO

transmission All transmissions are acknowledged with a single Multiuser ACK The

trigger frame is used to signal the selected STAs

t

t

t

a

s

t

f

f

c

3

s

s

t

n

n

A

d

w

t

fi

a

t

8

i

d

t

r

w

d

R

v

f

s

d

p

B

c

e

k

r

G

a

c

m

t

b

d

3 The concealment address protects legacy stations ie GCR-incapable stations

from receiving duplicated group-addressed frames

The challenge is that both the AP and the STA involved in a full-duplex

transmission have to start to transmit at the same time To achieve

that information about full duplex transmission capabilities can be

included in RTSCTS control packets to set-up a full-duplex transmis-

sion between the transmission initiator and its destination

Spatial reuse Dense WLAN deployments are necessary to offer a

continuous coverage with high transmission rates To improve both

the co-existence with those neighboring networks and the spatial

reuse of the spectrum a WLAN has two options (i) minimise its

area of influence by reducing its transmit power and (ii) accept

higher interference levels by increasing the Clear Channel Assess-

ment (CCA) level Use of both techniques may increase the number of

concurrent transmissions between neighbouring WLANs and there-

fore their capacity although it may also result in the opposite effect

since the achievable transmission rates may be negatively affected by

the higher interference levels observed which is the main challenge

to solve

Due the high WLAN dynamics the use of adaptive systems is cru-

cial but challenging as adaptivity requires extra complexity in terms

of computing and memory resources and there are no guarantees

that the implemented solution converges due to the decentralised

operation of each WLAN The use of DSC (Dynamic Sensitivity Con-

trol) to dynamically adjust the CCA level is one of the aspects cur-

rently under discussion in the IEEE 80211ax Task Group First studies

evaluating the performance of DSC for IEEE 80211ax WLANs show

a clear improvement on the spatial reuse and the area throughput

[61] Another example of the achievable throughput gains obtained

by adapting the CCA level can be found in [62] where the authors

show that gains of up to 100 can be achieved Moreover trans-

mit Power Control (TPC) to mitigate interference between WLANs in

dense scenarios is studied in [63] showing the need of jointly opti-

mising both TPC and CCA to maximise the network performance

Finally sectorisation by using beamforming is also under consid-

eration for the development of the IEEE 80211ax amendment as a

potential solution to improve spatial reuse [64] Using sectorisation

only the nodes of a given area are allowed to receive or transmit data

hence reducing the contention between different networks whenever

they activate non-overlapping sectors A challenge here is to coor-

dinate the different neighboring APs when they belong to different

administration domains Decentralised learning approaches may be

implemented to find feasible temporal patterns of non-overlapping

sectors

33 The IEEE 80211aa amendment

As discussed above legacy IEEE 80211 standards do not provide

robust and efficient delivery of audiovideo streaming services Thus

he IEEE 80211aa amendment was developed to include new fea-

ures and additional mechanisms to improve the performance of real-

ime multi-media content delivery [65] Specifically IEEE 80211aa

ddresses the following five shortcomings of previous 80211

tandards [1666]

(i) the lack of reliable and efficient support for multicast and

group communications

(ii) the incapacity of applying traffic prioritisation to different

multimedia streams or different types of frames from the same

stream

(iii) the absence of methods for cooperative resource sharing

among neighbouring APs

(iv) the lack of mechanisms for graceful degradation of audiovideo

streaming quality

(v) the non-interoperability with existing IEEE 8021 standards for

Audio Video Bridging (AVB)

In the following sections we present in detail the solutions to

hose problems introduced in the IEEE 80211aa amendment We

urther discuss the research studies that have provided the basis

or the IEEE 80211aa design and we identify the remaining open

hallenges

31 Novel features

Groupcast communication mechanisms In most audiovideo

treaming applications a group of clients must receive the same

tream simultaneously A multicast protocol is necessary to avoid that

he same content is replicated throughout the network In wireless

etworks multicast transmission can exploit the intrinsic broadcast

ature of the wireless channel ie broadcast transmissions from an

P are physically received by all other stations in the same collision

omain However multicast and broadcast frames in IEEE 80211 net-

orks are not protected by an acknowledgement mechanism as in

he case of unicast frames Thus layer-2 multicast transmissions de-

ned by legacy IEEE 80211 standards are unreliable and not suit-

ble for streaming applications To partially address this limitation

he Direct Multicast Service (DMS) was first specified in the IEEE

0211v amendment [67] Basically DMS converts multicast streams

nto unicast streams In this way frames destined to a multicast ad-

ress are individually transmitted as unicast frames to the stations

hat joined that multicast group Obviously DMS provides the same

eliability as unicast transmission services but the consumed band-

idth increases linearly with the number of group members To ad-

ress this scalability issue IEEE 80211aa includes the Groupcast with

etries (GCR) service in addition to DMS Specifically the GCR ser-

ice defines new mechanisms and the related management frames

or group formation which allows a set of stations to agree on a

hared (non-multicast) address called the groupcast concealment ad-

ress3 Furthermore the GCR service specifies two retransmission

olicies GCR Unsolicited Retry (GCR-UR) and GCR Block Ack (GCR-

A) When using GCR-UR the AP can proactively retransmit all group-

ast frames a number of times to mitigate the impact of channel

rrors (see Fig 5a)) while receivers are not requested to send ac-

nowledgements Intuitively this approach improves transmission

eliability but it still suffers from scalability issues In contrast when

CR-BA is used the AP sends a burst of consecutive groupcast frames

nd it requests the receivers to reply with a Block ACK frame which

ontains a bitmap to positively or negatively acknowledge trans-

itted frames (see Fig 5b)) The Block ACK mechanism defined for

he GCR-BA service is quite flexible because Block ACK frames can

e requested immediately after a transmission burst or after a ran-

omised back-off delay Furthermore the AP can request the Block

B Bellalta et al Computer Communications 75 (2016) 1ndash25 9

Fig 5 GCR service with different retransmission schemes

A

r

o

l

o

e

v

e

v

d

f

d

s

v

c

s

a

a

f

d

t

F

t

d

v

a

f

c

t

t

o

t

i

c

r

v

i

r

s

o

t

a

(

p

i

f

a

o

i

t

t

c

s

fi

o

c

I

n

i

a

t

a

t

f

i

B

T

f

t

a

g

a

a

i

s

a

d

s

p

m

m

i

a

t

b

t

i

(

r

a

S

i

c

s

c

T

Q

p

a

3

g

t

4 TSPEC is a traffic specification sent from a QoS capable wireless client that requests

a certain amount of network traffic from the AP for the traffic stream it represents

CK frame to all groupcast recipients or only to a subset of them to

educe overheads and delays The advantages of the GCR methods

ver broadcast and DMS have been extensively demonstrated in the

iterature [1668]

Intra-access category prioritisation The IEEE 80211e amendment

nly allows traffic differentiation between four different access cat-

gories (ACs) that are broadly mapped to four application classes

oice (VO) video (VD) best-effort (BE) and background (BK) How-

ver there is a variety of streaming services ranging from simple

ideoconferencing to HD streaming over IPTV systems which have

ifferent QoS requirements (see Table 3) To provide the ability to dif-

erentiate among individual streams IEEE 80211aa includes an ad-

itional scheduling layer with respect to IEEE 80211e IEEE 80211aa

plits each one of the transmission queues associated with voice and

ideo ACs into a primary and an alternate queue In this way spe-

ialised scheduling rules can be applied to decide which queue to

erve when the EDCA function for inter-AC collision resolution grants

n access opportunity to voice or video ACs To facilitate the man-

gement of service level agreements IEEE 80211aa follows the de-

ault mappings between user priority values and traffic types that are

efined in the IEEE 8021D standard [69] It is then straightforward

o further map traffic types onto transmission queues and ACs (see

ig 6) Finally it is important to point out that the intra-AC differen-

iation functionality can be used to provide more sophisticated traffic

ifferentiation than simple stream prioritisation For instance most

ideo applications use Scalable Video Coding (SVC) schemes that en-

ble the partitioning of a video sequence into multiple layers with dif-

erent qualities and rates [70] Typically an SVC-based video stream

ontains a base layer which provides a basic level of quality and mul-

iple enhancement layers which can only be decoded together with

he base layer to improve the video quality Thus the different layers

f the same encoded video steam can be easily mapped to different

ransmission queues to receive differentiated QoS [71]

Stream classification service The stream classification service (SCS)

s an optional service that can be provided by an AP to the asso-

iated stations to classify multimedia streams based on arbitrary

ules that are established directly by the stations instead of the con-

entional 8021D user priorities To this end the station request-

ng the use of SCS must initiate an SCS session by sending an SCS

equest frame to the AP which contains an identifier for the SCS

tream and the descriptor of the classification rule The AP may accept

r reject the requirements specified by the station Once accepted

he AP must assign all frames that match the classification rule to

specific AC When intra-access category prioritisation is enabled

see Section 331) calternate transmit queues so that finer grained

rioritisation can be applied Finally there is also a Drop Eligibil-

ty Indicator (DEI) bit in the SCS descriptor that indicates whether

rames from this traffic stream can be dropped in the case that there

re insufficient resources Specifically frames with the DEI bit set to

ne have a higher probability of being discarded because their max-

mum number of allowed retries is smaller than the default Note

hat how to combine intra-AC queues and frame dropping settings

o achieve graceful degradation of the audiovideo stream quality in

ase of bandwidth shortage is beyond the scope of the IEEE 80211aa

pecification

Overlapping Basic Service Set (OBSS) management Network densi-

cation ie a denser deployment of wireless infrastructure nodes is

ne of the key strategies that is used nowadays to easily increase the

apacity of wireless systems even for indoor WLANs [72] However

EEE 80211 networks have a limited number of orthogonal chan-

els available and even if optimised frequency planning is applied

t might happen that neighbouring APs are mutually interfering and

station may affect multiple overlapping BSSs In this case conges-

ion not only increases but it is also likely to observe an unfair us-

ge of wireless capacity with the channel retained by one AP for long

ime intervals This is mainly due to the neighbourhood capture ef-

ect ie hidden terminal phenomena among APs [73] To address this

ssue IEEE 80211aa specifies a new functionality called Overlapping

SS (OBSS) management which is based on two new mechanisms

he first defines a set of parameters to quantify the load and inter-

erence among neighbouring BSSs such as medium occupancy frac-

ion number of admitted audiovideo streams data traffic volumes

nd the number of BSSs that are using the same channel as the tar-

et one Note that the traffic load consists of two components the

llocated traffic which is derived on the basis of the TSPEC values of

dmitted streams4 and predicted traffic which is evaluated by track-

ng the maximum value of the allocated EDCA and HCCA traffic over

even-day periods Once load measurement reports are exchanged

mong the APs a second OBSS component is responsible for coor-

inated admission control procedures on the basis of two suggested

haring schemes proportional sharing and on-demand sharing The

urpose of both schemes is to keep the total allocated traffic below a

aximum value in order to provide some QoS protection to admitted

ultimedia streams Finally IEEE 80211aa recommends implement-

ng additional OBSS management procedures for channel selection

nd cooperatively creating HCCA schedules that do not collide

Interworking with IEEE 8021AVB Audio Video Bridging (AVB) is a

erm commonly used to denote a set of technical standards developed

y IEEE to support real-time streaming services with bounded latency

hrough IEEE 802 networks [74] This objective is achieved by spec-

fying mechanisms to allow the synchronisation of multiple streams

IEEE 8021AS [75]) and traffic shaping (IEEE 8021Qav [76]) and to

eserve network resources for specific audiovideo streams traversing

bridged local area network by using a signalling protocol called the

tream Reservation Protocol (SRP) (IEEE 8021Qat [77]) IEEE 80211aa

ntegrates the SRP operations with the EDCA admission control pro-

edures Specifically the SRP RequestResponse messages are encap-

ulated in the management frames that are used to carry the traffic

haracteristics and the QoS requirements during admission control

his enables the end-to-end management of resource reservation for

oS guaranteed streams even when one or more IEEE 80211 links are

art of a path from the stream producers (called IEEE 8021Q talkers)

nd the stream consumers (called IEEE 8021Q listeners)

32 Open challenges

In recent years several MAC enhancements have been investi-

ated to improve QoS guarantees for real-time multimedia applica-

ions in IEEE 80211 networks [20] and the IEEE 80211aa standard

10 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 6 Stream classification and inter-AC traffic prioritisation

a

l

t

r

n

a

t

n

i

t

i

t

r

d

p

t

t

e

m

b

t

a

i

s

o

s

a

e

t

i

f

e

m

8

t

I

i

fi

which was finalised in 2012 included several of these proposed im-

provements Significant research efforts have focused on improving

the transmission reliability of multicasting by integrating ARQ mech-

anisms in IEEE 80211-based multicast transmissions Modifications

to the legacy MAC protocol were proposed in [78] to enable the

RTSCTS option in multicast mode and to select one or more multi-

cast receivers (called leaders) for acknowledging multicast data pack-

ets However these enhancements require changes to the standard

specifications The main problems of leader-based ARQ schemes are

leader election and the trade-off between scalability and reliability

The authors in [79] propose selecting the multicast recipient operat-

ing in the worst channel conditions as the unique leader but this ap-

proach may perform poorly in lossy environments In the Batch mode

multicast MAC (BMMM) [80] all multicast recipients are polled by

the multicast originator to send individual ACKs but this scheme is

not suitable for large multicast groups The Enhanced Leader Based

Protocol (ELBP) is proposed in [81] on the basis of multiple ACK-

leaders and block acknowledgement techniques Analytical models

are then developed to help select optimal ACK-leaders to meet ap-

plication QoS requirements However the models apply only to sat-

urated traffic while multimedia streams are typically bursty Another

class of reliable multicast protocols relies on busy tones to reduce

packet losses due to collisions [82] but the additional radio inter-

face needed for the busy tone limits the practicality of such solu-

tions An alternative approach to avoid collisions of multicast pack-

ets is the multicast collision prevention (MCP) scheme [83] which

is based on the use of a shorter waiting time for transmitting mul-

ticast packets An interesting approach is also proposed in [84] to

retransmit lost packets using an online linear XOR coding algorithm

However a modification to the standard MAC protocol is required to

enable simultaneous ACK transmissions In summary several differ-

ent methods have been proposed to improve multicast transmission

reliability by integrating ARQ schemes into the protocol architecture

but there are not conclusive results on which is the best solution The

choice of the most efficient mechanism depends on a variety of in-

terdependent factors such as loss ratios channel congestion multi-

cast group size and QoS requirements of multimedia streams A com-

prehensive analytical framework is needed to optimise the setting of

the parameters for each scheme and to dynamically select the best

one

As discussed above one main difference between unicast services

nd multicast services in the legacy IEEE 80211 standard was the

ack of acknowledgements Another critical difference is that mul-

icast frames must be transmitted using a fixed rate in the basic

ate set while the transmission rate of unicast frames can be dy-

amically adapted to the channel and traffic conditions [85] Thus

group of research papers has investigated the use of rate adap-

ation to improve the throughput of multicast services in IEEE 80211

etworks [7186ndash89] For instance the authors in [86] propose us-

ng RTS frames to allow group members to estimate channel condi-

ions Each member will then send a dummy CTS frame with a length

nversely proportional to channel quality In this way the multicast

ransmitter can use the collision duration to predict the lowest data

ate that can be used for group transmissions The overhead intro-

uced by this mechanism is quite high however The solution pro-

osed in [88] called ARSM also relies on feedback messages sent by

he multicast receivers called multicast response frames to identify

he group member exhibiting the poorest channel conditions How-

ver in this case a different back off timer is associated with each

ulticast receiver depending on the SNR of previously received feed-

ack messages in order to prevent collision An approach similar to

he one employed in the Auto Rate Fallback (ARF) protocol a rate

daptation scheme originally proposed in [90] is used in [87] Specif-

cally the number of successful consecutive transmissions and con-

ecutive transmission failures are used to decide when to increase

r decrease the transmission data rate respectively A modified ARF

cheme is also proposed in [71] which can be applied to videos that

re encoded into two layers namely the base and enhancement lay-

rs However how to integrate rate adaptation with the different re-

ransmission policies that are defined in IEEE 80211aa is still an open

ssue

One research area that is expected to be crucial in the success-

ul development of IEEE 80211aa-based products is the design of

fficient scheduling algorithms for supporting voicevideo traffic Al-

ost all research work in this field has been triggered by the IEEE

0211e amendment that enhanced the original IEEE 80211 MAC with

wo new QoS-aware access mechanisms ie EDCA and HCCA [91]

n principle with a well-designed admission control and schedul-

ng scheme HCCA is able to provide hard QoS guarantees to traf-

c flows [9293] However HCCA is rarely implemented in IEEE

B Bellalta et al Computer Communications 75 (2016) 1ndash25 11

8

c

c

a

s

b

w

o

j

t

f

a

Q

t

t

A

p

t

s

a

f

s

H

a

i

c

i

t

s

f

s

s

m

i

t

s

l

t

t

[

t

i

o

b

r

W

v

m

8

s

i

o

p

q

4

c

e

t

d

a

t

m

Fig 7 WLANs for M2M communications STAs represent sensor and actuator devices

4

t

w

s

a

c

g

e

(

p

o

t

p

9

r

o

a

p

c

o

i

p

b

f

t

s

o

i

4

8

f

t

i

s

m

d

0211e-based WLANs owing to its higher complexity and cost con-

erns Instead EDCA is widely adopted Most papers have thus fo-

used on improving EDCA performance Many papers have proposed

nalytical models for various subsets of EDCA functionalities For in-

tance a saturation-based performance analysis is conducted in [94]

y differentiating the minimum back-off window size the back-off

indow-increasing factor and the retransmission limit The authors

f [9596] also model AIFS differentiation while the model in [97]

ointly captures all the four EDCA parameters for traffic differen-

iation More recent papers have analysed the EDCA performance

or non-saturated conditions and for arbitrary buffer sizes [98] The

uthors in [99] have developed an analytical model to predict the

oS levels that can be achieved once a new voicevideo flow is in-

roduced in the WLAN A Kalman filter is proposed in [100] to ob-

ain estimates on the number of active transmission queues of each

ccess Category in EDCA These analytical models can then be ex-

loited to derive the optimal configuration of the EDCA parameters

o achieve given performance criteria or to design admission control

chemes that preserve QoS constraints For instance a scheme that

ssigns contention-window values to achieve pre-defined weighted-

airness goals is proposed in [101] A control-theoretic scheme is de-

igned in [102] with the goal of minimising the video traffic delay

owever most of these solutions rely on non-realistic assumptions

bout video traffic dynamics An alternative class of solutions dynam-

cally updates the EDCA parameters based on the observed network

onditions In [103] the EDCA parameters are optimised consider-

ng a WLAN with rigid and elastic traffic simultaneously analysing

he interactions between both types of traffic The authors in [104]

pecify several bandwidth-sharing mechanisms with guaranteed QoS

or voice and video traffic Measurement-based admission control

chemes are proposed in [105] A TXOP adaptation method is de-

cribed in [106] that takes into account video frame sizes and trans-

it queue lengths However the main drawback of these solutions

s that they are based on heuristics and hence do not ensure op-

imal and guaranteed performance Finally a third category of re-

earch papers tries to improve video performance by designing cross-

ayer scheduling approaches Specifically these works take advan-

age of multi-layer video encoding to classify the frames according

o their importance and assign them to different access categories

107] For instance the authors in [108] define classifiers and waiting

ime priority schedulers that dynamically change the packet prior-

ties according to end-to-end delay measurements A disadvantage

f this approach however is that an additional adaptation layer may

e needed to implement the complex interactions that are typically

equired between the video coding applications and the MAC layer

e conclude this section by pointing out that existing studies pro-

ide the basic design principles and techniques for improving multi-

edia streaming performance in IEEE 80211 networks Still the IEEE

0211aa standard poses new research challenges that have not been

ufficiently explored and that will require innovative solutions For

nstance scheduling between primary and alternate queues is still an

pen research area as the mapping of individual frames to multi-

le queues in order to achieve graceful degradation of voicevideo

uality [16]

Sensor Networks and machine-type communications

As discussed in Section 2 M2M communications refer to any

ommunication technology that enables sensoractuator devices to

xchange information and perform actions without the manual assis-

ance of humans This section reviews the main features currently un-

er consideration in the development of the upcoming IEEE 80211ah

mendment which targets the main challenges of those networks as

he IoT in general such as the energy consumption or the manage-

ent of many devices

1 The IEEE 80211ah amendment

The IEEE 80211ah amendment [109] aims to provide WLANs with

he ability to both manage a large number of heterogeneous STAs

ithin a single BSS and minimise the energy consumption of the

ensor-type battery-powered STAs

The initial design requirements of the IEEE 80211ah amendment

re detailed in [110] these entail the support of up to 8192 STAs asso-

iated with a single AP the adoption of efficient power saving strate-

ies a minimum data rate of 100 kbps the operation in the license-

xempt sub 1 GHz band and a coverage up to 1 km in outdoor areas

see Fig 7 for an illustrative example) A preliminary assessment of

erformance of the IEEE 80211ah technology in terms of the number

f STAs that can be effectively supported in a single WLAN as well as

heir energy consumption is presented in [111]

IEEE 80211ah operates over different sub-1 GHz ISM bands de-

ending on country regulations 863ndash868 MHz in Europe 902ndash

28 MHz in the US and 9165ndash9275 MHz in Japan China South Ko-

ea and Singapore also have specific channelisations Channel widths

f 1 MHz and 2 MHz have been adopted although 4 8 and 16 MHz

re also supported in some countries IEEE 80211ah furthermore pro-

oses new PHY and MAC layers The IEEE 80211ah PHY layer can be

onsidered to some extent a sub-1 GHz version of the IEEE 80211ac

ne At the physical layer OFDM is the chosen modulation method us-

ng 32 or 64 tonessub-carriers that are spaced by 3125 kHz The sup-

orted modulations include BPSK QPSK and from 16 to 256-QAM A

road range of antenna technologies ranging from single-user beam-

orming to MIMO and DL-MU-MIMO which was first introduced in

he IEEE 80211ac amendment are also included in the IEEE 80211ah

pecification Similarly the IEEE 80211ah MAC protocol include most

f IEEE 80211 main characteristics further extending its power sav-

ng (PS) mechanisms

11 Novel features

This section overviews the extensions introduced by the IEEE

0211ah amendment to the IEEE 80211 PS mechanisms to account

or the specific characteristics of resource-constrained sensor and ac-

uator devices A more detailed review can be found in [110] includ-

ng a performance assessment of IEEE 80211ah in several of the key

cenarios for M2M communications such as agriculture and animal

onitoring smart metering and industrial automation plants In ad-

ition a detailed survey of the IEEE 80211ah is reported in [112]

12 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 8 IEEE 80211ah PS mode for TIM-STAs non-TIM STAs and Unscheduled STAs

a

d

w

i

a

a

c

m

i

a

e

t

s

m

t

t

u

b

i

I

s

d

f

a

b

s

a

b

S

T

c

t

w

I

4

a

a

which addresses aspects not considered in [110] such as the use of

sectorisation to avoid overlapping between multiple IEEE 80211ah

WLANs and the use of relays to further extend the IEEE 80211ah cov-

erage among other aspects Finally in [113] both IEEE 80211ah PHY

and MAC characteristics are introduced with emphasis on the ben-

efits of using the sub-1 GHz ISM band in terms of link budget the

location of the OFDM pilots for outdoors operation the use of bidi-

rectional transmission opportunities and the new packet formats to

minimise overheads among others

Enhanced power saving mechanism PS mechanisms for WLANs

were already considered in the development of the first IEEE 80211

standard with the goal of improving the lifetime of battery equipped

devices [114] In PS mode STAs keep the transceiver in sleeping mode

as much time as possible They periodically wake up to listen to the

beacons transmitted by the AP Those beacons indicate whether an

STA has packets waiting for it at the AP In the positive case that STA

remains awake and requests the delivery of those packets Otherwise

given it has nothing to receive it returns to sleep mode until the next

beacon is expected

In the IEEE 80211ah amendment time is divided into pages DTIM

(Delivery Traffic Indication Map) periods TIM (Traffic Indication Map)

periods and slots DTIM and TIM periods begin with the correspond-

ing DTIM and TIM beacons sent by the AP The functions of DTIM and

TIM beacons are described below

1 DTIM beacons They inform as to which TIM Groups (ie the

group of STAs assigned to the same TIM period) have pending

packets at the AP

2 TIM beacons Each TIM message informs a TIM Group about

which specific STA has pending data in the AP Between two

consecutive DTIMs there are as many TIM beacons as TIM

Groups

Using this DTIMTIM-based approach any STA can enter into a

power saving state if it does not have packets pending for transmis-

sion and one of two conditions is met (1) it observes in the DTIM

beacon that there is no downlink traffic addressed to its TIM Group or

(2) it observes in the DTIM beacon that there is some downlink traffic

addressed to its TIM Group but that STA does not explicitly appear in

the TIM beacon Compared to the preliminary IEEE 80211 PS mecha-

nism this approach reduces the size of the Traffic Indication Map in

each TIM beacon thus reducing the overhead and the time STAs need

to listen and process them In addition TIM periods can be organised

in pages which further increases the number of TIM groups between

two DTIM beacons

The temporal organisation of pages DTIM and TIM periods is re-

ported in Fig 8 which shows the division of TIM periods in RAW (Re-

stricted Access Window) and PRAW (Periodic RAW) The RAW is a

time interval in each TIM period where TIM stations can transmit and

receive data (see below the different types of STAs defined in IEEE

80211ah) In addition it can be divided into several downlink and

uplink slots for further granularity In the downlink the slots are as-

signed to a single STA or a group of STAs while the uplink slots are

randomly selected by the STAs with packets ready for transmission

The PRAW is the period of time in each TIM where non-TIM stations

can transmit and receive data

Types of STAs IEEE 80211ah supports three types of STAs TIM

non-TIM and Unscheduled STAs A TIM station is assigned to a TIM

Group Their data transmissions must be performed within a RAW

(Restricted Access Window) period Non-TIM stations do not have to

listen to beacons to transmit data During the association process

non-TIM devices directly negotiate with the AP to obtain a trans-

mission time allocated in a PRAW The following channel access can

either be renegotiated or occurs periodically depending on the re-

quirements set by the STA Unscheduled STAs do not need to listen to

any beacons similar to non-TIM stations Even inside any restricted

access window unscheduled STAs can send a poll frame to the AP

sking for immediate access to the channel The response frame in-

icates an interval (outside both restricted access windows) during

hich unscheduled stations can access the channel This procedure

s meant for STAs that transmit data very sporadically

Hierarchical station organisation To support a large number of STAs

nd their organisation in pages DTIM and TIM periods IEEE 80211ah

ssigns to each associated STA a unique identifier of 13 bits which is

alled the Association Identifier (AID) Using this new AID the maxi-

um number of supported STAs is increased from the original 2007

n IEEE 80211 to 8191 (= 213 minus 1) in IEEE 80211ah However it also

llows categorising STAs according to the type of application they are

xecuting their power level or even their desired QoS by assigning

hem to different TIM groups

Long sleeping periods IEEE 80211ah offers TIM Non-TIM and Un-

cheduled STAs the possibility to set very long doze times (up to

onths) The corresponding clock drift produced by such long doze

imes must be taken into consideration however as the higher the

ime an STA has been asleep the further in advance it should wake

p to avoid possible synchronisation problems with the network

Efficient small data transmission Three new enhancements have

een proposed to reduce the overhead when the data packet size

s small First while IEEE 80211 contains a 28-byte MAC header

EEE 80211ah proposes a short 18-byte version by using AIDs in-

tead of MAC addresses Second IEEE 80211 has defined several null

ata packet (NDP) frames which consist only of a PHY header These

rames can be used to create short ACKs short Block ACKs short CTSs

nd short PS-Polls Finally a Fast Frame Exchange mechanism has

een developed so that if an STA has data to transmit it can notify a

uccessful reception by transmitting its data frame instead of an ACK

Sectorisation Since the PHY layer is based on the IEEE 80211ac

mendment single and multi-user beam-forming are also supported

y IEEE 80211ah This allows the transmission of data to multiple

TAs simultaneously in the downlink increasing the system capacity

he use of the beam-forming capability of IEEE 80211ah APs is also

onsidered as a means to group STAs into different independent an-

enna sectors with the main goal of reducing interference issues This

ould be particularly useful in the case of overlapping with other

EEE 80211ah WLANs or in the presence of hidden nodes

12 Open challenges

A first open challenge is to understand the coverage and achiev-

ble transmission rates in IEEE 80211ah WLANs in both indoor

nd outdoor scenarios Propagation models for WLANs working at

B Bellalta et al Computer Communications 75 (2016) 1ndash25 13

f

c

8

p

R

p

i

m

l

a

m

c

p

p

f

m

t

a

e

p

t

m

a

l

T

m

a

i

n

t

t

t

u

p

c

w

f

a

t

8

a

m

a

s

l

a

m

b

m

d

S

i

a

o

a

R

t

n

G

fi

s

t

p

t

c

u

c

w

i

o

p

g

a

c

i

f

c

(

d

m

l

i

R

t

l

t

L

m

a

m

o

e

f

a

I

i

t

w

i

t

a

c

w

p

f

t

d

5

i

s

e

s

w

5

a

c

a

n

requencies lower than 1 GHz are evaluated in [115] The authors

ompare two path loss propagation models proposed by the IEEE

0211ah Task Group (one for macro and one for picohotzone de-

loyments) [116] with Lee and HatandashOkumura propagation models

esults show that the IEEE 80211ah channel models underestimate

ath loss with respect to Lee and Hata models Moreover in a compar-

son with empirical data it is observed that the IEEE 80211ah channel

odels also underestimate the initial loss and the slope of the path-

oss function A new model parameterisation is thus proposed by the

uthors In [117] the feasibility of an IEEE 80211ah WLAN deploy-

ent is also evaluated in terms of the achievable range and bit rate

omputed on the basis of the link budget using the same path loss

ropagation models as in [115] Results show that the transmission

ower limitations in the uplink can limit the overall network per-

ormance Finally in [118] the authors evaluate the achievable trans-

ission range for the different transmission rates and the achievable

hroughput for different combination of transmission power values

nd transmission rates Further studies are required due to the het-

rogeneity of scenarios in which IEEE 80211ah WLANs can be de-

loyed to characterise the effective coverage ranges with special at-

ention to mixed indoors and outdoors scenarios and in presence of

obile nodes

Due to the wide coverage IEEE 80211ah WLANs may be severely

ffected by the presence of hidden terminals To mitigate such a prob-

em smart solutions to guarantee that the STAs assigned to the same

IM Group are inside each other transmission range or the employ-

ent of RTSCTS-based solutions may contribute to mitigate such

problem An overview of long-range scenarios for IEEE 80211ah

s presented in [119] including a detailed description of the hidden

ode problem in them

Single-hop uplink transmissions from distant STAs require high

ransmission power to reach the AP Since not all STAs may be able

o transmit at the required power a solution could be the introduc-

ion of nodes able to relay transmissions from them However the

se of relays has to be efficiently harmonised with the operation in

ower saving mode allowing temporal periods in which the relays

an gather the data from their associated STAs and then periods in

hich they forward the data to the AP which is still an open challenge

or IEEE 80211ah WLANs In case relays are used data prediction and

ggregation techniques could be implemented to make more efficient

ransmissions [120]

Another challenge is the assessment of the efficiency of IEEE

0211ah PS mechanisms including their optimisation In [111] the

uthors evaluate the capacity of the TIM and page segmentation

echanisms in terms of the maximum number of nodes supported

nd the energy consumed given a certain network traffic profile Re-

ults confirm that a large number of STAs can be supported with

ow energy consumption The impact of the number of nodes oper-

ting in PS mode on the energy consumption and the delay perfor-

ance is analysed in [121] A detailed analysis of the TIM Group-

ased channel access adopted by the IEEE 80211ah task group is

ade in [122] where the STAs assigned to each TIM are uniformly

istributed between the slots of the RAW Since only one group of

TAs is allowed to transmit in each RAW slot the channel contention

s minimised The number of downlink and uplink slots in a RAW

s well as their duration is optimised on the basis of the number

f active STAs and the network traffic profile in [123] In [124] the

uthors propose an algorithm to determine the optimal size of the

AW interval for uplink transmissions based on the estimation of

he number of STAs transmitting and the duration of the RAW Fi-

ally in [125] the authors address the contention problems in Smart

rid communication networks with many nodes and periodic traf-

c when using an IEEE 80211ah WLAN It is still a challenge to con-

ider smart systems able to adapt the IEEE 80211ah parameters to

he instantaneous system state in order to save as much energy as

ossible

Since the PS mechanisms affect only the operation of TIM STAs

he performance of non-TIM and Unscheduled STAs has not yet been

onsidered in the literature To identify the scenarios in which the

se of non-TIM and Unscheduled STAs are of interest is still an open

hallenge as further investigating how non-TIM STAs negotiate over

hich PRAWs they can use to transmit and receive data Co-existence

ssues between the three types of STAs in a single network is also an

pen challenge that requires further work to guarantee a minimum

erformance level for all of them

The design of strategies to distribute the STAs between all the TIM

roups based on their specific traffic profile position battery level

nd application priority is another important challenge that is still

ompletely open EDCA is the default channel access scheme included

n the IEEE 80211ah amendment and provides some basic traffic dif-

erentiation capabilities at the packet level Besides the different ac-

ess categories (ACs) needing to be renamed and their parameters

Arbitration Inter-Frame Spacing or AIFS CWmin TXOP duration) up-

ated to fit the specific traffic profiles of M2M communications other

echanisms can be implemented A mechanism to assign the down-

ink RAW slots when packets from multiple priority levels are wait-

ng for transmission at the AP is required for example Also uplink

AW slots are currently selected randomly by the STAs that want to

ransmit a packet A mechanism to reserve some slots at each up-

ink RAW for high priority STAs may therefore be an option although

his may severely degrade the performance of the low priority STAs

astly different repetition patterns for different priority TIM groups

ay also allow priority STAs assigned to those priority TIM groups to

ccess the channel more often We expect this challenge will receive

uch attention in the upcoming years because of the heterogeneity

f sensors that will be connected though a single IEEE 80211ah AP

specially in urban scenarios

Similar to what has been discussed for 4G cellular networks used

or M2M communications [126] the development of mechanisms to

void and resolve congestion situations is another open challenge for

EEE 80211ah WLANs with many nodes For instance a mechanism

n which STAs have to wait a random number of DTIM periods before

hey can start a transmission may be implemented This approach

ould effectively distribute the traffic over a longer period of time

n overload conditions but otherwise would unnecessarily increase

he access delay Adaptive and load-aware solutions may play again

n important role to keep the optimal WLAN operation in all possible

ircumstances

Finally since IEEE 80211ah will compete with 4G5G cellular net-

orks and WSNs to provide M2M connectivity in many different ap-

lication domains such as smart cities or e-health comparative per-

ormance studies to determine the strong and weak points of each

echnology are required including also aspects such as the cost of the

evices and the system reliability

Cognitive radio technology for TV White Spaces

This section gives an overview of the IEEE 80211af amendment

ntroducing its most relevant features and open challenges Fig 9

hows the basic components and features considered for WLANs op-

rating in the TV White Spaces (TVWS) which include local spectrum

ensing by the AP and STAs and the use of geolocation data bases

ith information on channel availability

1 The IEEE 80211af amendment

Thanks to the transition from analog TV to digital TV several VHF

nd UHF spectrum channels used for decades for analog TV broad-

asting are now unused or are under-utilised in many geographical

reas [127ndash129] This ldquodigital dividendrdquo of spectrum has suggested

ational regulators such as the FCC in the US and Ofcom in the UK

14 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 9 WLAN operating in TVWS basic elements and functionalities

Fig 10 Hidden terminal problem in TVWS when the secondary network uses only in-

stantaneous channel sensing to decide whether or not a TV channel is occupied When

the STA is not able to detect the TV signal due to the presence of obstacles it may

initiate a transmission and create interference to the primary users

s

c

5

a

i

p

t

o

o

t

e

c

u

l

m

b

a

m

t

e

t

t

t

m

D

c

I

n

b

u

8

c

a

t

i

t

w

f

to discuss how to reuse these channels for unlicensed devicesrsquo com-

munications [130131] These potentially vacant channels in the VHF

and UHF bands are referred to as TVWS [132] and include spectrum

portions like the 470ndash790 MHz in Europe and non-contiguous 54ndash72

76ndash88 MHz 174ndash216 MHz 470ndash698 MHz and 698ndash806 MHz in USA

A snapshot of the TV spectrum occupancy in the city of Barcelona in

2012 is shown in [133]

The attractive characteristics of TVWS (not only for WLANs) in-

clude the ability to penetrate through walls and other obstacles much

more effectively than other widely used spectrum bands such as the

24 and 57 GHz ISM bands [134ndash137] This fact along with a progres-

sive spectrum scarcity in ISM bands has suggested the birth of the

IEEE 80211af amendment published in February 2014 [138] which

provides the IEEE 80211 operational characteristics for TVWS access

of unlicensed White Space Devices (WSD) A good summary of IEEE

80211af can be found in [137] The main advantage of operating IEEE

80211 WLANs in the TVWS comes from an increased coverage range

which can reach up to 1 km in rural areas and open fields [139]

and less energy needed to transmit However this comes at the cost

of an increased interference risk to other WSDs which creates co-

existence problems and demands new PHY and MAC layers to ef-

ficiently support channel access and operations preserving licensed

usersrsquo devices

WLANs operated in the TVWS could cover a number of interest-

ing and emerging use cases and scenarios eg Internet access in ru-

ral or sparsely populated areas Smart Grid sensor aggregation me-

tering and control Internet of Things advanced WLAN operations

and TVWS traffic offloading in indoor environments [140141] Nev-

ertheless the concept could be realised to create or extend com-

mercial or municipality WiFi services offered to citizens with cover-

age at the whole city level realised with reasonably limited network

infrastructures [142143]

IEEE 80211af will use the PHY and MAC layers derived from IEEE

80211ac It will adopt concepts such as the OFDM multi-user beam-

forming contiguous and non-contiguous channel bonding and packet

aggregation Among the mandatory and innovative behavioural and

operational parameters the most notable one is the channel acqui-

sition support realised through remote geolocation-based spectrum

allocation databases which maintain the channelsrsquo availability infor-

mation in any given area and time of day providing upon request the

list of free channels available for use

In the following section we highlight and summarise three main

novelties that IEEE 80211af introduces In Section 511 we explain

how the access infrastructure to the remote spectrum database is de-

signed and what the requirements are we discuss the coexistence is-

sues and methodologies adopted in IEEE 80211af and we also discuss

the novel concept of non-contiguous channel bonding The following

ections also contain an illustration of related works Open research

hallenges will be summarised in Section 512

11 Novel features

This section provides an illustration of novel features introduced

nd discussed in the path to IEEE 80211af and the directions provided

n related works and standardisation initiatives to resolve classical

roblems properly characterised in the new framework of TVWS

echnologies

Channel acquisition spectrum database and channel sensing Much

f the attention in operating in the TVWS is given to the protection

f the primary licensed users in the TVWS spectrum band In general

he primary users were considered as the Digital TV (DTV) broadcast-

rs and receivers In fact receivers would suffer the wasteful effect of

ollisions during the TV broadcast reception in the case of secondary

sersrsquo (SU) transmission interference (Fig 10)

It is a well-known problem that the definition of a coherent wire-

ess channel status (idlebusy) from the viewpoint of distributed pri-

ary users (eg DTV receivers) is made difficult by the difference

etween transmitters and receivers interpretation of signals in a vari-

ble time-space collision domain resulting in hiddenexposed ter-

inals concepts Solutions proposed to reduce the hiddenexposed

erminals included the RTSCTS mechanisms and beaconing How-

ver it is difficult to implement solutions that would effectively de-

ermine the channel status of TVWS channels based on localised dis-

ributed sensing activities performed by multiple secondary users in

he same collision domain On the other hand the use of RTSCTS

echanism would not be effective in this case because most of the

TV primary users (and specifically the receivers) were not originally

onceived to transmit signals or implement the RTSCTS handshake

n conclusion the actions to identify free (and busy) TVWS chan-

els should not be based on primary usersrsquo involvement but should

e almost totally implemented by specifically designed secondary

sersrsquo methodologies

There has been much debate on the methodology that IEEE

0211af secondary devices should adopt to get knowledge of free

hannels at a given time on a given target communication area in

n effort to guarantee verified optimised and reliable TVWS spec-

rum use The three main methodologies discussed include sens-

ng solutions geolocation databases (DB) and beaconing Moreover

wo approaches are possible in general distributed and centralised

ith their well known tradeoffs in terms of effectiveness required in-

rastructure deployment and coordination overhead The distributed

B Bellalta et al Computer Communications 75 (2016) 1ndash25 15

m

b

s

c

t

u

a

t

o

l

t

w

t

t

a

o

c

c

a

(

t

o

t

t

D

t

t

o

t

f

a

t

c

i

a

c

d

t

s

t

b

I

l

(

e

a

i

R

p

e

t

c

c

i

b

i

i

t

b

t

t

t

W

a

p

c

i

t

h

p

t

n

a

t

e

t

t

t

p

t

a

r

8

w

s

w

[

i

S

o

a

t

f

i

8

i

r

a

a

i

t

c

(

I

s

a

e

t

t

c

[

i

c

i

g

s

n

s

t

i

o

o

s

c

D

o

ethodology that is receiving more attention in the literature is

ased on distributed channel sensing solutions [127144145] These

olutions include the observation of spectrum use and possible

ooperative aggregation of spectrum sensing information to increase

he accuracy of detection and reduce the vulnerability of primary

sersrsquo transmissions The spectrum sensing capability is required in

ll the secondary devices with a minus114 dBm sensitivity Dynamic

ransmission power control must be provided and the upper limit

n emissions is 100 mW EIRP (20 dBm) for portable devices (further

imited in the case of adjacent channel use to reduce out-of-band in-

erference) However database (DB) coordination of TVWS combined

ith spectrum sensing is considered the most promising and effec-

ive technique compared with spectrum sensing alone The DB spec-

rum information has more chances to be effective (ie identifying

ll free channels opportunities) and reliable (ie not prone to attacks

r erroneous interpretation of channels status) In fact regulators de-

ided to push for the conservative solution of a remotely accessible

entralised spectrum DB that maintains the information on the avail-

bility of all TVWS channels at any given point in time and location

with a target accuracy around 50 m) under their responsibility For

his reason the spectrum databases have become a mandatory part

f many spectrum sharing systems and the dominant technical solu-

ion to support TV white spaces

The IEEE 80211af standard adheres to this vision and also has

he role of a common regulatory framework for different spectrum

B implementations defining a generalised coordination architec-

ure protocols and interfaces for spectrum queries and local spec-

rum information management The protocol explicitly considers out

f scope the communication protocol and technology adopted in

he background leaving the freedom to those players responsible

or the deployment to select any suitable Internet-based and local

ccess network technology Secondary devices that need to access

he TVWS spectrum should get information on which channels they

an effectively use and which parameters to adopt without impact-

ng primary usersrsquo transmissions (and receptions) by querying the

vailable spectrum DB The query must contain the transmission

haracteristics and accurate geographical position of the secondary

evice which can be determined by means of a geo-positioning sys-

em (GPS) regularly updated in the case of movement (or manually

et for static devices) The IEEE 80211af reference system architec-

ure for DB spectrum access is composed of multiple Geolocation-

ased Spectrum information Databases (GDBs) entities connected via

nternet to Registered Location Secure Servers (RLSS) RLSSs work as

ocal proxies of the GDB for a localised group of Basic Service Sets

BSS) RLSS are connected via a secure protocol architecture with the

quivalent of multiple Access Points (AP) in different BSS realising

trusted DB infrastructure APs locally coordinate the exchange of

nformation and channel access management between the GDB (via

LSS) and the secondary usersrsquo end stations (STA)

Mode I devices are those under the control of a device that em-

loys geo-location database access while Mode II devices are those

mploying geo-location database access by themselves The informa-

ion exchange provided between the GDB and a secondary user STA

an be provided in both an open-loop (eg adopted by FCC) and a

losed-loop (eg adopted by ETSI) implementation In an open loop

mplementation daily spectrum availability information is provided

y the GDB and no feedback on the spectrum information received

s provided by STAs thus the system approach for spectrum access

s much more conservative leading to low channelsrsquo utilisation po-

ential In a closed loop implementation the STA can provide feed-

ack to the GDB and there are more communication overheads due

o the high granularity of updates however the system is more effec-

ive and reliable in the exploitation of TVWS on behalf of STAs The

ypical information provided by the GDB includes (1) the updated

hite Space Maps (WSM) of frequencies allowed for secondary use

t the timespace of the querying STA and (2) the device-dependent

ower limitations for transmission (in general conservative and ac-

urate enough to avoid relevant interference effects on primary users

dentified in the area)

The basic IEEE 80211af mechanisms regulating the communica-

ion between STA and GDB can be found in [137138]

In parallel with IEEE 80211af standards such as the IEEE 19006

ave been created that consider interfaces and data structures sup-

orting spectrum sensing information exchange applicable to spec-

rum sensing (and particularly distributed spectrum sensing) sce-

arios In particular in October 2014 the IEEE 19006 WG initiated

project for a new standard called IEEE 19006b [146] concerning

he use of spectrum sensing information to support and optimise the

ffectiveness reliability and robustness of spectrum database solu-

ions The aim is to enhance the performance and capabilities of spec-

rum databases through the use of spectrum sensing information

Co-existence Support for co-existence mechanisms so that mul-

iple technologies can effectively utilise the TVWS spectrum is im-

ortant Self co-existence between network devices of a common

echnology (eg deployed by different operators in the same area)

nd co-existence among different technologies are relevant topic of

esearch for Cognitive Radio (CR) systems and specifically for IEEE

0211af Many solutions appeared on the research scene but no one

as so far finalised as the target solution for IEEE 80211af Specific

tandardisation has been started to regulate coexistence between

ireless standards of unlicensed devices including the IEEE 802191

147] The purpose of the IEEE 802191 standard is to enable the fam-

ly of IEEE 802 Wireless Standards to most effectively use TV White

pace by providing standard coexistence methods among dissimilar

r independently operated TVWS devices Early examples of gener-

lised coexistence mechanisms included Dynamic Frequency Selec-

ion (DFS) Transmission Power Control (TPC) listen before talk (eg

or contention based IEEE 80211 80215) time division multiplex-

ng (also among different techniques such as the IEEE 80216 80220

0222) and Message-based Spectrum Contention (that is beacon-

ng messages that carry coexistence information) Opportune met-

ics must be defined to assess the measurable coexistence achieved

mong different technologies as an example the hidden node prob-

bility for a target scenario or the estimate of percentage variation

n normalised network throughput and latency (before and during

he SU transmissions) On the other hand a centralised coexistence

ontrol mechanism could be effectively realised by a central manager

or coexistence-DB like the GDB) in critical scenarios To this end

EEE 19004 [148] (a standard for heterogeneous networks in dynamic

pectrum context part of IEEE Standards Coordinating Committee 41)

ims to standardise the overall system architecture and information

xchange between the network and mobile devices which will allow

hese elements to optimally choose from available radio resources

There are three possible classification of co-existence archi-

ectures and inter-network coordination channels for CR systems

entralised coordinated and autonomous co-existence mechanisms

149] In centralised co-existence schemes the co-existence is admin-

stered by a central entity (eg like in IEEE 802191) This solution

ould be applied in both homogeneous IEEE 80211af systems (eg

ncarnated by the centralised spectrum DB entity) and in hetero-

eneous inter-networks without requiring modifications to existing

tandards under the assumption that all the involved co-existing

etwork entities would adhere to the same centralised coordination

cheme and inter-network coordination protocol On the other hand

his assumption could be hard to satisfy in some practical scenar-

os The centralised schemes are effective in providing minimisation

f inter-network interference based on the availability and quality

f centralised co-existence information In coordinated co-existence

chemes the centralised entity could be present but not taking de-

isions in general The central entity could provide a co-existence

B with required information which can be used via in-band and

ut-band signalling for co-existence decisions taken by the cluster

16 B Bellalta et al Computer Communications 75 (2016) 1ndash25

c

e

n

c

o

s

i

d

a

S

8

n

a

b

o

b

i

t

a

8

b

[

t

i

n

b

5

t

a

c

b

s

a

c

n

s

p

i

c

l

r

c

f

a

v

t

a

e

c

t

Q

a

b

s

i

c

u

n

c

i

heads of many co-existing inter-networks for implementing a proper

common coordination protocol (eg out-band busy tone signalling)

These are often hybrid solutions realising a compromise between ef-

fectiveness and suitability In autonomous co-existence schemes all the

decisions and coordination are implemented in a distributed way

by the involved network entities In these schemes many policies

can be adopted to realise a sufficient level of co-existence under a

best effort approach Solutions that could be implemented include

the use of busy tones beaconing and other signalling protocols in

dedicated control channels dynamic distributed frequency selection

schemes listen before transmit token-based and dynamic reserva-

tion schemes All the above mentioned solutions have goods and

bads and a general illustration can be found in [149] Other proac-

tive co-existence techniques try to early detect and recovermitigate

co-existence issues with relaxed inter-network coordination eg re-

alised via spectrum sensing and interference avoidancesuppression

techniques

It must be clear that different aims exist regarding the co-

existence for primary TVWS users protection against secondary

users and for secondary devices mutual interference avoidance

Many co-existence problems have been analysed in the literature

for TVWS technologies in particular between the primary (licensed)

and secondary (unlicensed) devices The coexistence management in

TVWS is complex because primary users of DVB transmissions are in-

tended to be the pure receivers rather than the broadcasters In other

words potential protected users could be everywhere and the hid-

den terminal problem could arise for these systems The co-existence

problem is also complicated by many factors that cause asymme-

try and dynamicity in the TVWS such as the mobility variable den-

sity power asymmetry and heterogeneous MACPHY layers TVWS

co-existence problems originate between the IEEE 80211af technolo-

gies and the IEEE 80222 technologies for wireless regional area net-

works (WRAN) Results of analysis of TVWS usage in Europe show

that white spaces are typically present and fragmented They are typ-

ically more abundant in rural areas where larger contiguous blocks

of unused channels are available due to broadcast network planning

giving priorities linked to population density However the exploita-

tion of TVWS in urban areas is possible eg some recent research

was realised under the assumption of exploiting shadowing effects

created in the communication environment (eg by buildings obsta-

cles) in favour of frequency reuse [150]

Since different standards for opportunistic communication in the

TV White Spaces have now been published such as IEEE 80222 [151]

IEEE 802154m [152] Weightless [153] and of course IEEE 80211af

improved methods to guarantee the coexistence of different devices

operating on several protocols in the same bands must be deployed

We highlight here some work eg [150154ndash157] that focuses on

the coexistence between different technologies operating in the same

bands In particular [155] studies the performance degradation of

IEEE 80222 when an IEEE 80211af network is operated in the same

area The problem becomes even worse when the IEEE 80211af net-

work is located near 80222 user equipment causing both networks

to perceive a strong interference from each other One proposed solu-

tion is the Coexistence Beacon Protocol [158] studied for 80222 net-

works which foresees the exchange of a periodic beacon to identify

the neighbouring and possibly interfering networks

For what concerns the co-existence problems research has ad-

dressed other important related aspects of new WLANs on TV White

Spaces For instance the higher coverage range makes them attrac-

tive for applications in the smart grid and also for M2M on TVWS

[134150159] Here the increased range compared to standard tech-

nologies in the ISM bands can have beneficial effects for indoor

mobile devices [150] and M2M [159] Specifically indoor devices

communicating on TV White Spaces offer better propagation char-

acteristics and penetration through obstacles making it easier to

realise scalable home connectivity [135] although interference and

o-existence problems could be exacerbated and must be resolved

g see [139] However studies in the literature show how the sig-

al coming from an indoor transmitter on TV White Spaces remains

onstrained inside the house making it difficult to deploy indoor-to-

utdoor networks On the other hand this shadowing limit can be

een as an opportunity lowering the interference outside the build-

ng in which the TVWS network is locally operated [134150] Un-

er this approach HDTV streaming has recently been considered as

possible use case of WLAN networks operating in the TV White

pace [160]

Non-contiguous channel bonding A relevant novel feature of IEEE

0211af is the potential for contiguous and non-contiguous chan-

el bonding which permits aggregating basic channels (also non-

djacent ones) and leveraging the possible large frequency spread

etween multiple available channels Due to the rather static nature

f primary DTV transmissions where a busy channel is unlikely to

ecome free in the near future and state changes are coarse-grained

n general it is crucial to exploit the time-locality effect and exploit

he maximum physical channel availability that could be aggregated

t any given location With the methodology inherited from IEEE

0211ac IEEE 80211af is capable of bonding together two up to four

asic channels grouped in up to two different non-contiguous chunks

137] The spectrum bandwidth of a DVB-T basic channel can be ei-

her 6 7 or 8 MHz depending on the country in which the service

s operated As an example this creates a 144 to 168 OFDM chan-

elsrsquo bandwidth potential when up to four 6ndash7ndash8 MHz channels are

onded [137]

12 Open challenges

The deployment of WLANs in the TVWS at their maximum poten-

ial still requires the resolution of open problems Generally speaking

s mentioned in previous sections the challenges for such networks

an be divided into three main categories (1) spectrum sharing

etween opportunistic (secondary) and heterogeneous devices (2)

pectrum sensingmanagement and maximum exploitation of avail-

ble spectrum (potentially enabling the spectrum-on-demand con-

ept) and (3) co-existence and interference mitigation to the primary

etwork (primary user protection) and secondary networks [154]

The overall aim of spectrum sharing techniques is to maximise

eparation to avoid overlapping or contiguous channels operations

rovided that sufficient channels are available in the TVWS spectrum

n a given spacetime scenario [154] In general important research

ontributions have to be realised for the design of proper spectrum al-

ocation methodologies and techniques satisfying multi-factorial QoS

equirements at the system-level and the user-level Novel ideas in-

lude the spectrum sharing and spectrum sensing techniques being

urther divided into cooperative and non-cooperative

Non-cooperative solutions attempt to realise spectrum sharing

nd sensing on a local basis without direct cooperation between de-

ices Examples include the ldquolisten-before-transmitrdquo approach and

ransmission power control to limit the interference and spectrum

llocation policies that are based on local node feedback only Coop-

rative solutions rely on the existence of a common communication

hannel through which devices can tentatively agree on the spec-

rum allocation that provides the desired spectrum separation and

oS Major issues include the definition of a methodology to identify

common channel (or multiple channels mutually shared in space

y pairs of heterogeneous and distributed devices) and to efficiently

hare common information For the access to the spectrum DBs the

dentification of proper common access channels is still under dis-

ussion out-of-TVWS-band cellular communications can be widely

sed given their high coverage also indoor

Trying to avoid the use of complex and resource-hungry coordi-

ation schemes and dedicated control channels rendezvous proto-

ols play a significant role both for co-existence and for establish-

ng dynamic communication channels based on spectrum sharing

B Bellalta et al Computer Communications 75 (2016) 1ndash25 17

R

n

c

t

l

o

m

e

n

i

n

w

n

a

u

s

p

b

s

m

s

a

d

t

b

a

a

m

a

t

u

8

n

t

d

d

m

f

t

n

d

n

e

r

a

h

b

f

a

c

t

u

i

f

t

d

t

s

i

f

I

l

i

a

n

c

d

I

n

o

t

i

t

e

w

t

c

m

t

a

i

t

q

b

c

a

i

l

i

a

v

t

a

a

c

t

i

t

b

t

t

a

D

p

c

s

o

q

r

p

a

g

t

e

u

d

t

m

n

t

s

f

i

1

8

f

T

endezvous is considered a fundamental problem in generalised CR

etworks at the basis of many communication processes eg in-

luding neighbour discovery routing and broadcast Rendezvous pro-

ocols attempt to establish a new link for communication on a se-

ected and agreed frequency band (channel) identified within a set

f available resources (channels) thus creating the basis for com-

unication between two or more SUs This problem is often exac-

rbated by the high (and unknown) numbers of contending SUs high

umber of available channels and their ambiguous boundaries and

dentifiers and tough co-existence requirements satisfaction in cog-

itive radio systems [161] Preliminary approaches for rendezvous

ere based on the existence of a dedicated Common Control Chan-

el (CCC) agreed among all the SUs or the existence of a centralised

gent (decision maker) assigning channels to any pairs of requesting

sers This approach could be considered in the case of IEEE 80211af

olutions based on the spectrum DB implementation however the

roblem is complicated by the fact that the common channel must

e identified and made available for all the requesting SUs at the

ame time even when these are spread over multiple collision do-

ains in space On the other hand the scalability and reliability is-

ues which are caused by the congestion of CCCs and the vulner-

bility risks of centralised DBs motivate the research of alternative

istributed solutions called blind rendezvous protocols [162] Dis-

ributed solutions have been mainly based on beaconing andor slot-

ased channel hopping algorithms exploiting randomisation sensing

nd discoverysynchronisation messages spread in slotted time over

set of candidate available channels [161] These protocols aim to

inimise the convergence time to maximise the use of resources

nd to avoid collisions among SUs by identifying and agreeing on

he use of a given common channel among N non-overlapping and

nambiguously labelled channels The main complication in IEEE

0211af systems is given by the wide availability of many different

on-contiguous spectrum bands which cannot be uniformly quan-

ised in single channels and unambiguously labelled in space in or-

er to realise a common reference domain for all the IEEE 80211af

evices In fact the knowledge of the number of channels their com-

on labels the unique IDs of SUs and the number of SUs contending

or the channel assignment are the key factors determining the effec-

iveness and good properties of blind rendezvous protocols Recently

ew classes of rendezvous protocols have been proposed which are

istributed blind and oblivious Oblivious means that available chan-

els may be identified with different labels on behalf of the differ-

nt involved SUs A number of these distributed blind and oblivious

endezvous protocols have been proposed and have been referenced

nd analysed in [162] To conclude solutions for dynamic spectrum

andoff issues (eg in case PUs suddenly appear in a region) must

e resolved in advance to minimise the latency of re-establishing a

unctional communication channel eg by implementing a proactive

lternative rendezvous channel selection in background to ongoing

ommunication processes [163]

Another challenge is to provide cooperative techniques in a dis-

ributed vs centralised implementation by analysing their convergence

nder variable conditions overheads and tradeoffs In some cases as

n vehicular networks the exploitation of scenario characteristics and

actors such as the constrained mobility could help to realise effec-

ive dissemination of sensing information over an extended and pre-

ictable horizon under a cooperative approach [164ndash166] Coopera-

ion could be helpful in the realisation of the cooperative spectrum

ensing in combination with the Geolocation DB approach Another

nteresting direction of research is the investigation of mutual ef-

ects of coexisting cooperation-based vs non-cooperative techniques

n cooperation-based devices the exchange of information could al-

ow a quick convergence to a solution and avoidance of further

nterference [145] Certainly cooperative techniques have desirable

dvantages over non-cooperative ones However tight time synchro-

isation and high coordination overheads are required The common

hannel for cooperation purposes could become a bottleneck and re-

uce the potential advantages of dynamic spectrum allocation The

EEE 80219 foresees the presence of a shared common control chan-

el (CCC) to which devices need to tune in order to gather the state

f the network Solutions such as [141] propose the use of TVWS for

he CCC

A promising direction for research is the adoption of cluster-

ng schemes for secondary devices associated with a hybrid dis-

ributedcooperative vs centralisednon-cooperative approach in an

ffort to reduce overheads while maximising advantages In this

ay secondary nodes could implement cooperation for sensing func-

ions using properly identified common channels and delegate the

entralised decisions and spectrum-DB management to well instru-

ented cluster-leader nodes This concept is incarnated to some ex-

ent by RLSS in IEEE 80211af More recently there have been propos-

ls to build distributed databases that refresh their contents period-

cally by querying the remote spectrum database [167] Here Mas-

er devices (according to Ofcom terminology) periodically cache the

uery replies from the spectrum database in order to reply to and

roadcast the spectrum availability to neighbour Slave devices

Another challenging direction for research is the definition of ac-

urate models and efficient simulation tools enabling dynamic spectrum

nalysis for both rural and urban areas Recent developments of dig-

tal maps simulation tools and propagation models theoretically al-

ow improving the capability of predicting radio propagation effects

n complex scenarios with an acceptable computation time This en-

bling technology could be used in parallel to spectrum DBs to pro-

ide more accurate forecasting of dynamic frequency allocation in

imespace scenarios Another issue to be taken into account is the

ccuracy of the remote spectrum database regarding the channel

vailability estimation Several works have already shown the inac-

uracy of propagation models due to the complexity of parameters

o be considered and differences between modelled and real scenar-

os [132134168169] Better propagation models will make it possible

o more efficiently estimate the interference between devices and

uild Radio Environment Maps (REM) or White Space Maps (WSM)

hat will possibly lead to a more accurate sharing of the radio spec-

rum [144] To this end current research could focus on how to lever-

ge simulation models in order to build more accurate frequency

Bs

In general the adoption of remote spectrum DBs methodology

ushed by national regulators has made the spectrum sensing pro-

ess optional and conservatively realised (eg power control for

ensing-only devices is limited to below 17 dBm EIRP) Thus much

f the research has focused on techniques and solutions to successfully

uery the remote database also in challenging environments such as

ural areas and indoor However since the queries must provide the

osition of the mobile device the accuracy of positioning for mobile

nd handheld devices still needs more appropriate solutions Trian-

ulation can be used to estimate the position of mobile devices with

he accuracy needed by the current regulations outdoor (50 m) How-

ver a low GPS accuracy is possible indoors [170] Certified man-

al positioning can be provided by technicians for static and indoor

evices although limiting mobility Alternatives to costly infrastruc-

ures for positioning (eg based on short range beaconing devices)

ust be deployed in indoor scenarios

Finally regarding spectrum sensing and interference mitigation

ew challenges come from the heterogeneity and different charac-

eristics that networks operating in the TV White Spaces can have

uch as different transmitting power (4 W for static devices 100 mW

or mobile and portable devices and 40 mW for communication

n channels adjacent to occupied ones) different bandwidths (5

0 15 and 20 MHz are currently foreseen for IEEE 80211af IEEE

02154m can have narrower bandwidths or wider ones) and dif-

erent medium access schemes (protocols that use either CSMA or

DMA are required to co-exist in the TV White Spaces) With all this

18 B Bellalta et al Computer Communications 75 (2016) 1ndash25

d

a

T

i

a

c

t

l

b

o

t

a

d

p

a

r

p

n

a

m

t

p

p

d

d

o

t

i

D

w

p

i

a

a

d

b

fi

d

o

t

l

b

t

b

a

i

e

o

o

c

t

t

w

A

c

M

s

c

l

O

w

g

v

t

in place the problem of determining whether another secondary de-

vice is currently transmitting in the same channel is still an open

issue

Spectrum sensing is a very hot topic in the literature with many

proposals that leverage the cooperation between devices to raise the

accuracy of the detection process eg [144171] The current litera-

ture also offers room for proposals that suggest exploiting sensing to

build more accurate spectrum DBs [164172173] Lastly regarding the

interference mitigation to the primary network regulators provide a

conservative approach based on a remote spectrum database to avoid

high risk for the primary users However in particular for the indoor

scenario the concept of secondary usersrsquo transmission in TV Grey

Spaces has been proposed TV Grey Space identifies busy channels

that are formally used by primary users in the area of interest (eg

at the rooftop level of a building for DVB-T) but which could be con-

sidered usable indoors (eg at the basement level) without causing

effective interference to the primary users [134150174] This concept

is based on the generalised spectrum underlay paradigm [175] The

remaining research challenge is to provide TV Grey space access guar-

anteeing a sufficient amount of protection to the primary user while

providing a satisfactory QoS to opportunistic devices eg through

the use of customised and validated propagation models for indoor

TVWS networks and feedback mechanisms to constantly monitor the

interference generated to the primary network

6 Emerging new trends and technologies

In this last section we review three emerging trends that in our

opinion will have a large influence in the conception of future WLANs

as they change the way WLAN protocols and functionalities are de-

veloped implemented tested and integrated with other wireless

networks

61 Programmable Wireless LANs

Especially in the enterprise environment WLAN deployments

need to support a wide range of functionalities and services This is

intrinsically difficult because of the large number of APs that must be

managed which calls for scalable solutions Typical services include

channel assignment load balancing among APs authentication au-

thorisation and accounting (AAA) policy management support for

client mobility and interference coordination Another problem is the

fact that WLAN clients autonomously take several decisions such as

which APs to associate with when to hand-over etc Therefore sup-

porting roaming clients requires the management of a large number

of association states across several APs which is a challenge if sup-

port for real-time hand-over is desired Typically such management

schemes are centralised and most of them are proprietary such as

WLAN controller solutions from Aruba [176] or Cisco [177] although

the IEEE 80211u amendment has been released to allow mobile users

to seamless roam between WiFi networks with automatic authenti-

cation and handoff [178] For example Dyson [179] enables STAs to

send information such as radio channel conditions to a centralised

controller based on a custom API (eg based on Python) As the con-

troller has a centralised view of the network it can enforce a rich

set of policies to control the network also using historical informa-

tion A demo system has been implemented along with applications

such as airtime reservations for specific clients or optimised hand-

offs However Dyson requires STAs to be modified in order to use

those new services offered by the centralised controller TRANTOR

[180] is another example of a centralised management system that

requires changes of clients in order for the infrastructure to gather

information from them in terms of eg interference measurements

In addition control commands enable the infrastructure to exercise

control such as modifying the transmit power or influencing the as-

sociation procedure Clients still use standard CSMA MAC layer for

ata transmission In contrast to DenseAP [181] or DIRAC [182] which

re also based on a centralised controller but do not modify clients

RANTOR exercises much more control on clients by the use of a ded-

cated API which enables a significantly higher gain in coordination

nd thus capacity CENTAUR [183] is another example of a centralised

ontroller that centrally schedules hidden and exposed terminals in

he downlink while using standard CSMA MAC for uplink traffic and

egacy downlink traffic To achieve good performance it uses a fixed

ack-off packet staggering techniques and a hybrid data path that

nly schedules downlink transmissions towards hidden and exposed

erminals centrally All other traffic is sent using standard DCF with

standard back-off procedure In contrast to previous approaches it

oes not require any modifications to the clients but does require data

lane centralisation and it remains questionable how scalable such

solution is for large WLAN deployments with high PHY layer data

ates beyond 100 Mbps However new approaches for fast data plane

rocessing that are currently explored in the virtualisation commu-

ity which move all the packet processing into user space may be

n option to significantly increase speed Designing such manage-

ent systems raises several interesting research questions (a) Cen-

ralised infrastructures may incur high latency but are more simple to

rogram and maintain while designing a completely distributed ap-

roach that operates close to WLAN devices may result in significant

istributed coordination and consistency problems (b) what level of

istributed control is required in order to support the flexibility needs

f future WiFi based networks supporting a high level of mobility

Traditionally WLAN APs are built on proprietary operating sys-

ems that are tightly coupled with the hardware This design makes

t hard to create new applications on top of such networking devices

espite the fact that protocols and mechanisms are available that

ould greatly improve the utility of existing networks those new

rotocols are not deployed because the closed system design makes

t very difficult to extend their functionality An important aspect for

ll such new approaches is therefore how to provide open interfaces

nd open source to speed up innovation As an example Linux-based

evices are completely open source but in order to increase flexi-

ility much more work is needed in the area of open drivers and

rmware In fact the Atheros based ath9k driver has been the main

river of innovation in the open source community because it is the

nly driver that interworks with open firmware In general solutions

hat modify the OS driver focus mainly on programmable network

evel solutions for WLANs such as channel switching or handovers

etween APs While the ath9k driver is an excellent example in how

o speed up innovation it is still very cumbersome to support flexi-

le MAC engine reprogrammability with ath9k In contrast MAClets

nd Wireless MAC processors (WMP) [184] allow a much more flex-

ble reprogramming of MAC functionalities A MAC processor is an

ntity able to execute general MAC commands that specify the MAC

perations through a software-defined state machine The behaviour

f the MAC protocol can therefore be updated at run-time by simply

hanging the sequence in which those commands are executed (ie

he MAClets) As a proof-of-concept the authors of [184] implement

he proposed MAC processor solution in a commodity WLAN hard-

are card extending the basic DCF in three directions piggy-backing

CKs a pseudo TDMA and the use of multiple channels In [185] a

ontrol framework for this WMP system is also proposed to support

AClet code mobility ie for moving loading and activating MAC

oftware programs embedded into ordinary data packets (akin to the

apsule model of traditional active networks)

The difficulty in re-programming networking hardware has also

ed to the concept of Software Defined Networks (SDNs) based on the

penFlow protocol [186] The main idea of SDNs is to extend net-

orking devices with standardised APIs that allow third-party pro-

rammers to flexibly control the data path In addition SDNs pro-

ide higher level abstractions to network designers and programmers

hrough the use of a centralised control plane offered by a network

B Bellalta et al Computer Communications 75 (2016) 1ndash25 19

Fig 11 The CLOUDMAC architecture VAP stands for Virtual Access Point and WT

stands for Wireless Termination

c

a

c

t

m

W

n

t

t

c

p

t

u

h

a

i

e

s

o

c

a

a

f

T

r

s

a

s

p

o

M

C

i

T

t

fl

A

a

a

i

c

t

s

t

e

I

t

w

H

e

b

I

l

d

o

6

r

W

s

l

a

e

e

a

s

b

c

t

(

S

p

c

F

B

t

p

T

e

t

s

u

m

i

e

o

a

A

p

d

t

t

n

p

i

t

t

f

i

ontroller such as NOX [187] which allows reuse of components such

s topology discovery or network access control for different appli-

ations The SDN concept has recently been applied to WLAN archi-

ecture in order to enable fine grained control over mobility manage-

ent and data forwarding focusing on a programmable enterprise

LAN architecture An important part of the SDN architecture is the

etwork controller which provides a centralised view of (parts of)

he SDN enabled network and uses the OpenFlow protocol to install

he forwarding rules on SDN-enabled devices (routers switches ac-

ess points cellular base stations etc)

ODIN [188] is designed to support programmability in enter-

rise WLAN architecture by separating the association state from

he physical AP They implement Light Virtual Access Points (LVAP)

sing a Split-MAC approach where the infrastructure controls the

andover procedure among different WLAN APs By managing

ssociations through SDN controllers ODIN enables proactive mobil-

ty management and load balancing within the SDN enabled WLAN

nterprise network without the need for changes in the client WLAN

tack or IEEE 80211 MAC layer While ODIN requires agents to reside

n the APs that communicate with the Odin Master within the SDN

ontroller CLOUDMAC [189] is based purely on the concept of SDN

nd virtual APs Similar to ODIN CLOUDMAC implements a Split-MAC

pproach but in addition enables the processing of WLAN MAC layer

rames within a co-located Cloud using so-called Virtual APs (VAPs)

he physical APs in CLOUDMAC are lightweight WLAN APs that are

esponsible only for sending out IEEE 80211 based MAC layer ACKs to

tandard WLAN clients and tunnel WLAN MAC layer frames through

n SDN-enabled enterprise WLAN towards the VAPs As association

tates are kept in the VAPs fast mobility is supported using sim-

le OpenFlow forwarding rules Because of the additional processing

f IEEE 80211 WLAN MAC frames in the co-located Cloud CLOUD-

AC has higher latency than a standard WLAN deployment However

LOUDMAC offers a Webservice based API to third party applications

n order to program the enterprise WLAN and enable new services

he architecture of CLOUDMAC is depicted in Fig 11 It has been ex-

ended in [190] to support QoS management and in [191] to support

exible MAC management frame prioritisation based on IEEE 80211

system based on OpenFlow has been recently proposed in [192] to

llow the station to be associated with multiple APs simultaneously

nd to switch between APs with low overhead

An important aspect to consider in order to enable programmabil-

ty is backwards compatibility Several approaches (eg [180]) require

lients to be modified to utilise the features provided This is difficult

o do in practice because it requires changes in the operating system

oftware of all clients If not all clients can utilise those APIs it is ques-

ionable how usable the new architecture will be and how much ben-

fit in terms of aggregate performance such architecture will allow

n contrast the SDN based approaches are interesting in the sense

hat they do not require changes in the client WLAN stack and work

ith the standard IEEE 80211 MAC layer deployed within the clients

owever it remains questionable how scalable such solutions are For

xample an interesting open research topic is to evaluate the scala-

ility of approaches such as CLOUDMAC [189] In addition processing

EEE 80211 MAC frames within co-located private Cloud requires low

atency support from local Cloud solutions (such as OpenStack) in or-

er to reduce the IEEE 80211 MAC processing time which is an area

f active research

2 Prototyping and testing IEEE 80211 enhancements

Most of the new proposals for next-generation WLANs are cur-

ently only evaluated using mathematical analysis and simulation

hile both analysis and simulation are necessary to characterise and

tudy those enhancements in the initial design phase or to consider

arge-scale scenarios it is difficult to consider all practical aspects of

real-world scenario This can sometimes cause significant differ-

nces between what simulations and real experiments show How-

ver real experiments are challenging because of the high complexity

nd costs of building the new hardware and software for each specific

olution to test

To mitigate implementation complexity there are several flexi-

le hardware platforms where low-level MAC mechanisms can be

ompletely implemented in software Examples based on FPGAs are

he USRP (Universal Software Radio Peripheral) [193] and the WARP

Wireless Open-Access Research Platform) [194] As an alternative

ORA (Microsoft Research Software Radio) [195] works on general

urpose computers by taking some advantage from modern multiple

ore systems

A different approach is provided by OpenFWWF [196] Open-

WWF provides an open CSMACA firmware for specific models of

roadcom chipsets so the resulting firmware can be uploaded and

ested in real commercial hardware OpenFWWF implements a sim-

le State Machine (SM) for controlling the hardware in real time

he SM evolution is driven by a main loop that reacts to events by

xecuting specific handlers When a packet originally prepared by

he Linux kernel is ready in the NIC memory handler Packet_Ready

ets up the radio hardware according to the packet metadata sched-

les the transmission and jumps back to the main loop The Trans-

ission Engine (TXE) then takes care of accessing the channel ie

t decrements the back-off counter according to the DCF rules and

ventually starts the actual transmission This triggers the execution

f the TX_frame_now event that prepares the ACK time-out clock

nd finalises the MAC header If the ACK-frame is received or if the

CK time-out expires and the maximum number of attempts for this

acket is reached handler Update_Params resets the contention win-

ow to its minimum value or otherwise doubles it Finally it loads

he back-off counter with a fresh value

In the following we will review a list of selected papers that use

he OpenFWWF firmware and the WARP or USRP platforms to test

ew proposals for WLANs in real scenarios as examples A first im-

lementation of the groupcast mechanisms defined in IEEE 80211aa

s presented in [197] The authors modify the OpenFWWF firmware

o include those functionalities and evaluate them using a 30 STAs

estbed A second example of the use of OpenFWWF is the collision-

ree MAC protocol presented in [198] In [198] the authors design and

mplement in OpenFWWF a MAC protocol that is able to achieve a

20 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 12 LTE to WIFI offloading

a

n

c

t

[

d

a

W

v

t

t

o

b

f

S

g

l

i

I

s

I

b

n

s

m

a

p

d

c

N

o

v

t

t

n

l

i

a

t

a

collision-free operation by waiting a deterministic time after success-

ful transmissions before to start a new transmission This work shows

the experimental performance of a collision-free MAC (CF-MAC) pro-

tocol for WLANs using commercial hardware The results show that

the proposed CF-MAC protocol leads to a better distribution of the

available bandwidth among users a higher throughput and lower

losses than DCF

In [199] the authors use the WARP platform to test a variant of the

DCF ndash called IEEE 80211ec ndash that substitutes control packets such as

the RTS CTS and ACK with short detectable sequences Since those se-

quences are shorter than the control packets and can be detected cor-

rectly even at lower SNRs values a significant gain in performance is

achieved MIDAS (Multiple-Input Distributed Antenna Systems) [200]

implementation using the WARP platform shows the benefits of dis-

tributing the antennas over the area to cover instead of co-locating

them at the AP in terms of capacity when MU-MIMO is employed

The authors also propose a new MAC protocol to benefit from the

spatial reuse that the DAS allows taking as a basis the IEEE 80211ac

Experimental results show that their proposal is able to achieve up

to 200 gains versus the traditional approach where all antennas are

co-located at the AP

The USRP platform has recently been proposed to develop a first

implementation of IEEE 80211ah in order to obtain a preliminary ex-

perimental performance assessment of such a technology because

no commercial off-the-shelf hardware is yet available as the amend-

ment is still in progress [201] The USRP platform is also used in

[202] to evaluate TIMO (Technology Independent Multi-Output) a so-

lution to deal with high-power non-IEEE 80211 interferers in ISM

bands

To conclude this section it is worth mentioning here also to

MAClets and WMP as given such flexibility to implement and dis-

tribute new MAC protocols and other IEEE 80211 functionalities we

believe that if a use-friendly implementation of such a MAC proces-

sors framework is provided it would significantly contribute to the

development and testing of new MAC enhancements by the research

community

63 CellularWLAN interworking

Public hotspots that offer Internet access over a WLAN using IEEE

80211 technology are now nearly ubiquitous It is forecasted that the

cumulative installed base of WiFi hotspots worldwide will amount

to 551 million by 2018 excluding private hotspots (eg WiFi access

points deployed at home) [203] The sharp increase in the availabil-

ity of public WiFi was initially perceived by mobile cellular operators

as a threat due to the additional competition from wireline Internet

service providers or emerging crowdsourced WiFi networks such as

FON5 However as cellular operators are fighting to cope with the ex-

plosion of mobile data traffic created by the rising use of multimedia

content traffic over mobile devices [15] they are also starting to use

WLANs based on the IEEE 80211 technology to offload data from their

core and access networks In general mobile data offloading refers to

the use of complementary network technologies (in licensed or un-

licensed spectrum) for delivering data originally targeted to cellular

networks Intuitively the simplest type of offloading consists of ex-

ploiting connectivity to existing co-located WiFi networks and trans-

ferring data without any delay (Fig 12) Thus this offloading tech-

nique is known as on-the-spot offloading As a consequence of this

new trend the seamless integration of cellular (eg 3GLTE) and WiFi

technologies has attracted significant research interest in recent years

(see [204] for a survey) a few solutions have already been standard-

ised [205206] and roaming between cellular and WiFi is becoming

increasingly transparent to end users CellularWLAN interworking is

5 httpscorpfoncomen

fl

t

f

lso fostered by the support in the evolving 4G standards of heteroge-

eous network deployments (HetNets) in which the existing macro

ells are complemented with a number of small low-power base sta-

ions with the goal of increasing capacity in highly congested areas

207208] It is envisaged that small cells will be based on 4G stan-

ards (eg pico and femto cells) as well as IEEE 80211 technologies

nd multimode base stations that work simultaneously with LTE and

iFi are already entering the market

It is important to point out that LTE standards already support a

ariety of mechanisms that enable data offloading However most of

he existing solutions such as Local IP Access (LIPA) and Selected In-

ernet IP Traffic Offload (SIPTO) focus on data offloading at the core

f the network [205] For instance LIPA allows an IP-enabled mo-

ile device to transfer data to another device in the same pico or

emto cell without passing through the cellular access network while

IPTO enables the routing of selected IP data flows through different

ateways The only LTE offloading mechanism that supports seam-

ess interworking with IEEE 80211-based WLANs is IP Flow Mobil-

ty and Seamless Offload (IFOM) Specifically IFOM relies on Mobile

Pv6 technologies to allow a user terminal to simultaneously route

elected IP flows over different radio access technologies [209210]

n this way a user terminal can offload selected flows to a WLAN

ased on some operator-defined policy while keeping the LTE con-

ection running However to enable such an approach it is neces-

ary to support an entity in the cellular core network that can com-

unicate with the user terminal to exchange information about the

vailability and quality of neighbouring access networks as well as to

rovide the user terminal with predefined rules to manage the han-

over process This entity in current LTE standards is called the Ac-

ess Network Discovery and Selection Function (ANDSF) server [211]

ote that the problem of selecting the best communication technol-

gy in an heterogeneous wireless network has been extensively in-

estigated in the literature both with centralised as well as decen-

ralised approaches (see [212] for a survey) However the use of mul-

iple interfaces in parallel as well as per-flow offloading are relatively

ew concepts Thus the design of scalable and efficient network se-

ection strategies for the ANDSF framework is still an open issue For

nstance three offloading methods suitable for the ANDSF framework

re proposed in [213] which are based on coverage SNR and sys-

em load A reinforcement learning approach is designed in [214] that

llows multimode base stations to autonomously steer their traffic

ows across different access technologies depending on the traffic

ype the usersrsquo QoS requirements the network load and the inter-

erence levels A number of studies have also tried to quantify the

B Bellalta et al Computer Communications 75 (2016) 1ndash25 5

Fig 2 High-throughput demanding multimedia devices associated to an IEEE 80211acax AP

s

(

3

s

w

b

i

t

s

t

c

o

f

C

t

A

p

c

f

u

c

i

t

u

i

t

t

I

s

m

[

a

t

u

s

T

e

S

Fig 3 A DL-MU-MIMO transmission in IEEE 80211ac Note that the AP transmits two

spatial streams in SU-MIMO mode to STA B and a single spatial stream to STA A Packet

aggregation is used in both transmissions The PHY header is transmitted ommnidirec-

tionally to inform the selected STAs about the next MU-MIMO transmission

n

C

c

F

D

t

t

s

h

d

d

b

m

k

c

M

A

D

d

3

n

(

e

i

i

I

cheme ie a 256-QAM modulation and downlink multiuser MIMO

DL-MU-MIMO)

11 Novel features

The most relevant new features included in IEEE 80211ac are de-

cribed in the following

Channel bonding IEEE 80211ac enables the use of channel band-

idths of 20 40 80 (mandatory) and 160 MHz (optional) Channel

andwidths larger than 20 MHz are created by ldquobondingrdquo (ie group-

ng) a group of consecutive 20 MHz channels and aim to offer higher

ransmission rates

Two extensions have been proposed in IEEE 80211ac for the ba-

ic DCF (Distributed Coordination Function) access method in order

o support channel bonding (i) the Static Bandwidth Channel Ac-

ess Protocol (SBCA) which always transmits over the same group

f 20 MHz channels and requires that all sub-channels are idle be-

ore starting a packet transmission and (ii) the Dynamic Bandwidth

hannel Access scheme (DBCA) which is able to dynamically adapt

he channel width to the instantaneous spectrum availability [3031]

s expected in dense scenarios the use of DBCA offer a much better

erformance than SBCA due to adaptability [32]

To avoid hidden terminals operating in any of the 20 MHz bonded

hannels the IEEE 80211ac amendment includes extended RTSCTS

rames in order to signal the maximum channel width that can be

sed at both the transmitter and the receiver In case the CTS in-

ludes a lower channel width than the RTS the transmitter will adopt

t Similarly to the ACK frames when the RTS and CTS frames are

ransmitted they are duplicated over all the 20 MHz sub-channels

sed The operation and performance of channel bonding in WLANs

s thoroughly analysed in [27] showing the new interactions be-

ween neighbouring WLANs that may appear and their impact in the

hroughput of each one

Downlink multiuser MIMO The main novelty introduced by the

EEE 80211ac amendment compared with the IEEE 80211n one is the

upport of MU-MIMO transmissions in the downlink hence allowing

ultiple simultaneous transmissions from the AP to different STAs

58] In the IEEE 80211ac amendment the AP can be equipped with

maximum of eight antennas and send up to four spatial streams

o two different users or up to two spatial streams to four different

sers at the same time

When an IEEE 80211ac AP performs a multi-user transmission it

pecifies the group of STAs to which that transmission is directed

his information is contained in the new IEEE 80211ac PHY head-

rs which are broadcast omni-directionally to all STAs The way

TAs are grouped is decided by the AP after obtaining the chan-

el state information (CSI) feedback from all STAs To gather the

SI information by the AP IEEE 80211ac considers only an explicit

hannel sounding feedback mechanism called Explicit Compressed

eedBack (ECFB) The channel access is governed by EDCA (Enhanced

istributed Channel Access) At each transmission attempt the mul-

iple access categories (AC) managed by the AP should contend for

he channel medium as only one AC can be served for each transmis-

ion attempt In the case that the queue associated with the AC that

as won the internal contention does not contain packets to enough

ifferent destinations to fill all the available spatial streams it can

ecide to share the remaining ones with the other ACs

Packet aggregation To increase the efficiency of each transmission

y reducing unnecessary overheads IEEE 80211ac allows the trans-

ission of several MPDUs aggregated in a single A-MPDU Then to ac-

nowledge each MPDU individually a Block ACK packet is used which

ontains a bitmap to indicate the correct reception of all included

PDUs Thus leveraging on the information contained in the Block

CK the transmitter is able to selectively retransmit only those MP-

Us that have failed instead of the whole A-MPDU Fig 3 illustrates a

owlink MU-MIMO transmission in which packet aggregation is used

12 Open challenges

Since the IEEE 80211ac amendment has recently been fi-

alised current research around it should cover two main aspects

a) understanding the performance bounds of IEEE 80211ac which

ntails the development of new models simulation tools and exper-

mental platforms of IEEE 80211ac-based WLANs and (b) propos-

ng specific solutions for those aspects that are not defined by the

EEE 80211ac amendment on purpose such as the mechanism for

6 B Bellalta et al Computer Communications 75 (2016) 1ndash25

g

i

l

a

b

b

f

o

w

o

H

t

e

n

a

f

t

m

a

w

s

3

a

8

2

c

s

m

W

r

a

[

o

m

2

n

c

p

r

a

creating the groups of STAs for DL-MU-MIMO transmissions smart

packet schedullers able to decide when the use of DL-MU-MIMO out-

performs SU-MIMO transmissions and the implementation of the

TXOP sharing feature between several ACs The results and conclu-

sions obtained in both cases will be very valuable in the development

of IEEE 80211ac technologies as well as in the conception of the fu-

ture amendments that will substitute IEEE 80211ac in four to five

years such as the recently initiated IEEE 80211ax

Following the first mentioned research direction there are several

efforts that have focused on understanding both theoretical and exper-

imental performance bounds of IEEE 80211ac The maximum downlink

throughput that an IEEE 80211ac AP can achieve when packet aggre-

gation channel bonding and different spatial stream configurations

are considered is presented in [33] In [34] the authors evaluate the

IEEE 80211ac performance experimentally using commodity devices

focusing on the effects that the use of wider channels the 256-QAM

modulation and the number of SU-MIMO spatial streams have in

terms of throughput and energy consumption It is worth mentioning

that DL-MU-MIMO was not yet implemented in the equipment they

were using and that feature was therefore not considered The evalu-

ation of a DL-MU-MIMO implementation for WLANs using the WARP

platform is presented in [35] where a deep evaluation of the potential

benefits of DL-MU-MIMO transmissions is done in terms of the loca-

tion of the receivers number of users and user mobility among other

aspects A solution that combines both packet aggregation and DL-

MU-MIMO transmissions is presented in [36] Results show the need

of properly dimensioning the buffer space to achieve the full potential

of such a combination In [37] the authors compare the throughput

achieved by IEEE 80211n and IEEE 80211ac when packet aggrega-

tion is used with and without channel errors They show that in most

cases the packet aggregation mechanism introduced in IEEE 80211ac

outperforms the one in IEEE 80211n An analytical model to evalu-

ate the performance of the IEEE 80211ac TXOP sharing mechanism in

DL-MU-MIMO communications is developed in [38] The main goal of

this study is to identify how the TXOP sharing mechanism could im-

prove the system efficiency while achieving channel access fairness

among the different ACs

How to optimally exploit the new DL-MU-MIMO capabilities pro-

vided by IEEE 80211ac is still an open challenge First due to the need

of frequent CSI exchanges between STAs and the AP it is not yet clear

in which conditions DL-MU-MIMO outperforms SU-MIMO [39ndash42]

or even whether MU-MIMO does or does not outperform multi-user

packet aggregation when the amount of data directed to each des-

tination is not balanced [43] Packet aggregation can be a solution

to balance the duration of the multi-user spatial streams as shown

in [36] although it will always depend on the amount of traffic di-

rected to each destination and the buffer capacity at the AP In [44]

the authors compare different strategies to assign the spatial streams

between the available destinations at each transmission in a fully

connected mesh network showing in ideal channel conditions the

theoretical benefits of MU-MIMO vs SU-MIMO

Closely related to the previous point a second open challenge is

the design of efficient schedulers that consider traffic priorities the

buffer state the different MIMO strategies TXOP sharing policies

grouping of STAs and the availability of fresh CSI feedbacks to max-

imise the throughput and guarantee the required QoS for each ac-

tive traffic flow It is important to consider that the availability of up-

dated CSI estimates from all STAs allows the AP to reduce the mutual

interference between the transmitted spatial streams which means

lower packet error probabilities and higher transmission rates How-

ever the overheads for obtaining the CSI from all STAs is large and

increases linearly with the channel sounding rate and the number

of STAs Proposals for reducing the CSI overhead are under develop-

ment For example in [40] the CSI overhead is reduced by inhibiting

the channel sounding whenever possible based on the estimation of

the channel stability for all users Another related problem is how to

roup the STAs as the goal is to find groups of STAs with compat-

ble (ie orthogonal) channels In [45] the authors show the chal-

enges inherent to the group assignment problem and they propose

n heuristic method to solve them TXOP sharing is considered in [46]

y presenting two alternative approaches to enhance the considered

ack-off procedure for the purpose of improving both throughput and

airness

A third key challenge for IEEE 80211ac is to achieve an efficient use

f the spectrum when several channel widths are used in scenarios

ith multiple overlapping WLANs Increasing the channel width the-

retically allows individual WLANs to achieve a higher throughput

owever the presence of other WLANs in the vicinity also increases

he chances of frequency overlapping which may cause the opposite

ffect as there appears inter-WLAN contention [27] Adaptive mecha-

isms to select the channel centre frequency and the channel width

nd MAC protocols to choose the instantaneous channel width used

or each transmission are thus required For instance in [47] the au-

hors focus on the channel selection problem when WLANs can use

ultiple channel widths using a game-theoretic framework In [48]

scheme is proposed to enable the communication between nodes

ith partially overlapping channels which may provide stronger re-

ilience to channel interferences

2 The IEEE 80211ax amendment

In 2014 the High Efficiency WLANs (HEW) Task Group [49] initi-

ted the development of a new IEEE 80211 amendment called IEEE

0211ax The IEEE 80211ax amendment is expected to be released in

019 and to some extent it will be the IEEE 80211 response to the

hallenges of future dense and high-bandwidth demanding WLAN

cenarios [5051]

The challenges in the development of the IEEE 80211ax amend-

ent are to

(i) Improve the WLANs performance by providing at least a four-

fold capacity increase compared to IEEE 80211ac

(ii) Provide support for dense networks considering both the ex-

istence of multiple overlapping WLANs and many STAs in each

of them Spatial reuse of the transmission resources is a must

(iii) Achieve an efficient use of the transmission resources by min-

imising the exchange of management and control packets re-

visiting the structure of the packets and improving channel

access and retransmission mechanisms among others aspects

(iv) Provide backward compatibility with previous amendments

This is achieved by the mandatory transmission of the legacy

PHY preamble in all frames and by keeping EDCA as the basic

channel access scheme

(v) Introduce effective energy saving mechanisms to minimise the

energy consumption

(vi) Support multi-user transmission strategies by further devel-

oping MU-MIMO and Orthogonal Frequency Division Multiple

Access (OFDMA) capabilities in both downlink and uplink

In addition to the aforementioned challenges next-generation

LANs will have to implement some other functionalities beyond the

aw packet transmission and reception Examples are a fast efficient

nd robust handoff between APs in the same administration domain

52] device-to-device communication (D2D) [53] and coordination

f multi-AP networks [54] In the first case the IEEE 80211ai amend-

ent called Fast Initial Link Setup is in progress and expected for

016 Its target is to complete a handoff in less than 100 ms including

ew AP discovery user authentication and configuration Using D2D

ommunication we can avoid the use of the AP as a relay hence im-

roving the overall efficiency as the number of packet transmissions

equired is reduced Finally the virtualisation of network functions

dds a new dimension in the management of multiple APs which in

B Bellalta et al Computer Communications 75 (2016) 1ndash25 7

d

e

I

w

f

t

s

3

P

s

G

t

8

p

C

c

m

c

i

t

a

s

w

f

t

t

e

l

p

e

a

H

d

t

c

o

s

p

n

t

b

a

w

c

t

w

n

8

a

c

w

d

H

n

a

h

p

w

i

t

fi

e

c

u

t

m

p

p

b

p

i

t

8

d

I

m

M

u

t

m

i

t

D

W

I

u

M

i

w

p

h

t

n

o

p

t

s

u

s

r

b

A

t

O

t

t

w

s

p

o

i

W

t

t

a

[

m

ense scenarios can contribute to notably improving the user experi-

nce We further discuss this last topic in Section 6

Different from the other amendments covered in this survey the

EEE 80211ax amendment is just in its initial stages of development

ith only very few technical aspects consolidated at this stage There-

ore in the following subsection we will overview both the new fea-

ures and open challenges of the IEEE 80211ax amendment at the

ame time

21 Novel features and open challenges

The IEEE 80111ax Task Group is currently working in four areas

HY MAC Multi-user and Spatial Reuse [49] Next we will overview

ome of the topics currently under discussion in the IEEE 80211 Task

roup in each category

PHY layer The IEEE 80211ax PHY layer will be an evolution of

he IEEE 80211ac one The challenges in the design of the IEEE

0211ax PHY layer are related with the extensions required to sup-

ort multi-user MU-MIMO and OFDMA transmissions and Dynamic

CA Also improvements in the supported modulation and channel

oding techniques will be likely considered to allow for higher trans-

ission rates at lower SNR values For example IEEE 80211ax may

onsider LDPC (Low-Density Parity Check) coding which is optional

n IEEE 80211ac instead of the traditional convolutional codes as

hey provide a coding gain of 1ndash2 dB Moreover the PHY layer may

lso include some functionalities to support the use of Hybrid ARQ

chemes to improve the efficiency of packet retransmissions

Medium acess control In order to keep backward compatibility

ith previous IEEE 80211 amendments besides a common PHY

rame preamble compatible MAC protocols are required This means

hat it is likely that EDCA will be kept as the main channel access

echnique in the IEEE 80211ax amendment Therefore the most rel-

vant open challenges are related to EDCA extensions to support a

arge number of STAs improve traffic differentiation capabilities im-

rove the energy consumption and provide mechanisms to fairly co-

xist with neighboring wireless networks

To support a large number of contenders with a low collision prob-

bility a simple solution is to use large backoff contention windows

owever it would also increase the time a STA is in backoff so re-

ucing the number of packets it can transmit A solution to reduce

he backoff duration when using large contention windows is to de-

rease the duration of the backoff slots Latest technological advances

n electronics should require less time to perform a CCA check to

witch between reception and transmission modes and to process a

acket thus enabling such a possibility

Another approach to enhance the underlying CSMACA mecha-

ism in EDCA and achieve a higher efficiency is to consider decen-

ralised collision-free MAC strategies Those MAC protocols are able to

uild collision-free schedules thus improving the network efficiency

s collisions are reduced while preserving backward compatibility

ith the default EDCA implementation An overview of several de-

entralised collision-free MAC protocols can be found in [55] Among

hem CSMAECA [56] is especially relevant since it is fully compatible

ith EDCA and latest upgrades made it adaptive to the instantaneous

umber of contenders in a single WLAN

In addition to EDCA IEEE 80211ax WLANs can rely on the IEEE

0211aa amendment to further improve the WLAN traffic differenti-

tion capabilities with intra-AC traffic differentiation and groupcast

ommunication mechanisms among other features In Section 33

e will provide further details about the IEEE 80211aa amendment

IEEE 80211ax will likely keep the same channel widths that were

efined in the IEEE 80211ac amendment ie 20 40 80 and 160 MHz

owever it is expected that IEEE 80211ax will extend current chan-

el bonding mechanisms to further improve the spectrum utilisation

nd the coexistence between neighbouring WLANs For example it

as been shown in [32] that the use of dynamic channel bonding

rovides significant throughput gains in dense scenarios compared

ith the static approach while minimising the inter-WLAN negative

nteractions [27] Furthermore additional mechanisms are required

o fully exploit the use of channel bonding such as the use of ef-

cient algorithms to select the position of the primary channel or

ven to consider the existence of multiple primary channels to in-

rease the number of bonded channel combinations that a node can

se for transmission

The MAC layer in IEEE 80211ax may work with the PHY layer

o implement an efficient Hybrid ARQ mechanism able to retrans-

it only incremental redundancy bits in short packets Opportunistic

iggy backing of data packets in ACKs and vice versa may further im-

rove the efficiency of IEEE 80211ax WLANs by reducing the num-

er of transmissions in a bidirectional data exchange [57] Finally

acket headers can be reduced if shorter STA identificators are used

nstead of MAC addresses and unnecessary fields are avoided in all

ransmissions

Finally in terms of power saving mechanisms it is expected IEEE

0211ax will follow the same trends as in previous IEEE 80211 stan-

ards including the extensions currently under development for the

EEE 80211ah amendment that are discussed in Section 4

Multi-user Multi-user communications will likely be one of the

ain characteristics of IEEE 80211ax as both uplink and downlink

U-MUMO and OFDMA are under consideration The use of multi-

ser communication techniques does not necessarily represent a sys-

em capacity increase because the available transmission resources

ay be the same as in the single-user communication case However

n WLANs the simultaneous transmission from different users is able

o parallelise the large temporal overheads of each transmission (ie

IFS SIFS ACKs packet headers etc) which can notably improve the

LAN efficiency

IEEE 80211ax will further develop the MU-MIMO capabilities of

EEE 80211ac by allowing multiple simultaneous transmissions in the

plink which is known as uplink (UL-) MU-MIMO [58] Similar to DL-

U-MIMO transmissions an open challenge to enable UL-MU-MIMO

s to design a mechanism able to efficiently schedule the users that

ill transmit at the same time In one hand a pure decentralised ap-

roach would be easy to implement with minimal signalling over-

eads However since it requires that all STAs finish their backoff at

he same time it may be very inefficient besides that those STAs may

ot be compatible in terms of their respective spatial channels In the

ther hand a pure centralised approach requires that the AP has com-

lete CSI and buffer occupancy information from all STAs to select

he most suitable group to perform a multi-user transmission Once a

uitable group of STAs is selected by the AP a ldquoTriggerrdquo frame may be

sed to notify the group of selected users that can initiate a transmis-

ion This approach guarantees efficient multi-user transmissions but

equires some extra overheads to collect all the required information

y the AP and signal the selected STAs In both cases new multi-user

CKs will be likely introduced by IEEE 80211ax to acknowledge all

ransmissions with a single control packet

Multi-user OFDMA is also in the agenda for IEEE 80211ax Using

FDMA a channel can be split in several sub-channels and assigned

o different users Likely OFDMA will be implemented in combina-

ion with channel bonding where each of the 20 MHz subchannels

ill be assigned to a different user in both downlink and uplink Be-

ides that a similar operation as in the multi-user MIMO case is ex-

ected as there are almost the same challenges to solve A survey

f current OFDMA proposals for WLANs is presented in [59] show-

ng also how the use of OFDMA is able to significantly improve the

LAN efficiency Fig 4 illustrates an uplink OFDMA and MU-MIMO

ransmission

In addition to Multi-user MIMO and OFDMA the use of Simul-

aneous Transmit and Receive (STR) techniques commonly known

s full-duplex transmission have been suggested for IEEE 80211ax

5051] Using STR a pair of nodes is able to transmit and receive si-

ultaneously [60] which theoretically doubles the channel capacity

8 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 4 Multiuser uplink transmission using two different OFDM subchannels Trans-

missions from STA B and C over the same subchannel represent an uplink MU-MIMO

transmission All transmissions are acknowledged with a single Multiuser ACK The

trigger frame is used to signal the selected STAs

t

t

t

a

s

t

f

f

c

3

s

s

t

n

n

A

d

w

t

fi

a

t

8

i

d

t

r

w

d

R

v

f

s

d

p

B

c

e

k

r

G

a

c

m

t

b

d

3 The concealment address protects legacy stations ie GCR-incapable stations

from receiving duplicated group-addressed frames

The challenge is that both the AP and the STA involved in a full-duplex

transmission have to start to transmit at the same time To achieve

that information about full duplex transmission capabilities can be

included in RTSCTS control packets to set-up a full-duplex transmis-

sion between the transmission initiator and its destination

Spatial reuse Dense WLAN deployments are necessary to offer a

continuous coverage with high transmission rates To improve both

the co-existence with those neighboring networks and the spatial

reuse of the spectrum a WLAN has two options (i) minimise its

area of influence by reducing its transmit power and (ii) accept

higher interference levels by increasing the Clear Channel Assess-

ment (CCA) level Use of both techniques may increase the number of

concurrent transmissions between neighbouring WLANs and there-

fore their capacity although it may also result in the opposite effect

since the achievable transmission rates may be negatively affected by

the higher interference levels observed which is the main challenge

to solve

Due the high WLAN dynamics the use of adaptive systems is cru-

cial but challenging as adaptivity requires extra complexity in terms

of computing and memory resources and there are no guarantees

that the implemented solution converges due to the decentralised

operation of each WLAN The use of DSC (Dynamic Sensitivity Con-

trol) to dynamically adjust the CCA level is one of the aspects cur-

rently under discussion in the IEEE 80211ax Task Group First studies

evaluating the performance of DSC for IEEE 80211ax WLANs show

a clear improvement on the spatial reuse and the area throughput

[61] Another example of the achievable throughput gains obtained

by adapting the CCA level can be found in [62] where the authors

show that gains of up to 100 can be achieved Moreover trans-

mit Power Control (TPC) to mitigate interference between WLANs in

dense scenarios is studied in [63] showing the need of jointly opti-

mising both TPC and CCA to maximise the network performance

Finally sectorisation by using beamforming is also under consid-

eration for the development of the IEEE 80211ax amendment as a

potential solution to improve spatial reuse [64] Using sectorisation

only the nodes of a given area are allowed to receive or transmit data

hence reducing the contention between different networks whenever

they activate non-overlapping sectors A challenge here is to coor-

dinate the different neighboring APs when they belong to different

administration domains Decentralised learning approaches may be

implemented to find feasible temporal patterns of non-overlapping

sectors

33 The IEEE 80211aa amendment

As discussed above legacy IEEE 80211 standards do not provide

robust and efficient delivery of audiovideo streaming services Thus

he IEEE 80211aa amendment was developed to include new fea-

ures and additional mechanisms to improve the performance of real-

ime multi-media content delivery [65] Specifically IEEE 80211aa

ddresses the following five shortcomings of previous 80211

tandards [1666]

(i) the lack of reliable and efficient support for multicast and

group communications

(ii) the incapacity of applying traffic prioritisation to different

multimedia streams or different types of frames from the same

stream

(iii) the absence of methods for cooperative resource sharing

among neighbouring APs

(iv) the lack of mechanisms for graceful degradation of audiovideo

streaming quality

(v) the non-interoperability with existing IEEE 8021 standards for

Audio Video Bridging (AVB)

In the following sections we present in detail the solutions to

hose problems introduced in the IEEE 80211aa amendment We

urther discuss the research studies that have provided the basis

or the IEEE 80211aa design and we identify the remaining open

hallenges

31 Novel features

Groupcast communication mechanisms In most audiovideo

treaming applications a group of clients must receive the same

tream simultaneously A multicast protocol is necessary to avoid that

he same content is replicated throughout the network In wireless

etworks multicast transmission can exploit the intrinsic broadcast

ature of the wireless channel ie broadcast transmissions from an

P are physically received by all other stations in the same collision

omain However multicast and broadcast frames in IEEE 80211 net-

orks are not protected by an acknowledgement mechanism as in

he case of unicast frames Thus layer-2 multicast transmissions de-

ned by legacy IEEE 80211 standards are unreliable and not suit-

ble for streaming applications To partially address this limitation

he Direct Multicast Service (DMS) was first specified in the IEEE

0211v amendment [67] Basically DMS converts multicast streams

nto unicast streams In this way frames destined to a multicast ad-

ress are individually transmitted as unicast frames to the stations

hat joined that multicast group Obviously DMS provides the same

eliability as unicast transmission services but the consumed band-

idth increases linearly with the number of group members To ad-

ress this scalability issue IEEE 80211aa includes the Groupcast with

etries (GCR) service in addition to DMS Specifically the GCR ser-

ice defines new mechanisms and the related management frames

or group formation which allows a set of stations to agree on a

hared (non-multicast) address called the groupcast concealment ad-

ress3 Furthermore the GCR service specifies two retransmission

olicies GCR Unsolicited Retry (GCR-UR) and GCR Block Ack (GCR-

A) When using GCR-UR the AP can proactively retransmit all group-

ast frames a number of times to mitigate the impact of channel

rrors (see Fig 5a)) while receivers are not requested to send ac-

nowledgements Intuitively this approach improves transmission

eliability but it still suffers from scalability issues In contrast when

CR-BA is used the AP sends a burst of consecutive groupcast frames

nd it requests the receivers to reply with a Block ACK frame which

ontains a bitmap to positively or negatively acknowledge trans-

itted frames (see Fig 5b)) The Block ACK mechanism defined for

he GCR-BA service is quite flexible because Block ACK frames can

e requested immediately after a transmission burst or after a ran-

omised back-off delay Furthermore the AP can request the Block

B Bellalta et al Computer Communications 75 (2016) 1ndash25 9

Fig 5 GCR service with different retransmission schemes

A

r

o

l

o

e

v

e

v

d

f

d

s

v

c

s

a

a

f

d

t

F

t

d

v

a

f

c

t

t

o

t

i

c

r

v

i

r

s

o

t

a

(

p

i

f

a

o

i

t

t

c

s

fi

o

c

I

n

i

a

t

a

t

f

i

B

T

f

t

a

g

a

a

i

s

a

d

s

p

m

m

i

a

t

b

t

i

(

r

a

S

i

c

s

c

T

Q

p

a

3

g

t

4 TSPEC is a traffic specification sent from a QoS capable wireless client that requests

a certain amount of network traffic from the AP for the traffic stream it represents

CK frame to all groupcast recipients or only to a subset of them to

educe overheads and delays The advantages of the GCR methods

ver broadcast and DMS have been extensively demonstrated in the

iterature [1668]

Intra-access category prioritisation The IEEE 80211e amendment

nly allows traffic differentiation between four different access cat-

gories (ACs) that are broadly mapped to four application classes

oice (VO) video (VD) best-effort (BE) and background (BK) How-

ver there is a variety of streaming services ranging from simple

ideoconferencing to HD streaming over IPTV systems which have

ifferent QoS requirements (see Table 3) To provide the ability to dif-

erentiate among individual streams IEEE 80211aa includes an ad-

itional scheduling layer with respect to IEEE 80211e IEEE 80211aa

plits each one of the transmission queues associated with voice and

ideo ACs into a primary and an alternate queue In this way spe-

ialised scheduling rules can be applied to decide which queue to

erve when the EDCA function for inter-AC collision resolution grants

n access opportunity to voice or video ACs To facilitate the man-

gement of service level agreements IEEE 80211aa follows the de-

ault mappings between user priority values and traffic types that are

efined in the IEEE 8021D standard [69] It is then straightforward

o further map traffic types onto transmission queues and ACs (see

ig 6) Finally it is important to point out that the intra-AC differen-

iation functionality can be used to provide more sophisticated traffic

ifferentiation than simple stream prioritisation For instance most

ideo applications use Scalable Video Coding (SVC) schemes that en-

ble the partitioning of a video sequence into multiple layers with dif-

erent qualities and rates [70] Typically an SVC-based video stream

ontains a base layer which provides a basic level of quality and mul-

iple enhancement layers which can only be decoded together with

he base layer to improve the video quality Thus the different layers

f the same encoded video steam can be easily mapped to different

ransmission queues to receive differentiated QoS [71]

Stream classification service The stream classification service (SCS)

s an optional service that can be provided by an AP to the asso-

iated stations to classify multimedia streams based on arbitrary

ules that are established directly by the stations instead of the con-

entional 8021D user priorities To this end the station request-

ng the use of SCS must initiate an SCS session by sending an SCS

equest frame to the AP which contains an identifier for the SCS

tream and the descriptor of the classification rule The AP may accept

r reject the requirements specified by the station Once accepted

he AP must assign all frames that match the classification rule to

specific AC When intra-access category prioritisation is enabled

see Section 331) calternate transmit queues so that finer grained

rioritisation can be applied Finally there is also a Drop Eligibil-

ty Indicator (DEI) bit in the SCS descriptor that indicates whether

rames from this traffic stream can be dropped in the case that there

re insufficient resources Specifically frames with the DEI bit set to

ne have a higher probability of being discarded because their max-

mum number of allowed retries is smaller than the default Note

hat how to combine intra-AC queues and frame dropping settings

o achieve graceful degradation of the audiovideo stream quality in

ase of bandwidth shortage is beyond the scope of the IEEE 80211aa

pecification

Overlapping Basic Service Set (OBSS) management Network densi-

cation ie a denser deployment of wireless infrastructure nodes is

ne of the key strategies that is used nowadays to easily increase the

apacity of wireless systems even for indoor WLANs [72] However

EEE 80211 networks have a limited number of orthogonal chan-

els available and even if optimised frequency planning is applied

t might happen that neighbouring APs are mutually interfering and

station may affect multiple overlapping BSSs In this case conges-

ion not only increases but it is also likely to observe an unfair us-

ge of wireless capacity with the channel retained by one AP for long

ime intervals This is mainly due to the neighbourhood capture ef-

ect ie hidden terminal phenomena among APs [73] To address this

ssue IEEE 80211aa specifies a new functionality called Overlapping

SS (OBSS) management which is based on two new mechanisms

he first defines a set of parameters to quantify the load and inter-

erence among neighbouring BSSs such as medium occupancy frac-

ion number of admitted audiovideo streams data traffic volumes

nd the number of BSSs that are using the same channel as the tar-

et one Note that the traffic load consists of two components the

llocated traffic which is derived on the basis of the TSPEC values of

dmitted streams4 and predicted traffic which is evaluated by track-

ng the maximum value of the allocated EDCA and HCCA traffic over

even-day periods Once load measurement reports are exchanged

mong the APs a second OBSS component is responsible for coor-

inated admission control procedures on the basis of two suggested

haring schemes proportional sharing and on-demand sharing The

urpose of both schemes is to keep the total allocated traffic below a

aximum value in order to provide some QoS protection to admitted

ultimedia streams Finally IEEE 80211aa recommends implement-

ng additional OBSS management procedures for channel selection

nd cooperatively creating HCCA schedules that do not collide

Interworking with IEEE 8021AVB Audio Video Bridging (AVB) is a

erm commonly used to denote a set of technical standards developed

y IEEE to support real-time streaming services with bounded latency

hrough IEEE 802 networks [74] This objective is achieved by spec-

fying mechanisms to allow the synchronisation of multiple streams

IEEE 8021AS [75]) and traffic shaping (IEEE 8021Qav [76]) and to

eserve network resources for specific audiovideo streams traversing

bridged local area network by using a signalling protocol called the

tream Reservation Protocol (SRP) (IEEE 8021Qat [77]) IEEE 80211aa

ntegrates the SRP operations with the EDCA admission control pro-

edures Specifically the SRP RequestResponse messages are encap-

ulated in the management frames that are used to carry the traffic

haracteristics and the QoS requirements during admission control

his enables the end-to-end management of resource reservation for

oS guaranteed streams even when one or more IEEE 80211 links are

art of a path from the stream producers (called IEEE 8021Q talkers)

nd the stream consumers (called IEEE 8021Q listeners)

32 Open challenges

In recent years several MAC enhancements have been investi-

ated to improve QoS guarantees for real-time multimedia applica-

ions in IEEE 80211 networks [20] and the IEEE 80211aa standard

10 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 6 Stream classification and inter-AC traffic prioritisation

a

l

t

r

n

a

t

n

i

t

i

t

r

d

p

t

t

e

m

b

t

a

i

s

o

s

a

e

t

i

f

e

m

8

t

I

i

fi

which was finalised in 2012 included several of these proposed im-

provements Significant research efforts have focused on improving

the transmission reliability of multicasting by integrating ARQ mech-

anisms in IEEE 80211-based multicast transmissions Modifications

to the legacy MAC protocol were proposed in [78] to enable the

RTSCTS option in multicast mode and to select one or more multi-

cast receivers (called leaders) for acknowledging multicast data pack-

ets However these enhancements require changes to the standard

specifications The main problems of leader-based ARQ schemes are

leader election and the trade-off between scalability and reliability

The authors in [79] propose selecting the multicast recipient operat-

ing in the worst channel conditions as the unique leader but this ap-

proach may perform poorly in lossy environments In the Batch mode

multicast MAC (BMMM) [80] all multicast recipients are polled by

the multicast originator to send individual ACKs but this scheme is

not suitable for large multicast groups The Enhanced Leader Based

Protocol (ELBP) is proposed in [81] on the basis of multiple ACK-

leaders and block acknowledgement techniques Analytical models

are then developed to help select optimal ACK-leaders to meet ap-

plication QoS requirements However the models apply only to sat-

urated traffic while multimedia streams are typically bursty Another

class of reliable multicast protocols relies on busy tones to reduce

packet losses due to collisions [82] but the additional radio inter-

face needed for the busy tone limits the practicality of such solu-

tions An alternative approach to avoid collisions of multicast pack-

ets is the multicast collision prevention (MCP) scheme [83] which

is based on the use of a shorter waiting time for transmitting mul-

ticast packets An interesting approach is also proposed in [84] to

retransmit lost packets using an online linear XOR coding algorithm

However a modification to the standard MAC protocol is required to

enable simultaneous ACK transmissions In summary several differ-

ent methods have been proposed to improve multicast transmission

reliability by integrating ARQ schemes into the protocol architecture

but there are not conclusive results on which is the best solution The

choice of the most efficient mechanism depends on a variety of in-

terdependent factors such as loss ratios channel congestion multi-

cast group size and QoS requirements of multimedia streams A com-

prehensive analytical framework is needed to optimise the setting of

the parameters for each scheme and to dynamically select the best

one

As discussed above one main difference between unicast services

nd multicast services in the legacy IEEE 80211 standard was the

ack of acknowledgements Another critical difference is that mul-

icast frames must be transmitted using a fixed rate in the basic

ate set while the transmission rate of unicast frames can be dy-

amically adapted to the channel and traffic conditions [85] Thus

group of research papers has investigated the use of rate adap-

ation to improve the throughput of multicast services in IEEE 80211

etworks [7186ndash89] For instance the authors in [86] propose us-

ng RTS frames to allow group members to estimate channel condi-

ions Each member will then send a dummy CTS frame with a length

nversely proportional to channel quality In this way the multicast

ransmitter can use the collision duration to predict the lowest data

ate that can be used for group transmissions The overhead intro-

uced by this mechanism is quite high however The solution pro-

osed in [88] called ARSM also relies on feedback messages sent by

he multicast receivers called multicast response frames to identify

he group member exhibiting the poorest channel conditions How-

ver in this case a different back off timer is associated with each

ulticast receiver depending on the SNR of previously received feed-

ack messages in order to prevent collision An approach similar to

he one employed in the Auto Rate Fallback (ARF) protocol a rate

daptation scheme originally proposed in [90] is used in [87] Specif-

cally the number of successful consecutive transmissions and con-

ecutive transmission failures are used to decide when to increase

r decrease the transmission data rate respectively A modified ARF

cheme is also proposed in [71] which can be applied to videos that

re encoded into two layers namely the base and enhancement lay-

rs However how to integrate rate adaptation with the different re-

ransmission policies that are defined in IEEE 80211aa is still an open

ssue

One research area that is expected to be crucial in the success-

ul development of IEEE 80211aa-based products is the design of

fficient scheduling algorithms for supporting voicevideo traffic Al-

ost all research work in this field has been triggered by the IEEE

0211e amendment that enhanced the original IEEE 80211 MAC with

wo new QoS-aware access mechanisms ie EDCA and HCCA [91]

n principle with a well-designed admission control and schedul-

ng scheme HCCA is able to provide hard QoS guarantees to traf-

c flows [9293] However HCCA is rarely implemented in IEEE

B Bellalta et al Computer Communications 75 (2016) 1ndash25 11

8

c

c

a

s

b

w

o

j

t

f

a

Q

t

t

A

p

t

s

a

f

s

H

a

i

c

i

t

s

f

s

s

m

i

t

s

l

t

t

[

t

i

o

b

r

W

v

m

8

s

i

o

p

q

4

c

e

t

d

a

t

m

Fig 7 WLANs for M2M communications STAs represent sensor and actuator devices

4

t

w

s

a

c

g

e

(

p

o

t

p

9

r

o

a

p

c

o

i

p

b

f

t

s

o

i

4

8

f

t

i

s

m

d

0211e-based WLANs owing to its higher complexity and cost con-

erns Instead EDCA is widely adopted Most papers have thus fo-

used on improving EDCA performance Many papers have proposed

nalytical models for various subsets of EDCA functionalities For in-

tance a saturation-based performance analysis is conducted in [94]

y differentiating the minimum back-off window size the back-off

indow-increasing factor and the retransmission limit The authors

f [9596] also model AIFS differentiation while the model in [97]

ointly captures all the four EDCA parameters for traffic differen-

iation More recent papers have analysed the EDCA performance

or non-saturated conditions and for arbitrary buffer sizes [98] The

uthors in [99] have developed an analytical model to predict the

oS levels that can be achieved once a new voicevideo flow is in-

roduced in the WLAN A Kalman filter is proposed in [100] to ob-

ain estimates on the number of active transmission queues of each

ccess Category in EDCA These analytical models can then be ex-

loited to derive the optimal configuration of the EDCA parameters

o achieve given performance criteria or to design admission control

chemes that preserve QoS constraints For instance a scheme that

ssigns contention-window values to achieve pre-defined weighted-

airness goals is proposed in [101] A control-theoretic scheme is de-

igned in [102] with the goal of minimising the video traffic delay

owever most of these solutions rely on non-realistic assumptions

bout video traffic dynamics An alternative class of solutions dynam-

cally updates the EDCA parameters based on the observed network

onditions In [103] the EDCA parameters are optimised consider-

ng a WLAN with rigid and elastic traffic simultaneously analysing

he interactions between both types of traffic The authors in [104]

pecify several bandwidth-sharing mechanisms with guaranteed QoS

or voice and video traffic Measurement-based admission control

chemes are proposed in [105] A TXOP adaptation method is de-

cribed in [106] that takes into account video frame sizes and trans-

it queue lengths However the main drawback of these solutions

s that they are based on heuristics and hence do not ensure op-

imal and guaranteed performance Finally a third category of re-

earch papers tries to improve video performance by designing cross-

ayer scheduling approaches Specifically these works take advan-

age of multi-layer video encoding to classify the frames according

o their importance and assign them to different access categories

107] For instance the authors in [108] define classifiers and waiting

ime priority schedulers that dynamically change the packet prior-

ties according to end-to-end delay measurements A disadvantage

f this approach however is that an additional adaptation layer may

e needed to implement the complex interactions that are typically

equired between the video coding applications and the MAC layer

e conclude this section by pointing out that existing studies pro-

ide the basic design principles and techniques for improving multi-

edia streaming performance in IEEE 80211 networks Still the IEEE

0211aa standard poses new research challenges that have not been

ufficiently explored and that will require innovative solutions For

nstance scheduling between primary and alternate queues is still an

pen research area as the mapping of individual frames to multi-

le queues in order to achieve graceful degradation of voicevideo

uality [16]

Sensor Networks and machine-type communications

As discussed in Section 2 M2M communications refer to any

ommunication technology that enables sensoractuator devices to

xchange information and perform actions without the manual assis-

ance of humans This section reviews the main features currently un-

er consideration in the development of the upcoming IEEE 80211ah

mendment which targets the main challenges of those networks as

he IoT in general such as the energy consumption or the manage-

ent of many devices

1 The IEEE 80211ah amendment

The IEEE 80211ah amendment [109] aims to provide WLANs with

he ability to both manage a large number of heterogeneous STAs

ithin a single BSS and minimise the energy consumption of the

ensor-type battery-powered STAs

The initial design requirements of the IEEE 80211ah amendment

re detailed in [110] these entail the support of up to 8192 STAs asso-

iated with a single AP the adoption of efficient power saving strate-

ies a minimum data rate of 100 kbps the operation in the license-

xempt sub 1 GHz band and a coverage up to 1 km in outdoor areas

see Fig 7 for an illustrative example) A preliminary assessment of

erformance of the IEEE 80211ah technology in terms of the number

f STAs that can be effectively supported in a single WLAN as well as

heir energy consumption is presented in [111]

IEEE 80211ah operates over different sub-1 GHz ISM bands de-

ending on country regulations 863ndash868 MHz in Europe 902ndash

28 MHz in the US and 9165ndash9275 MHz in Japan China South Ko-

ea and Singapore also have specific channelisations Channel widths

f 1 MHz and 2 MHz have been adopted although 4 8 and 16 MHz

re also supported in some countries IEEE 80211ah furthermore pro-

oses new PHY and MAC layers The IEEE 80211ah PHY layer can be

onsidered to some extent a sub-1 GHz version of the IEEE 80211ac

ne At the physical layer OFDM is the chosen modulation method us-

ng 32 or 64 tonessub-carriers that are spaced by 3125 kHz The sup-

orted modulations include BPSK QPSK and from 16 to 256-QAM A

road range of antenna technologies ranging from single-user beam-

orming to MIMO and DL-MU-MIMO which was first introduced in

he IEEE 80211ac amendment are also included in the IEEE 80211ah

pecification Similarly the IEEE 80211ah MAC protocol include most

f IEEE 80211 main characteristics further extending its power sav-

ng (PS) mechanisms

11 Novel features

This section overviews the extensions introduced by the IEEE

0211ah amendment to the IEEE 80211 PS mechanisms to account

or the specific characteristics of resource-constrained sensor and ac-

uator devices A more detailed review can be found in [110] includ-

ng a performance assessment of IEEE 80211ah in several of the key

cenarios for M2M communications such as agriculture and animal

onitoring smart metering and industrial automation plants In ad-

ition a detailed survey of the IEEE 80211ah is reported in [112]

12 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 8 IEEE 80211ah PS mode for TIM-STAs non-TIM STAs and Unscheduled STAs

a

d

w

i

a

a

c

m

i

a

e

t

s

m

t

t

u

b

i

I

s

d

f

a

b

s

a

b

S

T

c

t

w

I

4

a

a

which addresses aspects not considered in [110] such as the use of

sectorisation to avoid overlapping between multiple IEEE 80211ah

WLANs and the use of relays to further extend the IEEE 80211ah cov-

erage among other aspects Finally in [113] both IEEE 80211ah PHY

and MAC characteristics are introduced with emphasis on the ben-

efits of using the sub-1 GHz ISM band in terms of link budget the

location of the OFDM pilots for outdoors operation the use of bidi-

rectional transmission opportunities and the new packet formats to

minimise overheads among others

Enhanced power saving mechanism PS mechanisms for WLANs

were already considered in the development of the first IEEE 80211

standard with the goal of improving the lifetime of battery equipped

devices [114] In PS mode STAs keep the transceiver in sleeping mode

as much time as possible They periodically wake up to listen to the

beacons transmitted by the AP Those beacons indicate whether an

STA has packets waiting for it at the AP In the positive case that STA

remains awake and requests the delivery of those packets Otherwise

given it has nothing to receive it returns to sleep mode until the next

beacon is expected

In the IEEE 80211ah amendment time is divided into pages DTIM

(Delivery Traffic Indication Map) periods TIM (Traffic Indication Map)

periods and slots DTIM and TIM periods begin with the correspond-

ing DTIM and TIM beacons sent by the AP The functions of DTIM and

TIM beacons are described below

1 DTIM beacons They inform as to which TIM Groups (ie the

group of STAs assigned to the same TIM period) have pending

packets at the AP

2 TIM beacons Each TIM message informs a TIM Group about

which specific STA has pending data in the AP Between two

consecutive DTIMs there are as many TIM beacons as TIM

Groups

Using this DTIMTIM-based approach any STA can enter into a

power saving state if it does not have packets pending for transmis-

sion and one of two conditions is met (1) it observes in the DTIM

beacon that there is no downlink traffic addressed to its TIM Group or

(2) it observes in the DTIM beacon that there is some downlink traffic

addressed to its TIM Group but that STA does not explicitly appear in

the TIM beacon Compared to the preliminary IEEE 80211 PS mecha-

nism this approach reduces the size of the Traffic Indication Map in

each TIM beacon thus reducing the overhead and the time STAs need

to listen and process them In addition TIM periods can be organised

in pages which further increases the number of TIM groups between

two DTIM beacons

The temporal organisation of pages DTIM and TIM periods is re-

ported in Fig 8 which shows the division of TIM periods in RAW (Re-

stricted Access Window) and PRAW (Periodic RAW) The RAW is a

time interval in each TIM period where TIM stations can transmit and

receive data (see below the different types of STAs defined in IEEE

80211ah) In addition it can be divided into several downlink and

uplink slots for further granularity In the downlink the slots are as-

signed to a single STA or a group of STAs while the uplink slots are

randomly selected by the STAs with packets ready for transmission

The PRAW is the period of time in each TIM where non-TIM stations

can transmit and receive data

Types of STAs IEEE 80211ah supports three types of STAs TIM

non-TIM and Unscheduled STAs A TIM station is assigned to a TIM

Group Their data transmissions must be performed within a RAW

(Restricted Access Window) period Non-TIM stations do not have to

listen to beacons to transmit data During the association process

non-TIM devices directly negotiate with the AP to obtain a trans-

mission time allocated in a PRAW The following channel access can

either be renegotiated or occurs periodically depending on the re-

quirements set by the STA Unscheduled STAs do not need to listen to

any beacons similar to non-TIM stations Even inside any restricted

access window unscheduled STAs can send a poll frame to the AP

sking for immediate access to the channel The response frame in-

icates an interval (outside both restricted access windows) during

hich unscheduled stations can access the channel This procedure

s meant for STAs that transmit data very sporadically

Hierarchical station organisation To support a large number of STAs

nd their organisation in pages DTIM and TIM periods IEEE 80211ah

ssigns to each associated STA a unique identifier of 13 bits which is

alled the Association Identifier (AID) Using this new AID the maxi-

um number of supported STAs is increased from the original 2007

n IEEE 80211 to 8191 (= 213 minus 1) in IEEE 80211ah However it also

llows categorising STAs according to the type of application they are

xecuting their power level or even their desired QoS by assigning

hem to different TIM groups

Long sleeping periods IEEE 80211ah offers TIM Non-TIM and Un-

cheduled STAs the possibility to set very long doze times (up to

onths) The corresponding clock drift produced by such long doze

imes must be taken into consideration however as the higher the

ime an STA has been asleep the further in advance it should wake

p to avoid possible synchronisation problems with the network

Efficient small data transmission Three new enhancements have

een proposed to reduce the overhead when the data packet size

s small First while IEEE 80211 contains a 28-byte MAC header

EEE 80211ah proposes a short 18-byte version by using AIDs in-

tead of MAC addresses Second IEEE 80211 has defined several null

ata packet (NDP) frames which consist only of a PHY header These

rames can be used to create short ACKs short Block ACKs short CTSs

nd short PS-Polls Finally a Fast Frame Exchange mechanism has

een developed so that if an STA has data to transmit it can notify a

uccessful reception by transmitting its data frame instead of an ACK

Sectorisation Since the PHY layer is based on the IEEE 80211ac

mendment single and multi-user beam-forming are also supported

y IEEE 80211ah This allows the transmission of data to multiple

TAs simultaneously in the downlink increasing the system capacity

he use of the beam-forming capability of IEEE 80211ah APs is also

onsidered as a means to group STAs into different independent an-

enna sectors with the main goal of reducing interference issues This

ould be particularly useful in the case of overlapping with other

EEE 80211ah WLANs or in the presence of hidden nodes

12 Open challenges

A first open challenge is to understand the coverage and achiev-

ble transmission rates in IEEE 80211ah WLANs in both indoor

nd outdoor scenarios Propagation models for WLANs working at

B Bellalta et al Computer Communications 75 (2016) 1ndash25 13

f

c

8

p

R

p

i

m

l

a

m

c

p

p

f

m

t

a

e

p

t

m

a

l

T

m

a

i

n

t

t

t

u

p

c

w

f

a

t

8

a

m

a

s

l

a

m

b

m

d

S

i

a

o

a

R

t

n

G

fi

s

t

p

t

c

u

c

w

i

o

p

g

a

c

i

f

c

(

d

m

l

i

R

t

l

t

L

m

a

m

o

e

f

a

I

i

t

w

i

t

a

c

w

p

f

t

d

5

i

s

e

s

w

5

a

c

a

n

requencies lower than 1 GHz are evaluated in [115] The authors

ompare two path loss propagation models proposed by the IEEE

0211ah Task Group (one for macro and one for picohotzone de-

loyments) [116] with Lee and HatandashOkumura propagation models

esults show that the IEEE 80211ah channel models underestimate

ath loss with respect to Lee and Hata models Moreover in a compar-

son with empirical data it is observed that the IEEE 80211ah channel

odels also underestimate the initial loss and the slope of the path-

oss function A new model parameterisation is thus proposed by the

uthors In [117] the feasibility of an IEEE 80211ah WLAN deploy-

ent is also evaluated in terms of the achievable range and bit rate

omputed on the basis of the link budget using the same path loss

ropagation models as in [115] Results show that the transmission

ower limitations in the uplink can limit the overall network per-

ormance Finally in [118] the authors evaluate the achievable trans-

ission range for the different transmission rates and the achievable

hroughput for different combination of transmission power values

nd transmission rates Further studies are required due to the het-

rogeneity of scenarios in which IEEE 80211ah WLANs can be de-

loyed to characterise the effective coverage ranges with special at-

ention to mixed indoors and outdoors scenarios and in presence of

obile nodes

Due to the wide coverage IEEE 80211ah WLANs may be severely

ffected by the presence of hidden terminals To mitigate such a prob-

em smart solutions to guarantee that the STAs assigned to the same

IM Group are inside each other transmission range or the employ-

ent of RTSCTS-based solutions may contribute to mitigate such

problem An overview of long-range scenarios for IEEE 80211ah

s presented in [119] including a detailed description of the hidden

ode problem in them

Single-hop uplink transmissions from distant STAs require high

ransmission power to reach the AP Since not all STAs may be able

o transmit at the required power a solution could be the introduc-

ion of nodes able to relay transmissions from them However the

se of relays has to be efficiently harmonised with the operation in

ower saving mode allowing temporal periods in which the relays

an gather the data from their associated STAs and then periods in

hich they forward the data to the AP which is still an open challenge

or IEEE 80211ah WLANs In case relays are used data prediction and

ggregation techniques could be implemented to make more efficient

ransmissions [120]

Another challenge is the assessment of the efficiency of IEEE

0211ah PS mechanisms including their optimisation In [111] the

uthors evaluate the capacity of the TIM and page segmentation

echanisms in terms of the maximum number of nodes supported

nd the energy consumed given a certain network traffic profile Re-

ults confirm that a large number of STAs can be supported with

ow energy consumption The impact of the number of nodes oper-

ting in PS mode on the energy consumption and the delay perfor-

ance is analysed in [121] A detailed analysis of the TIM Group-

ased channel access adopted by the IEEE 80211ah task group is

ade in [122] where the STAs assigned to each TIM are uniformly

istributed between the slots of the RAW Since only one group of

TAs is allowed to transmit in each RAW slot the channel contention

s minimised The number of downlink and uplink slots in a RAW

s well as their duration is optimised on the basis of the number

f active STAs and the network traffic profile in [123] In [124] the

uthors propose an algorithm to determine the optimal size of the

AW interval for uplink transmissions based on the estimation of

he number of STAs transmitting and the duration of the RAW Fi-

ally in [125] the authors address the contention problems in Smart

rid communication networks with many nodes and periodic traf-

c when using an IEEE 80211ah WLAN It is still a challenge to con-

ider smart systems able to adapt the IEEE 80211ah parameters to

he instantaneous system state in order to save as much energy as

ossible

Since the PS mechanisms affect only the operation of TIM STAs

he performance of non-TIM and Unscheduled STAs has not yet been

onsidered in the literature To identify the scenarios in which the

se of non-TIM and Unscheduled STAs are of interest is still an open

hallenge as further investigating how non-TIM STAs negotiate over

hich PRAWs they can use to transmit and receive data Co-existence

ssues between the three types of STAs in a single network is also an

pen challenge that requires further work to guarantee a minimum

erformance level for all of them

The design of strategies to distribute the STAs between all the TIM

roups based on their specific traffic profile position battery level

nd application priority is another important challenge that is still

ompletely open EDCA is the default channel access scheme included

n the IEEE 80211ah amendment and provides some basic traffic dif-

erentiation capabilities at the packet level Besides the different ac-

ess categories (ACs) needing to be renamed and their parameters

Arbitration Inter-Frame Spacing or AIFS CWmin TXOP duration) up-

ated to fit the specific traffic profiles of M2M communications other

echanisms can be implemented A mechanism to assign the down-

ink RAW slots when packets from multiple priority levels are wait-

ng for transmission at the AP is required for example Also uplink

AW slots are currently selected randomly by the STAs that want to

ransmit a packet A mechanism to reserve some slots at each up-

ink RAW for high priority STAs may therefore be an option although

his may severely degrade the performance of the low priority STAs

astly different repetition patterns for different priority TIM groups

ay also allow priority STAs assigned to those priority TIM groups to

ccess the channel more often We expect this challenge will receive

uch attention in the upcoming years because of the heterogeneity

f sensors that will be connected though a single IEEE 80211ah AP

specially in urban scenarios

Similar to what has been discussed for 4G cellular networks used

or M2M communications [126] the development of mechanisms to

void and resolve congestion situations is another open challenge for

EEE 80211ah WLANs with many nodes For instance a mechanism

n which STAs have to wait a random number of DTIM periods before

hey can start a transmission may be implemented This approach

ould effectively distribute the traffic over a longer period of time

n overload conditions but otherwise would unnecessarily increase

he access delay Adaptive and load-aware solutions may play again

n important role to keep the optimal WLAN operation in all possible

ircumstances

Finally since IEEE 80211ah will compete with 4G5G cellular net-

orks and WSNs to provide M2M connectivity in many different ap-

lication domains such as smart cities or e-health comparative per-

ormance studies to determine the strong and weak points of each

echnology are required including also aspects such as the cost of the

evices and the system reliability

Cognitive radio technology for TV White Spaces

This section gives an overview of the IEEE 80211af amendment

ntroducing its most relevant features and open challenges Fig 9

hows the basic components and features considered for WLANs op-

rating in the TV White Spaces (TVWS) which include local spectrum

ensing by the AP and STAs and the use of geolocation data bases

ith information on channel availability

1 The IEEE 80211af amendment

Thanks to the transition from analog TV to digital TV several VHF

nd UHF spectrum channels used for decades for analog TV broad-

asting are now unused or are under-utilised in many geographical

reas [127ndash129] This ldquodigital dividendrdquo of spectrum has suggested

ational regulators such as the FCC in the US and Ofcom in the UK

14 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 9 WLAN operating in TVWS basic elements and functionalities

Fig 10 Hidden terminal problem in TVWS when the secondary network uses only in-

stantaneous channel sensing to decide whether or not a TV channel is occupied When

the STA is not able to detect the TV signal due to the presence of obstacles it may

initiate a transmission and create interference to the primary users

s

c

5

a

i

p

t

o

o

t

e

c

u

l

m

b

a

m

t

e

t

t

t

m

D

c

I

n

b

u

8

c

a

t

i

t

w

f

to discuss how to reuse these channels for unlicensed devicesrsquo com-

munications [130131] These potentially vacant channels in the VHF

and UHF bands are referred to as TVWS [132] and include spectrum

portions like the 470ndash790 MHz in Europe and non-contiguous 54ndash72

76ndash88 MHz 174ndash216 MHz 470ndash698 MHz and 698ndash806 MHz in USA

A snapshot of the TV spectrum occupancy in the city of Barcelona in

2012 is shown in [133]

The attractive characteristics of TVWS (not only for WLANs) in-

clude the ability to penetrate through walls and other obstacles much

more effectively than other widely used spectrum bands such as the

24 and 57 GHz ISM bands [134ndash137] This fact along with a progres-

sive spectrum scarcity in ISM bands has suggested the birth of the

IEEE 80211af amendment published in February 2014 [138] which

provides the IEEE 80211 operational characteristics for TVWS access

of unlicensed White Space Devices (WSD) A good summary of IEEE

80211af can be found in [137] The main advantage of operating IEEE

80211 WLANs in the TVWS comes from an increased coverage range

which can reach up to 1 km in rural areas and open fields [139]

and less energy needed to transmit However this comes at the cost

of an increased interference risk to other WSDs which creates co-

existence problems and demands new PHY and MAC layers to ef-

ficiently support channel access and operations preserving licensed

usersrsquo devices

WLANs operated in the TVWS could cover a number of interest-

ing and emerging use cases and scenarios eg Internet access in ru-

ral or sparsely populated areas Smart Grid sensor aggregation me-

tering and control Internet of Things advanced WLAN operations

and TVWS traffic offloading in indoor environments [140141] Nev-

ertheless the concept could be realised to create or extend com-

mercial or municipality WiFi services offered to citizens with cover-

age at the whole city level realised with reasonably limited network

infrastructures [142143]

IEEE 80211af will use the PHY and MAC layers derived from IEEE

80211ac It will adopt concepts such as the OFDM multi-user beam-

forming contiguous and non-contiguous channel bonding and packet

aggregation Among the mandatory and innovative behavioural and

operational parameters the most notable one is the channel acqui-

sition support realised through remote geolocation-based spectrum

allocation databases which maintain the channelsrsquo availability infor-

mation in any given area and time of day providing upon request the

list of free channels available for use

In the following section we highlight and summarise three main

novelties that IEEE 80211af introduces In Section 511 we explain

how the access infrastructure to the remote spectrum database is de-

signed and what the requirements are we discuss the coexistence is-

sues and methodologies adopted in IEEE 80211af and we also discuss

the novel concept of non-contiguous channel bonding The following

ections also contain an illustration of related works Open research

hallenges will be summarised in Section 512

11 Novel features

This section provides an illustration of novel features introduced

nd discussed in the path to IEEE 80211af and the directions provided

n related works and standardisation initiatives to resolve classical

roblems properly characterised in the new framework of TVWS

echnologies

Channel acquisition spectrum database and channel sensing Much

f the attention in operating in the TVWS is given to the protection

f the primary licensed users in the TVWS spectrum band In general

he primary users were considered as the Digital TV (DTV) broadcast-

rs and receivers In fact receivers would suffer the wasteful effect of

ollisions during the TV broadcast reception in the case of secondary

sersrsquo (SU) transmission interference (Fig 10)

It is a well-known problem that the definition of a coherent wire-

ess channel status (idlebusy) from the viewpoint of distributed pri-

ary users (eg DTV receivers) is made difficult by the difference

etween transmitters and receivers interpretation of signals in a vari-

ble time-space collision domain resulting in hiddenexposed ter-

inals concepts Solutions proposed to reduce the hiddenexposed

erminals included the RTSCTS mechanisms and beaconing How-

ver it is difficult to implement solutions that would effectively de-

ermine the channel status of TVWS channels based on localised dis-

ributed sensing activities performed by multiple secondary users in

he same collision domain On the other hand the use of RTSCTS

echanism would not be effective in this case because most of the

TV primary users (and specifically the receivers) were not originally

onceived to transmit signals or implement the RTSCTS handshake

n conclusion the actions to identify free (and busy) TVWS chan-

els should not be based on primary usersrsquo involvement but should

e almost totally implemented by specifically designed secondary

sersrsquo methodologies

There has been much debate on the methodology that IEEE

0211af secondary devices should adopt to get knowledge of free

hannels at a given time on a given target communication area in

n effort to guarantee verified optimised and reliable TVWS spec-

rum use The three main methodologies discussed include sens-

ng solutions geolocation databases (DB) and beaconing Moreover

wo approaches are possible in general distributed and centralised

ith their well known tradeoffs in terms of effectiveness required in-

rastructure deployment and coordination overhead The distributed

B Bellalta et al Computer Communications 75 (2016) 1ndash25 15

m

b

s

c

t

u

a

t

o

l

t

w

t

t

a

o

c

c

a

(

t

o

t

t

D

t

t

o

t

f

a

t

c

i

a

c

d

t

s

t

b

I

l

(

e

a

i

R

p

e

t

c

c

i

b

i

i

t

b

t

t

t

W

a

p

c

i

t

h

p

t

n

a

t

e

t

t

t

p

t

a

r

8

w

s

w

[

i

S

o

a

t

f

i

8

i

r

a

a

i

t

c

(

I

s

a

e

t

t

c

[

i

c

i

g

s

n

s

t

i

o

o

s

c

D

o

ethodology that is receiving more attention in the literature is

ased on distributed channel sensing solutions [127144145] These

olutions include the observation of spectrum use and possible

ooperative aggregation of spectrum sensing information to increase

he accuracy of detection and reduce the vulnerability of primary

sersrsquo transmissions The spectrum sensing capability is required in

ll the secondary devices with a minus114 dBm sensitivity Dynamic

ransmission power control must be provided and the upper limit

n emissions is 100 mW EIRP (20 dBm) for portable devices (further

imited in the case of adjacent channel use to reduce out-of-band in-

erference) However database (DB) coordination of TVWS combined

ith spectrum sensing is considered the most promising and effec-

ive technique compared with spectrum sensing alone The DB spec-

rum information has more chances to be effective (ie identifying

ll free channels opportunities) and reliable (ie not prone to attacks

r erroneous interpretation of channels status) In fact regulators de-

ided to push for the conservative solution of a remotely accessible

entralised spectrum DB that maintains the information on the avail-

bility of all TVWS channels at any given point in time and location

with a target accuracy around 50 m) under their responsibility For

his reason the spectrum databases have become a mandatory part

f many spectrum sharing systems and the dominant technical solu-

ion to support TV white spaces

The IEEE 80211af standard adheres to this vision and also has

he role of a common regulatory framework for different spectrum

B implementations defining a generalised coordination architec-

ure protocols and interfaces for spectrum queries and local spec-

rum information management The protocol explicitly considers out

f scope the communication protocol and technology adopted in

he background leaving the freedom to those players responsible

or the deployment to select any suitable Internet-based and local

ccess network technology Secondary devices that need to access

he TVWS spectrum should get information on which channels they

an effectively use and which parameters to adopt without impact-

ng primary usersrsquo transmissions (and receptions) by querying the

vailable spectrum DB The query must contain the transmission

haracteristics and accurate geographical position of the secondary

evice which can be determined by means of a geo-positioning sys-

em (GPS) regularly updated in the case of movement (or manually

et for static devices) The IEEE 80211af reference system architec-

ure for DB spectrum access is composed of multiple Geolocation-

ased Spectrum information Databases (GDBs) entities connected via

nternet to Registered Location Secure Servers (RLSS) RLSSs work as

ocal proxies of the GDB for a localised group of Basic Service Sets

BSS) RLSS are connected via a secure protocol architecture with the

quivalent of multiple Access Points (AP) in different BSS realising

trusted DB infrastructure APs locally coordinate the exchange of

nformation and channel access management between the GDB (via

LSS) and the secondary usersrsquo end stations (STA)

Mode I devices are those under the control of a device that em-

loys geo-location database access while Mode II devices are those

mploying geo-location database access by themselves The informa-

ion exchange provided between the GDB and a secondary user STA

an be provided in both an open-loop (eg adopted by FCC) and a

losed-loop (eg adopted by ETSI) implementation In an open loop

mplementation daily spectrum availability information is provided

y the GDB and no feedback on the spectrum information received

s provided by STAs thus the system approach for spectrum access

s much more conservative leading to low channelsrsquo utilisation po-

ential In a closed loop implementation the STA can provide feed-

ack to the GDB and there are more communication overheads due

o the high granularity of updates however the system is more effec-

ive and reliable in the exploitation of TVWS on behalf of STAs The

ypical information provided by the GDB includes (1) the updated

hite Space Maps (WSM) of frequencies allowed for secondary use

t the timespace of the querying STA and (2) the device-dependent

ower limitations for transmission (in general conservative and ac-

urate enough to avoid relevant interference effects on primary users

dentified in the area)

The basic IEEE 80211af mechanisms regulating the communica-

ion between STA and GDB can be found in [137138]

In parallel with IEEE 80211af standards such as the IEEE 19006

ave been created that consider interfaces and data structures sup-

orting spectrum sensing information exchange applicable to spec-

rum sensing (and particularly distributed spectrum sensing) sce-

arios In particular in October 2014 the IEEE 19006 WG initiated

project for a new standard called IEEE 19006b [146] concerning

he use of spectrum sensing information to support and optimise the

ffectiveness reliability and robustness of spectrum database solu-

ions The aim is to enhance the performance and capabilities of spec-

rum databases through the use of spectrum sensing information

Co-existence Support for co-existence mechanisms so that mul-

iple technologies can effectively utilise the TVWS spectrum is im-

ortant Self co-existence between network devices of a common

echnology (eg deployed by different operators in the same area)

nd co-existence among different technologies are relevant topic of

esearch for Cognitive Radio (CR) systems and specifically for IEEE

0211af Many solutions appeared on the research scene but no one

as so far finalised as the target solution for IEEE 80211af Specific

tandardisation has been started to regulate coexistence between

ireless standards of unlicensed devices including the IEEE 802191

147] The purpose of the IEEE 802191 standard is to enable the fam-

ly of IEEE 802 Wireless Standards to most effectively use TV White

pace by providing standard coexistence methods among dissimilar

r independently operated TVWS devices Early examples of gener-

lised coexistence mechanisms included Dynamic Frequency Selec-

ion (DFS) Transmission Power Control (TPC) listen before talk (eg

or contention based IEEE 80211 80215) time division multiplex-

ng (also among different techniques such as the IEEE 80216 80220

0222) and Message-based Spectrum Contention (that is beacon-

ng messages that carry coexistence information) Opportune met-

ics must be defined to assess the measurable coexistence achieved

mong different technologies as an example the hidden node prob-

bility for a target scenario or the estimate of percentage variation

n normalised network throughput and latency (before and during

he SU transmissions) On the other hand a centralised coexistence

ontrol mechanism could be effectively realised by a central manager

or coexistence-DB like the GDB) in critical scenarios To this end

EEE 19004 [148] (a standard for heterogeneous networks in dynamic

pectrum context part of IEEE Standards Coordinating Committee 41)

ims to standardise the overall system architecture and information

xchange between the network and mobile devices which will allow

hese elements to optimally choose from available radio resources

There are three possible classification of co-existence archi-

ectures and inter-network coordination channels for CR systems

entralised coordinated and autonomous co-existence mechanisms

149] In centralised co-existence schemes the co-existence is admin-

stered by a central entity (eg like in IEEE 802191) This solution

ould be applied in both homogeneous IEEE 80211af systems (eg

ncarnated by the centralised spectrum DB entity) and in hetero-

eneous inter-networks without requiring modifications to existing

tandards under the assumption that all the involved co-existing

etwork entities would adhere to the same centralised coordination

cheme and inter-network coordination protocol On the other hand

his assumption could be hard to satisfy in some practical scenar-

os The centralised schemes are effective in providing minimisation

f inter-network interference based on the availability and quality

f centralised co-existence information In coordinated co-existence

chemes the centralised entity could be present but not taking de-

isions in general The central entity could provide a co-existence

B with required information which can be used via in-band and

ut-band signalling for co-existence decisions taken by the cluster

16 B Bellalta et al Computer Communications 75 (2016) 1ndash25

c

e

n

c

o

s

i

d

a

S

8

n

a

b

o

b

i

t

a

8

b

[

t

i

n

b

5

t

a

c

b

s

a

c

n

s

p

i

c

l

r

c

f

a

v

t

a

e

c

t

Q

a

b

s

i

c

u

n

c

i

heads of many co-existing inter-networks for implementing a proper

common coordination protocol (eg out-band busy tone signalling)

These are often hybrid solutions realising a compromise between ef-

fectiveness and suitability In autonomous co-existence schemes all the

decisions and coordination are implemented in a distributed way

by the involved network entities In these schemes many policies

can be adopted to realise a sufficient level of co-existence under a

best effort approach Solutions that could be implemented include

the use of busy tones beaconing and other signalling protocols in

dedicated control channels dynamic distributed frequency selection

schemes listen before transmit token-based and dynamic reserva-

tion schemes All the above mentioned solutions have goods and

bads and a general illustration can be found in [149] Other proac-

tive co-existence techniques try to early detect and recovermitigate

co-existence issues with relaxed inter-network coordination eg re-

alised via spectrum sensing and interference avoidancesuppression

techniques

It must be clear that different aims exist regarding the co-

existence for primary TVWS users protection against secondary

users and for secondary devices mutual interference avoidance

Many co-existence problems have been analysed in the literature

for TVWS technologies in particular between the primary (licensed)

and secondary (unlicensed) devices The coexistence management in

TVWS is complex because primary users of DVB transmissions are in-

tended to be the pure receivers rather than the broadcasters In other

words potential protected users could be everywhere and the hid-

den terminal problem could arise for these systems The co-existence

problem is also complicated by many factors that cause asymme-

try and dynamicity in the TVWS such as the mobility variable den-

sity power asymmetry and heterogeneous MACPHY layers TVWS

co-existence problems originate between the IEEE 80211af technolo-

gies and the IEEE 80222 technologies for wireless regional area net-

works (WRAN) Results of analysis of TVWS usage in Europe show

that white spaces are typically present and fragmented They are typ-

ically more abundant in rural areas where larger contiguous blocks

of unused channels are available due to broadcast network planning

giving priorities linked to population density However the exploita-

tion of TVWS in urban areas is possible eg some recent research

was realised under the assumption of exploiting shadowing effects

created in the communication environment (eg by buildings obsta-

cles) in favour of frequency reuse [150]

Since different standards for opportunistic communication in the

TV White Spaces have now been published such as IEEE 80222 [151]

IEEE 802154m [152] Weightless [153] and of course IEEE 80211af

improved methods to guarantee the coexistence of different devices

operating on several protocols in the same bands must be deployed

We highlight here some work eg [150154ndash157] that focuses on

the coexistence between different technologies operating in the same

bands In particular [155] studies the performance degradation of

IEEE 80222 when an IEEE 80211af network is operated in the same

area The problem becomes even worse when the IEEE 80211af net-

work is located near 80222 user equipment causing both networks

to perceive a strong interference from each other One proposed solu-

tion is the Coexistence Beacon Protocol [158] studied for 80222 net-

works which foresees the exchange of a periodic beacon to identify

the neighbouring and possibly interfering networks

For what concerns the co-existence problems research has ad-

dressed other important related aspects of new WLANs on TV White

Spaces For instance the higher coverage range makes them attrac-

tive for applications in the smart grid and also for M2M on TVWS

[134150159] Here the increased range compared to standard tech-

nologies in the ISM bands can have beneficial effects for indoor

mobile devices [150] and M2M [159] Specifically indoor devices

communicating on TV White Spaces offer better propagation char-

acteristics and penetration through obstacles making it easier to

realise scalable home connectivity [135] although interference and

o-existence problems could be exacerbated and must be resolved

g see [139] However studies in the literature show how the sig-

al coming from an indoor transmitter on TV White Spaces remains

onstrained inside the house making it difficult to deploy indoor-to-

utdoor networks On the other hand this shadowing limit can be

een as an opportunity lowering the interference outside the build-

ng in which the TVWS network is locally operated [134150] Un-

er this approach HDTV streaming has recently been considered as

possible use case of WLAN networks operating in the TV White

pace [160]

Non-contiguous channel bonding A relevant novel feature of IEEE

0211af is the potential for contiguous and non-contiguous chan-

el bonding which permits aggregating basic channels (also non-

djacent ones) and leveraging the possible large frequency spread

etween multiple available channels Due to the rather static nature

f primary DTV transmissions where a busy channel is unlikely to

ecome free in the near future and state changes are coarse-grained

n general it is crucial to exploit the time-locality effect and exploit

he maximum physical channel availability that could be aggregated

t any given location With the methodology inherited from IEEE

0211ac IEEE 80211af is capable of bonding together two up to four

asic channels grouped in up to two different non-contiguous chunks

137] The spectrum bandwidth of a DVB-T basic channel can be ei-

her 6 7 or 8 MHz depending on the country in which the service

s operated As an example this creates a 144 to 168 OFDM chan-

elsrsquo bandwidth potential when up to four 6ndash7ndash8 MHz channels are

onded [137]

12 Open challenges

The deployment of WLANs in the TVWS at their maximum poten-

ial still requires the resolution of open problems Generally speaking

s mentioned in previous sections the challenges for such networks

an be divided into three main categories (1) spectrum sharing

etween opportunistic (secondary) and heterogeneous devices (2)

pectrum sensingmanagement and maximum exploitation of avail-

ble spectrum (potentially enabling the spectrum-on-demand con-

ept) and (3) co-existence and interference mitigation to the primary

etwork (primary user protection) and secondary networks [154]

The overall aim of spectrum sharing techniques is to maximise

eparation to avoid overlapping or contiguous channels operations

rovided that sufficient channels are available in the TVWS spectrum

n a given spacetime scenario [154] In general important research

ontributions have to be realised for the design of proper spectrum al-

ocation methodologies and techniques satisfying multi-factorial QoS

equirements at the system-level and the user-level Novel ideas in-

lude the spectrum sharing and spectrum sensing techniques being

urther divided into cooperative and non-cooperative

Non-cooperative solutions attempt to realise spectrum sharing

nd sensing on a local basis without direct cooperation between de-

ices Examples include the ldquolisten-before-transmitrdquo approach and

ransmission power control to limit the interference and spectrum

llocation policies that are based on local node feedback only Coop-

rative solutions rely on the existence of a common communication

hannel through which devices can tentatively agree on the spec-

rum allocation that provides the desired spectrum separation and

oS Major issues include the definition of a methodology to identify

common channel (or multiple channels mutually shared in space

y pairs of heterogeneous and distributed devices) and to efficiently

hare common information For the access to the spectrum DBs the

dentification of proper common access channels is still under dis-

ussion out-of-TVWS-band cellular communications can be widely

sed given their high coverage also indoor

Trying to avoid the use of complex and resource-hungry coordi-

ation schemes and dedicated control channels rendezvous proto-

ols play a significant role both for co-existence and for establish-

ng dynamic communication channels based on spectrum sharing

B Bellalta et al Computer Communications 75 (2016) 1ndash25 17

R

n

c

t

l

o

m

e

n

i

n

w

n

a

u

s

p

b

s

m

s

a

d

t

b

a

a

m

a

t

u

8

n

t

d

d

m

f

t

n

d

n

e

r

a

h

b

f

a

c

t

u

i

f

t

d

t

s

i

f

I

l

i

a

n

c

d

I

n

o

t

i

t

e

w

t

c

m

t

a

i

t

q

b

c

a

i

l

i

a

v

t

a

a

c

t

i

t

b

t

t

a

D

p

c

s

o

q

r

p

a

g

t

e

u

d

t

m

n

t

s

f

i

1

8

f

T

endezvous is considered a fundamental problem in generalised CR

etworks at the basis of many communication processes eg in-

luding neighbour discovery routing and broadcast Rendezvous pro-

ocols attempt to establish a new link for communication on a se-

ected and agreed frequency band (channel) identified within a set

f available resources (channels) thus creating the basis for com-

unication between two or more SUs This problem is often exac-

rbated by the high (and unknown) numbers of contending SUs high

umber of available channels and their ambiguous boundaries and

dentifiers and tough co-existence requirements satisfaction in cog-

itive radio systems [161] Preliminary approaches for rendezvous

ere based on the existence of a dedicated Common Control Chan-

el (CCC) agreed among all the SUs or the existence of a centralised

gent (decision maker) assigning channels to any pairs of requesting

sers This approach could be considered in the case of IEEE 80211af

olutions based on the spectrum DB implementation however the

roblem is complicated by the fact that the common channel must

e identified and made available for all the requesting SUs at the

ame time even when these are spread over multiple collision do-

ains in space On the other hand the scalability and reliability is-

ues which are caused by the congestion of CCCs and the vulner-

bility risks of centralised DBs motivate the research of alternative

istributed solutions called blind rendezvous protocols [162] Dis-

ributed solutions have been mainly based on beaconing andor slot-

ased channel hopping algorithms exploiting randomisation sensing

nd discoverysynchronisation messages spread in slotted time over

set of candidate available channels [161] These protocols aim to

inimise the convergence time to maximise the use of resources

nd to avoid collisions among SUs by identifying and agreeing on

he use of a given common channel among N non-overlapping and

nambiguously labelled channels The main complication in IEEE

0211af systems is given by the wide availability of many different

on-contiguous spectrum bands which cannot be uniformly quan-

ised in single channels and unambiguously labelled in space in or-

er to realise a common reference domain for all the IEEE 80211af

evices In fact the knowledge of the number of channels their com-

on labels the unique IDs of SUs and the number of SUs contending

or the channel assignment are the key factors determining the effec-

iveness and good properties of blind rendezvous protocols Recently

ew classes of rendezvous protocols have been proposed which are

istributed blind and oblivious Oblivious means that available chan-

els may be identified with different labels on behalf of the differ-

nt involved SUs A number of these distributed blind and oblivious

endezvous protocols have been proposed and have been referenced

nd analysed in [162] To conclude solutions for dynamic spectrum

andoff issues (eg in case PUs suddenly appear in a region) must

e resolved in advance to minimise the latency of re-establishing a

unctional communication channel eg by implementing a proactive

lternative rendezvous channel selection in background to ongoing

ommunication processes [163]

Another challenge is to provide cooperative techniques in a dis-

ributed vs centralised implementation by analysing their convergence

nder variable conditions overheads and tradeoffs In some cases as

n vehicular networks the exploitation of scenario characteristics and

actors such as the constrained mobility could help to realise effec-

ive dissemination of sensing information over an extended and pre-

ictable horizon under a cooperative approach [164ndash166] Coopera-

ion could be helpful in the realisation of the cooperative spectrum

ensing in combination with the Geolocation DB approach Another

nteresting direction of research is the investigation of mutual ef-

ects of coexisting cooperation-based vs non-cooperative techniques

n cooperation-based devices the exchange of information could al-

ow a quick convergence to a solution and avoidance of further

nterference [145] Certainly cooperative techniques have desirable

dvantages over non-cooperative ones However tight time synchro-

isation and high coordination overheads are required The common

hannel for cooperation purposes could become a bottleneck and re-

uce the potential advantages of dynamic spectrum allocation The

EEE 80219 foresees the presence of a shared common control chan-

el (CCC) to which devices need to tune in order to gather the state

f the network Solutions such as [141] propose the use of TVWS for

he CCC

A promising direction for research is the adoption of cluster-

ng schemes for secondary devices associated with a hybrid dis-

ributedcooperative vs centralisednon-cooperative approach in an

ffort to reduce overheads while maximising advantages In this

ay secondary nodes could implement cooperation for sensing func-

ions using properly identified common channels and delegate the

entralised decisions and spectrum-DB management to well instru-

ented cluster-leader nodes This concept is incarnated to some ex-

ent by RLSS in IEEE 80211af More recently there have been propos-

ls to build distributed databases that refresh their contents period-

cally by querying the remote spectrum database [167] Here Mas-

er devices (according to Ofcom terminology) periodically cache the

uery replies from the spectrum database in order to reply to and

roadcast the spectrum availability to neighbour Slave devices

Another challenging direction for research is the definition of ac-

urate models and efficient simulation tools enabling dynamic spectrum

nalysis for both rural and urban areas Recent developments of dig-

tal maps simulation tools and propagation models theoretically al-

ow improving the capability of predicting radio propagation effects

n complex scenarios with an acceptable computation time This en-

bling technology could be used in parallel to spectrum DBs to pro-

ide more accurate forecasting of dynamic frequency allocation in

imespace scenarios Another issue to be taken into account is the

ccuracy of the remote spectrum database regarding the channel

vailability estimation Several works have already shown the inac-

uracy of propagation models due to the complexity of parameters

o be considered and differences between modelled and real scenar-

os [132134168169] Better propagation models will make it possible

o more efficiently estimate the interference between devices and

uild Radio Environment Maps (REM) or White Space Maps (WSM)

hat will possibly lead to a more accurate sharing of the radio spec-

rum [144] To this end current research could focus on how to lever-

ge simulation models in order to build more accurate frequency

Bs

In general the adoption of remote spectrum DBs methodology

ushed by national regulators has made the spectrum sensing pro-

ess optional and conservatively realised (eg power control for

ensing-only devices is limited to below 17 dBm EIRP) Thus much

f the research has focused on techniques and solutions to successfully

uery the remote database also in challenging environments such as

ural areas and indoor However since the queries must provide the

osition of the mobile device the accuracy of positioning for mobile

nd handheld devices still needs more appropriate solutions Trian-

ulation can be used to estimate the position of mobile devices with

he accuracy needed by the current regulations outdoor (50 m) How-

ver a low GPS accuracy is possible indoors [170] Certified man-

al positioning can be provided by technicians for static and indoor

evices although limiting mobility Alternatives to costly infrastruc-

ures for positioning (eg based on short range beaconing devices)

ust be deployed in indoor scenarios

Finally regarding spectrum sensing and interference mitigation

ew challenges come from the heterogeneity and different charac-

eristics that networks operating in the TV White Spaces can have

uch as different transmitting power (4 W for static devices 100 mW

or mobile and portable devices and 40 mW for communication

n channels adjacent to occupied ones) different bandwidths (5

0 15 and 20 MHz are currently foreseen for IEEE 80211af IEEE

02154m can have narrower bandwidths or wider ones) and dif-

erent medium access schemes (protocols that use either CSMA or

DMA are required to co-exist in the TV White Spaces) With all this

18 B Bellalta et al Computer Communications 75 (2016) 1ndash25

d

a

T

i

a

c

t

l

b

o

t

a

d

p

a

r

p

n

a

m

t

p

p

d

d

o

t

i

D

w

p

i

a

a

d

b

fi

d

o

t

l

b

t

b

a

i

e

o

o

c

t

t

w

A

c

M

s

c

l

O

w

g

v

t

in place the problem of determining whether another secondary de-

vice is currently transmitting in the same channel is still an open

issue

Spectrum sensing is a very hot topic in the literature with many

proposals that leverage the cooperation between devices to raise the

accuracy of the detection process eg [144171] The current litera-

ture also offers room for proposals that suggest exploiting sensing to

build more accurate spectrum DBs [164172173] Lastly regarding the

interference mitigation to the primary network regulators provide a

conservative approach based on a remote spectrum database to avoid

high risk for the primary users However in particular for the indoor

scenario the concept of secondary usersrsquo transmission in TV Grey

Spaces has been proposed TV Grey Space identifies busy channels

that are formally used by primary users in the area of interest (eg

at the rooftop level of a building for DVB-T) but which could be con-

sidered usable indoors (eg at the basement level) without causing

effective interference to the primary users [134150174] This concept

is based on the generalised spectrum underlay paradigm [175] The

remaining research challenge is to provide TV Grey space access guar-

anteeing a sufficient amount of protection to the primary user while

providing a satisfactory QoS to opportunistic devices eg through

the use of customised and validated propagation models for indoor

TVWS networks and feedback mechanisms to constantly monitor the

interference generated to the primary network

6 Emerging new trends and technologies

In this last section we review three emerging trends that in our

opinion will have a large influence in the conception of future WLANs

as they change the way WLAN protocols and functionalities are de-

veloped implemented tested and integrated with other wireless

networks

61 Programmable Wireless LANs

Especially in the enterprise environment WLAN deployments

need to support a wide range of functionalities and services This is

intrinsically difficult because of the large number of APs that must be

managed which calls for scalable solutions Typical services include

channel assignment load balancing among APs authentication au-

thorisation and accounting (AAA) policy management support for

client mobility and interference coordination Another problem is the

fact that WLAN clients autonomously take several decisions such as

which APs to associate with when to hand-over etc Therefore sup-

porting roaming clients requires the management of a large number

of association states across several APs which is a challenge if sup-

port for real-time hand-over is desired Typically such management

schemes are centralised and most of them are proprietary such as

WLAN controller solutions from Aruba [176] or Cisco [177] although

the IEEE 80211u amendment has been released to allow mobile users

to seamless roam between WiFi networks with automatic authenti-

cation and handoff [178] For example Dyson [179] enables STAs to

send information such as radio channel conditions to a centralised

controller based on a custom API (eg based on Python) As the con-

troller has a centralised view of the network it can enforce a rich

set of policies to control the network also using historical informa-

tion A demo system has been implemented along with applications

such as airtime reservations for specific clients or optimised hand-

offs However Dyson requires STAs to be modified in order to use

those new services offered by the centralised controller TRANTOR

[180] is another example of a centralised management system that

requires changes of clients in order for the infrastructure to gather

information from them in terms of eg interference measurements

In addition control commands enable the infrastructure to exercise

control such as modifying the transmit power or influencing the as-

sociation procedure Clients still use standard CSMA MAC layer for

ata transmission In contrast to DenseAP [181] or DIRAC [182] which

re also based on a centralised controller but do not modify clients

RANTOR exercises much more control on clients by the use of a ded-

cated API which enables a significantly higher gain in coordination

nd thus capacity CENTAUR [183] is another example of a centralised

ontroller that centrally schedules hidden and exposed terminals in

he downlink while using standard CSMA MAC for uplink traffic and

egacy downlink traffic To achieve good performance it uses a fixed

ack-off packet staggering techniques and a hybrid data path that

nly schedules downlink transmissions towards hidden and exposed

erminals centrally All other traffic is sent using standard DCF with

standard back-off procedure In contrast to previous approaches it

oes not require any modifications to the clients but does require data

lane centralisation and it remains questionable how scalable such

solution is for large WLAN deployments with high PHY layer data

ates beyond 100 Mbps However new approaches for fast data plane

rocessing that are currently explored in the virtualisation commu-

ity which move all the packet processing into user space may be

n option to significantly increase speed Designing such manage-

ent systems raises several interesting research questions (a) Cen-

ralised infrastructures may incur high latency but are more simple to

rogram and maintain while designing a completely distributed ap-

roach that operates close to WLAN devices may result in significant

istributed coordination and consistency problems (b) what level of

istributed control is required in order to support the flexibility needs

f future WiFi based networks supporting a high level of mobility

Traditionally WLAN APs are built on proprietary operating sys-

ems that are tightly coupled with the hardware This design makes

t hard to create new applications on top of such networking devices

espite the fact that protocols and mechanisms are available that

ould greatly improve the utility of existing networks those new

rotocols are not deployed because the closed system design makes

t very difficult to extend their functionality An important aspect for

ll such new approaches is therefore how to provide open interfaces

nd open source to speed up innovation As an example Linux-based

evices are completely open source but in order to increase flexi-

ility much more work is needed in the area of open drivers and

rmware In fact the Atheros based ath9k driver has been the main

river of innovation in the open source community because it is the

nly driver that interworks with open firmware In general solutions

hat modify the OS driver focus mainly on programmable network

evel solutions for WLANs such as channel switching or handovers

etween APs While the ath9k driver is an excellent example in how

o speed up innovation it is still very cumbersome to support flexi-

le MAC engine reprogrammability with ath9k In contrast MAClets

nd Wireless MAC processors (WMP) [184] allow a much more flex-

ble reprogramming of MAC functionalities A MAC processor is an

ntity able to execute general MAC commands that specify the MAC

perations through a software-defined state machine The behaviour

f the MAC protocol can therefore be updated at run-time by simply

hanging the sequence in which those commands are executed (ie

he MAClets) As a proof-of-concept the authors of [184] implement

he proposed MAC processor solution in a commodity WLAN hard-

are card extending the basic DCF in three directions piggy-backing

CKs a pseudo TDMA and the use of multiple channels In [185] a

ontrol framework for this WMP system is also proposed to support

AClet code mobility ie for moving loading and activating MAC

oftware programs embedded into ordinary data packets (akin to the

apsule model of traditional active networks)

The difficulty in re-programming networking hardware has also

ed to the concept of Software Defined Networks (SDNs) based on the

penFlow protocol [186] The main idea of SDNs is to extend net-

orking devices with standardised APIs that allow third-party pro-

rammers to flexibly control the data path In addition SDNs pro-

ide higher level abstractions to network designers and programmers

hrough the use of a centralised control plane offered by a network

B Bellalta et al Computer Communications 75 (2016) 1ndash25 19

Fig 11 The CLOUDMAC architecture VAP stands for Virtual Access Point and WT

stands for Wireless Termination

c

a

c

t

m

W

n

t

t

c

p

t

u

h

a

i

e

s

o

c

a

a

f

T

r

s

a

s

p

o

M

C

i

T

t

fl

A

a

a

i

c

t

s

t

e

I

t

w

H

e

b

I

l

d

o

6

r

W

s

l

a

e

e

a

s

b

c

t

(

S

p

c

F

B

t

p

T

e

t

s

u

m

i

e

o

a

A

p

d

t

t

n

p

i

t

t

f

i

ontroller such as NOX [187] which allows reuse of components such

s topology discovery or network access control for different appli-

ations The SDN concept has recently been applied to WLAN archi-

ecture in order to enable fine grained control over mobility manage-

ent and data forwarding focusing on a programmable enterprise

LAN architecture An important part of the SDN architecture is the

etwork controller which provides a centralised view of (parts of)

he SDN enabled network and uses the OpenFlow protocol to install

he forwarding rules on SDN-enabled devices (routers switches ac-

ess points cellular base stations etc)

ODIN [188] is designed to support programmability in enter-

rise WLAN architecture by separating the association state from

he physical AP They implement Light Virtual Access Points (LVAP)

sing a Split-MAC approach where the infrastructure controls the

andover procedure among different WLAN APs By managing

ssociations through SDN controllers ODIN enables proactive mobil-

ty management and load balancing within the SDN enabled WLAN

nterprise network without the need for changes in the client WLAN

tack or IEEE 80211 MAC layer While ODIN requires agents to reside

n the APs that communicate with the Odin Master within the SDN

ontroller CLOUDMAC [189] is based purely on the concept of SDN

nd virtual APs Similar to ODIN CLOUDMAC implements a Split-MAC

pproach but in addition enables the processing of WLAN MAC layer

rames within a co-located Cloud using so-called Virtual APs (VAPs)

he physical APs in CLOUDMAC are lightweight WLAN APs that are

esponsible only for sending out IEEE 80211 based MAC layer ACKs to

tandard WLAN clients and tunnel WLAN MAC layer frames through

n SDN-enabled enterprise WLAN towards the VAPs As association

tates are kept in the VAPs fast mobility is supported using sim-

le OpenFlow forwarding rules Because of the additional processing

f IEEE 80211 WLAN MAC frames in the co-located Cloud CLOUD-

AC has higher latency than a standard WLAN deployment However

LOUDMAC offers a Webservice based API to third party applications

n order to program the enterprise WLAN and enable new services

he architecture of CLOUDMAC is depicted in Fig 11 It has been ex-

ended in [190] to support QoS management and in [191] to support

exible MAC management frame prioritisation based on IEEE 80211

system based on OpenFlow has been recently proposed in [192] to

llow the station to be associated with multiple APs simultaneously

nd to switch between APs with low overhead

An important aspect to consider in order to enable programmabil-

ty is backwards compatibility Several approaches (eg [180]) require

lients to be modified to utilise the features provided This is difficult

o do in practice because it requires changes in the operating system

oftware of all clients If not all clients can utilise those APIs it is ques-

ionable how usable the new architecture will be and how much ben-

fit in terms of aggregate performance such architecture will allow

n contrast the SDN based approaches are interesting in the sense

hat they do not require changes in the client WLAN stack and work

ith the standard IEEE 80211 MAC layer deployed within the clients

owever it remains questionable how scalable such solutions are For

xample an interesting open research topic is to evaluate the scala-

ility of approaches such as CLOUDMAC [189] In addition processing

EEE 80211 MAC frames within co-located private Cloud requires low

atency support from local Cloud solutions (such as OpenStack) in or-

er to reduce the IEEE 80211 MAC processing time which is an area

f active research

2 Prototyping and testing IEEE 80211 enhancements

Most of the new proposals for next-generation WLANs are cur-

ently only evaluated using mathematical analysis and simulation

hile both analysis and simulation are necessary to characterise and

tudy those enhancements in the initial design phase or to consider

arge-scale scenarios it is difficult to consider all practical aspects of

real-world scenario This can sometimes cause significant differ-

nces between what simulations and real experiments show How-

ver real experiments are challenging because of the high complexity

nd costs of building the new hardware and software for each specific

olution to test

To mitigate implementation complexity there are several flexi-

le hardware platforms where low-level MAC mechanisms can be

ompletely implemented in software Examples based on FPGAs are

he USRP (Universal Software Radio Peripheral) [193] and the WARP

Wireless Open-Access Research Platform) [194] As an alternative

ORA (Microsoft Research Software Radio) [195] works on general

urpose computers by taking some advantage from modern multiple

ore systems

A different approach is provided by OpenFWWF [196] Open-

WWF provides an open CSMACA firmware for specific models of

roadcom chipsets so the resulting firmware can be uploaded and

ested in real commercial hardware OpenFWWF implements a sim-

le State Machine (SM) for controlling the hardware in real time

he SM evolution is driven by a main loop that reacts to events by

xecuting specific handlers When a packet originally prepared by

he Linux kernel is ready in the NIC memory handler Packet_Ready

ets up the radio hardware according to the packet metadata sched-

les the transmission and jumps back to the main loop The Trans-

ission Engine (TXE) then takes care of accessing the channel ie

t decrements the back-off counter according to the DCF rules and

ventually starts the actual transmission This triggers the execution

f the TX_frame_now event that prepares the ACK time-out clock

nd finalises the MAC header If the ACK-frame is received or if the

CK time-out expires and the maximum number of attempts for this

acket is reached handler Update_Params resets the contention win-

ow to its minimum value or otherwise doubles it Finally it loads

he back-off counter with a fresh value

In the following we will review a list of selected papers that use

he OpenFWWF firmware and the WARP or USRP platforms to test

ew proposals for WLANs in real scenarios as examples A first im-

lementation of the groupcast mechanisms defined in IEEE 80211aa

s presented in [197] The authors modify the OpenFWWF firmware

o include those functionalities and evaluate them using a 30 STAs

estbed A second example of the use of OpenFWWF is the collision-

ree MAC protocol presented in [198] In [198] the authors design and

mplement in OpenFWWF a MAC protocol that is able to achieve a

20 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 12 LTE to WIFI offloading

a

n

c

t

[

d

a

W

v

t

t

o

b

f

S

g

l

i

I

s

I

b

n

s

m

a

p

d

c

N

o

v

t

t

n

l

i

a

t

a

collision-free operation by waiting a deterministic time after success-

ful transmissions before to start a new transmission This work shows

the experimental performance of a collision-free MAC (CF-MAC) pro-

tocol for WLANs using commercial hardware The results show that

the proposed CF-MAC protocol leads to a better distribution of the

available bandwidth among users a higher throughput and lower

losses than DCF

In [199] the authors use the WARP platform to test a variant of the

DCF ndash called IEEE 80211ec ndash that substitutes control packets such as

the RTS CTS and ACK with short detectable sequences Since those se-

quences are shorter than the control packets and can be detected cor-

rectly even at lower SNRs values a significant gain in performance is

achieved MIDAS (Multiple-Input Distributed Antenna Systems) [200]

implementation using the WARP platform shows the benefits of dis-

tributing the antennas over the area to cover instead of co-locating

them at the AP in terms of capacity when MU-MIMO is employed

The authors also propose a new MAC protocol to benefit from the

spatial reuse that the DAS allows taking as a basis the IEEE 80211ac

Experimental results show that their proposal is able to achieve up

to 200 gains versus the traditional approach where all antennas are

co-located at the AP

The USRP platform has recently been proposed to develop a first

implementation of IEEE 80211ah in order to obtain a preliminary ex-

perimental performance assessment of such a technology because

no commercial off-the-shelf hardware is yet available as the amend-

ment is still in progress [201] The USRP platform is also used in

[202] to evaluate TIMO (Technology Independent Multi-Output) a so-

lution to deal with high-power non-IEEE 80211 interferers in ISM

bands

To conclude this section it is worth mentioning here also to

MAClets and WMP as given such flexibility to implement and dis-

tribute new MAC protocols and other IEEE 80211 functionalities we

believe that if a use-friendly implementation of such a MAC proces-

sors framework is provided it would significantly contribute to the

development and testing of new MAC enhancements by the research

community

63 CellularWLAN interworking

Public hotspots that offer Internet access over a WLAN using IEEE

80211 technology are now nearly ubiquitous It is forecasted that the

cumulative installed base of WiFi hotspots worldwide will amount

to 551 million by 2018 excluding private hotspots (eg WiFi access

points deployed at home) [203] The sharp increase in the availabil-

ity of public WiFi was initially perceived by mobile cellular operators

as a threat due to the additional competition from wireline Internet

service providers or emerging crowdsourced WiFi networks such as

FON5 However as cellular operators are fighting to cope with the ex-

plosion of mobile data traffic created by the rising use of multimedia

content traffic over mobile devices [15] they are also starting to use

WLANs based on the IEEE 80211 technology to offload data from their

core and access networks In general mobile data offloading refers to

the use of complementary network technologies (in licensed or un-

licensed spectrum) for delivering data originally targeted to cellular

networks Intuitively the simplest type of offloading consists of ex-

ploiting connectivity to existing co-located WiFi networks and trans-

ferring data without any delay (Fig 12) Thus this offloading tech-

nique is known as on-the-spot offloading As a consequence of this

new trend the seamless integration of cellular (eg 3GLTE) and WiFi

technologies has attracted significant research interest in recent years

(see [204] for a survey) a few solutions have already been standard-

ised [205206] and roaming between cellular and WiFi is becoming

increasingly transparent to end users CellularWLAN interworking is

5 httpscorpfoncomen

fl

t

f

lso fostered by the support in the evolving 4G standards of heteroge-

eous network deployments (HetNets) in which the existing macro

ells are complemented with a number of small low-power base sta-

ions with the goal of increasing capacity in highly congested areas

207208] It is envisaged that small cells will be based on 4G stan-

ards (eg pico and femto cells) as well as IEEE 80211 technologies

nd multimode base stations that work simultaneously with LTE and

iFi are already entering the market

It is important to point out that LTE standards already support a

ariety of mechanisms that enable data offloading However most of

he existing solutions such as Local IP Access (LIPA) and Selected In-

ernet IP Traffic Offload (SIPTO) focus on data offloading at the core

f the network [205] For instance LIPA allows an IP-enabled mo-

ile device to transfer data to another device in the same pico or

emto cell without passing through the cellular access network while

IPTO enables the routing of selected IP data flows through different

ateways The only LTE offloading mechanism that supports seam-

ess interworking with IEEE 80211-based WLANs is IP Flow Mobil-

ty and Seamless Offload (IFOM) Specifically IFOM relies on Mobile

Pv6 technologies to allow a user terminal to simultaneously route

elected IP flows over different radio access technologies [209210]

n this way a user terminal can offload selected flows to a WLAN

ased on some operator-defined policy while keeping the LTE con-

ection running However to enable such an approach it is neces-

ary to support an entity in the cellular core network that can com-

unicate with the user terminal to exchange information about the

vailability and quality of neighbouring access networks as well as to

rovide the user terminal with predefined rules to manage the han-

over process This entity in current LTE standards is called the Ac-

ess Network Discovery and Selection Function (ANDSF) server [211]

ote that the problem of selecting the best communication technol-

gy in an heterogeneous wireless network has been extensively in-

estigated in the literature both with centralised as well as decen-

ralised approaches (see [212] for a survey) However the use of mul-

iple interfaces in parallel as well as per-flow offloading are relatively

ew concepts Thus the design of scalable and efficient network se-

ection strategies for the ANDSF framework is still an open issue For

nstance three offloading methods suitable for the ANDSF framework

re proposed in [213] which are based on coverage SNR and sys-

em load A reinforcement learning approach is designed in [214] that

llows multimode base stations to autonomously steer their traffic

ows across different access technologies depending on the traffic

ype the usersrsquo QoS requirements the network load and the inter-

erence levels A number of studies have also tried to quantify the

6 B Bellalta et al Computer Communications 75 (2016) 1ndash25

g

i

l

a

b

b

f

o

w

o

H

t

e

n

a

f

t

m

a

w

s

3

a

8

2

c

s

m

W

r

a

[

o

m

2

n

c

p

r

a

creating the groups of STAs for DL-MU-MIMO transmissions smart

packet schedullers able to decide when the use of DL-MU-MIMO out-

performs SU-MIMO transmissions and the implementation of the

TXOP sharing feature between several ACs The results and conclu-

sions obtained in both cases will be very valuable in the development

of IEEE 80211ac technologies as well as in the conception of the fu-

ture amendments that will substitute IEEE 80211ac in four to five

years such as the recently initiated IEEE 80211ax

Following the first mentioned research direction there are several

efforts that have focused on understanding both theoretical and exper-

imental performance bounds of IEEE 80211ac The maximum downlink

throughput that an IEEE 80211ac AP can achieve when packet aggre-

gation channel bonding and different spatial stream configurations

are considered is presented in [33] In [34] the authors evaluate the

IEEE 80211ac performance experimentally using commodity devices

focusing on the effects that the use of wider channels the 256-QAM

modulation and the number of SU-MIMO spatial streams have in

terms of throughput and energy consumption It is worth mentioning

that DL-MU-MIMO was not yet implemented in the equipment they

were using and that feature was therefore not considered The evalu-

ation of a DL-MU-MIMO implementation for WLANs using the WARP

platform is presented in [35] where a deep evaluation of the potential

benefits of DL-MU-MIMO transmissions is done in terms of the loca-

tion of the receivers number of users and user mobility among other

aspects A solution that combines both packet aggregation and DL-

MU-MIMO transmissions is presented in [36] Results show the need

of properly dimensioning the buffer space to achieve the full potential

of such a combination In [37] the authors compare the throughput

achieved by IEEE 80211n and IEEE 80211ac when packet aggrega-

tion is used with and without channel errors They show that in most

cases the packet aggregation mechanism introduced in IEEE 80211ac

outperforms the one in IEEE 80211n An analytical model to evalu-

ate the performance of the IEEE 80211ac TXOP sharing mechanism in

DL-MU-MIMO communications is developed in [38] The main goal of

this study is to identify how the TXOP sharing mechanism could im-

prove the system efficiency while achieving channel access fairness

among the different ACs

How to optimally exploit the new DL-MU-MIMO capabilities pro-

vided by IEEE 80211ac is still an open challenge First due to the need

of frequent CSI exchanges between STAs and the AP it is not yet clear

in which conditions DL-MU-MIMO outperforms SU-MIMO [39ndash42]

or even whether MU-MIMO does or does not outperform multi-user

packet aggregation when the amount of data directed to each des-

tination is not balanced [43] Packet aggregation can be a solution

to balance the duration of the multi-user spatial streams as shown

in [36] although it will always depend on the amount of traffic di-

rected to each destination and the buffer capacity at the AP In [44]

the authors compare different strategies to assign the spatial streams

between the available destinations at each transmission in a fully

connected mesh network showing in ideal channel conditions the

theoretical benefits of MU-MIMO vs SU-MIMO

Closely related to the previous point a second open challenge is

the design of efficient schedulers that consider traffic priorities the

buffer state the different MIMO strategies TXOP sharing policies

grouping of STAs and the availability of fresh CSI feedbacks to max-

imise the throughput and guarantee the required QoS for each ac-

tive traffic flow It is important to consider that the availability of up-

dated CSI estimates from all STAs allows the AP to reduce the mutual

interference between the transmitted spatial streams which means

lower packet error probabilities and higher transmission rates How-

ever the overheads for obtaining the CSI from all STAs is large and

increases linearly with the channel sounding rate and the number

of STAs Proposals for reducing the CSI overhead are under develop-

ment For example in [40] the CSI overhead is reduced by inhibiting

the channel sounding whenever possible based on the estimation of

the channel stability for all users Another related problem is how to

roup the STAs as the goal is to find groups of STAs with compat-

ble (ie orthogonal) channels In [45] the authors show the chal-

enges inherent to the group assignment problem and they propose

n heuristic method to solve them TXOP sharing is considered in [46]

y presenting two alternative approaches to enhance the considered

ack-off procedure for the purpose of improving both throughput and

airness

A third key challenge for IEEE 80211ac is to achieve an efficient use

f the spectrum when several channel widths are used in scenarios

ith multiple overlapping WLANs Increasing the channel width the-

retically allows individual WLANs to achieve a higher throughput

owever the presence of other WLANs in the vicinity also increases

he chances of frequency overlapping which may cause the opposite

ffect as there appears inter-WLAN contention [27] Adaptive mecha-

isms to select the channel centre frequency and the channel width

nd MAC protocols to choose the instantaneous channel width used

or each transmission are thus required For instance in [47] the au-

hors focus on the channel selection problem when WLANs can use

ultiple channel widths using a game-theoretic framework In [48]

scheme is proposed to enable the communication between nodes

ith partially overlapping channels which may provide stronger re-

ilience to channel interferences

2 The IEEE 80211ax amendment

In 2014 the High Efficiency WLANs (HEW) Task Group [49] initi-

ted the development of a new IEEE 80211 amendment called IEEE

0211ax The IEEE 80211ax amendment is expected to be released in

019 and to some extent it will be the IEEE 80211 response to the

hallenges of future dense and high-bandwidth demanding WLAN

cenarios [5051]

The challenges in the development of the IEEE 80211ax amend-

ent are to

(i) Improve the WLANs performance by providing at least a four-

fold capacity increase compared to IEEE 80211ac

(ii) Provide support for dense networks considering both the ex-

istence of multiple overlapping WLANs and many STAs in each

of them Spatial reuse of the transmission resources is a must

(iii) Achieve an efficient use of the transmission resources by min-

imising the exchange of management and control packets re-

visiting the structure of the packets and improving channel

access and retransmission mechanisms among others aspects

(iv) Provide backward compatibility with previous amendments

This is achieved by the mandatory transmission of the legacy

PHY preamble in all frames and by keeping EDCA as the basic

channel access scheme

(v) Introduce effective energy saving mechanisms to minimise the

energy consumption

(vi) Support multi-user transmission strategies by further devel-

oping MU-MIMO and Orthogonal Frequency Division Multiple

Access (OFDMA) capabilities in both downlink and uplink

In addition to the aforementioned challenges next-generation

LANs will have to implement some other functionalities beyond the

aw packet transmission and reception Examples are a fast efficient

nd robust handoff between APs in the same administration domain

52] device-to-device communication (D2D) [53] and coordination

f multi-AP networks [54] In the first case the IEEE 80211ai amend-

ent called Fast Initial Link Setup is in progress and expected for

016 Its target is to complete a handoff in less than 100 ms including

ew AP discovery user authentication and configuration Using D2D

ommunication we can avoid the use of the AP as a relay hence im-

roving the overall efficiency as the number of packet transmissions

equired is reduced Finally the virtualisation of network functions

dds a new dimension in the management of multiple APs which in

B Bellalta et al Computer Communications 75 (2016) 1ndash25 7

d

e

I

w

f

t

s

3

P

s

G

t

8

p

C

c

m

c

i

t

a

s

w

f

t

t

e

l

p

e

a

H

d

t

c

o

s

p

n

t

b

a

w

c

t

w

n

8

a

c

w

d

H

n

a

h

p

w

i

t

fi

e

c

u

t

m

p

p

b

p

i

t

8

d

I

m

M

u

t

m

i

t

D

W

I

u

M

i

w

p

h

t

n

o

p

t

s

u

s

r

b

A

t

O

t

t

w

s

p

o

i

W

t

t

a

[

m

ense scenarios can contribute to notably improving the user experi-

nce We further discuss this last topic in Section 6

Different from the other amendments covered in this survey the

EEE 80211ax amendment is just in its initial stages of development

ith only very few technical aspects consolidated at this stage There-

ore in the following subsection we will overview both the new fea-

ures and open challenges of the IEEE 80211ax amendment at the

ame time

21 Novel features and open challenges

The IEEE 80111ax Task Group is currently working in four areas

HY MAC Multi-user and Spatial Reuse [49] Next we will overview

ome of the topics currently under discussion in the IEEE 80211 Task

roup in each category

PHY layer The IEEE 80211ax PHY layer will be an evolution of

he IEEE 80211ac one The challenges in the design of the IEEE

0211ax PHY layer are related with the extensions required to sup-

ort multi-user MU-MIMO and OFDMA transmissions and Dynamic

CA Also improvements in the supported modulation and channel

oding techniques will be likely considered to allow for higher trans-

ission rates at lower SNR values For example IEEE 80211ax may

onsider LDPC (Low-Density Parity Check) coding which is optional

n IEEE 80211ac instead of the traditional convolutional codes as

hey provide a coding gain of 1ndash2 dB Moreover the PHY layer may

lso include some functionalities to support the use of Hybrid ARQ

chemes to improve the efficiency of packet retransmissions

Medium acess control In order to keep backward compatibility

ith previous IEEE 80211 amendments besides a common PHY

rame preamble compatible MAC protocols are required This means

hat it is likely that EDCA will be kept as the main channel access

echnique in the IEEE 80211ax amendment Therefore the most rel-

vant open challenges are related to EDCA extensions to support a

arge number of STAs improve traffic differentiation capabilities im-

rove the energy consumption and provide mechanisms to fairly co-

xist with neighboring wireless networks

To support a large number of contenders with a low collision prob-

bility a simple solution is to use large backoff contention windows

owever it would also increase the time a STA is in backoff so re-

ucing the number of packets it can transmit A solution to reduce

he backoff duration when using large contention windows is to de-

rease the duration of the backoff slots Latest technological advances

n electronics should require less time to perform a CCA check to

witch between reception and transmission modes and to process a

acket thus enabling such a possibility

Another approach to enhance the underlying CSMACA mecha-

ism in EDCA and achieve a higher efficiency is to consider decen-

ralised collision-free MAC strategies Those MAC protocols are able to

uild collision-free schedules thus improving the network efficiency

s collisions are reduced while preserving backward compatibility

ith the default EDCA implementation An overview of several de-

entralised collision-free MAC protocols can be found in [55] Among

hem CSMAECA [56] is especially relevant since it is fully compatible

ith EDCA and latest upgrades made it adaptive to the instantaneous

umber of contenders in a single WLAN

In addition to EDCA IEEE 80211ax WLANs can rely on the IEEE

0211aa amendment to further improve the WLAN traffic differenti-

tion capabilities with intra-AC traffic differentiation and groupcast

ommunication mechanisms among other features In Section 33

e will provide further details about the IEEE 80211aa amendment

IEEE 80211ax will likely keep the same channel widths that were

efined in the IEEE 80211ac amendment ie 20 40 80 and 160 MHz

owever it is expected that IEEE 80211ax will extend current chan-

el bonding mechanisms to further improve the spectrum utilisation

nd the coexistence between neighbouring WLANs For example it

as been shown in [32] that the use of dynamic channel bonding

rovides significant throughput gains in dense scenarios compared

ith the static approach while minimising the inter-WLAN negative

nteractions [27] Furthermore additional mechanisms are required

o fully exploit the use of channel bonding such as the use of ef-

cient algorithms to select the position of the primary channel or

ven to consider the existence of multiple primary channels to in-

rease the number of bonded channel combinations that a node can

se for transmission

The MAC layer in IEEE 80211ax may work with the PHY layer

o implement an efficient Hybrid ARQ mechanism able to retrans-

it only incremental redundancy bits in short packets Opportunistic

iggy backing of data packets in ACKs and vice versa may further im-

rove the efficiency of IEEE 80211ax WLANs by reducing the num-

er of transmissions in a bidirectional data exchange [57] Finally

acket headers can be reduced if shorter STA identificators are used

nstead of MAC addresses and unnecessary fields are avoided in all

ransmissions

Finally in terms of power saving mechanisms it is expected IEEE

0211ax will follow the same trends as in previous IEEE 80211 stan-

ards including the extensions currently under development for the

EEE 80211ah amendment that are discussed in Section 4

Multi-user Multi-user communications will likely be one of the

ain characteristics of IEEE 80211ax as both uplink and downlink

U-MUMO and OFDMA are under consideration The use of multi-

ser communication techniques does not necessarily represent a sys-

em capacity increase because the available transmission resources

ay be the same as in the single-user communication case However

n WLANs the simultaneous transmission from different users is able

o parallelise the large temporal overheads of each transmission (ie

IFS SIFS ACKs packet headers etc) which can notably improve the

LAN efficiency

IEEE 80211ax will further develop the MU-MIMO capabilities of

EEE 80211ac by allowing multiple simultaneous transmissions in the

plink which is known as uplink (UL-) MU-MIMO [58] Similar to DL-

U-MIMO transmissions an open challenge to enable UL-MU-MIMO

s to design a mechanism able to efficiently schedule the users that

ill transmit at the same time In one hand a pure decentralised ap-

roach would be easy to implement with minimal signalling over-

eads However since it requires that all STAs finish their backoff at

he same time it may be very inefficient besides that those STAs may

ot be compatible in terms of their respective spatial channels In the

ther hand a pure centralised approach requires that the AP has com-

lete CSI and buffer occupancy information from all STAs to select

he most suitable group to perform a multi-user transmission Once a

uitable group of STAs is selected by the AP a ldquoTriggerrdquo frame may be

sed to notify the group of selected users that can initiate a transmis-

ion This approach guarantees efficient multi-user transmissions but

equires some extra overheads to collect all the required information

y the AP and signal the selected STAs In both cases new multi-user

CKs will be likely introduced by IEEE 80211ax to acknowledge all

ransmissions with a single control packet

Multi-user OFDMA is also in the agenda for IEEE 80211ax Using

FDMA a channel can be split in several sub-channels and assigned

o different users Likely OFDMA will be implemented in combina-

ion with channel bonding where each of the 20 MHz subchannels

ill be assigned to a different user in both downlink and uplink Be-

ides that a similar operation as in the multi-user MIMO case is ex-

ected as there are almost the same challenges to solve A survey

f current OFDMA proposals for WLANs is presented in [59] show-

ng also how the use of OFDMA is able to significantly improve the

LAN efficiency Fig 4 illustrates an uplink OFDMA and MU-MIMO

ransmission

In addition to Multi-user MIMO and OFDMA the use of Simul-

aneous Transmit and Receive (STR) techniques commonly known

s full-duplex transmission have been suggested for IEEE 80211ax

5051] Using STR a pair of nodes is able to transmit and receive si-

ultaneously [60] which theoretically doubles the channel capacity

8 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 4 Multiuser uplink transmission using two different OFDM subchannels Trans-

missions from STA B and C over the same subchannel represent an uplink MU-MIMO

transmission All transmissions are acknowledged with a single Multiuser ACK The

trigger frame is used to signal the selected STAs

t

t

t

a

s

t

f

f

c

3

s

s

t

n

n

A

d

w

t

fi

a

t

8

i

d

t

r

w

d

R

v

f

s

d

p

B

c

e

k

r

G

a

c

m

t

b

d

3 The concealment address protects legacy stations ie GCR-incapable stations

from receiving duplicated group-addressed frames

The challenge is that both the AP and the STA involved in a full-duplex

transmission have to start to transmit at the same time To achieve

that information about full duplex transmission capabilities can be

included in RTSCTS control packets to set-up a full-duplex transmis-

sion between the transmission initiator and its destination

Spatial reuse Dense WLAN deployments are necessary to offer a

continuous coverage with high transmission rates To improve both

the co-existence with those neighboring networks and the spatial

reuse of the spectrum a WLAN has two options (i) minimise its

area of influence by reducing its transmit power and (ii) accept

higher interference levels by increasing the Clear Channel Assess-

ment (CCA) level Use of both techniques may increase the number of

concurrent transmissions between neighbouring WLANs and there-

fore their capacity although it may also result in the opposite effect

since the achievable transmission rates may be negatively affected by

the higher interference levels observed which is the main challenge

to solve

Due the high WLAN dynamics the use of adaptive systems is cru-

cial but challenging as adaptivity requires extra complexity in terms

of computing and memory resources and there are no guarantees

that the implemented solution converges due to the decentralised

operation of each WLAN The use of DSC (Dynamic Sensitivity Con-

trol) to dynamically adjust the CCA level is one of the aspects cur-

rently under discussion in the IEEE 80211ax Task Group First studies

evaluating the performance of DSC for IEEE 80211ax WLANs show

a clear improvement on the spatial reuse and the area throughput

[61] Another example of the achievable throughput gains obtained

by adapting the CCA level can be found in [62] where the authors

show that gains of up to 100 can be achieved Moreover trans-

mit Power Control (TPC) to mitigate interference between WLANs in

dense scenarios is studied in [63] showing the need of jointly opti-

mising both TPC and CCA to maximise the network performance

Finally sectorisation by using beamforming is also under consid-

eration for the development of the IEEE 80211ax amendment as a

potential solution to improve spatial reuse [64] Using sectorisation

only the nodes of a given area are allowed to receive or transmit data

hence reducing the contention between different networks whenever

they activate non-overlapping sectors A challenge here is to coor-

dinate the different neighboring APs when they belong to different

administration domains Decentralised learning approaches may be

implemented to find feasible temporal patterns of non-overlapping

sectors

33 The IEEE 80211aa amendment

As discussed above legacy IEEE 80211 standards do not provide

robust and efficient delivery of audiovideo streaming services Thus

he IEEE 80211aa amendment was developed to include new fea-

ures and additional mechanisms to improve the performance of real-

ime multi-media content delivery [65] Specifically IEEE 80211aa

ddresses the following five shortcomings of previous 80211

tandards [1666]

(i) the lack of reliable and efficient support for multicast and

group communications

(ii) the incapacity of applying traffic prioritisation to different

multimedia streams or different types of frames from the same

stream

(iii) the absence of methods for cooperative resource sharing

among neighbouring APs

(iv) the lack of mechanisms for graceful degradation of audiovideo

streaming quality

(v) the non-interoperability with existing IEEE 8021 standards for

Audio Video Bridging (AVB)

In the following sections we present in detail the solutions to

hose problems introduced in the IEEE 80211aa amendment We

urther discuss the research studies that have provided the basis

or the IEEE 80211aa design and we identify the remaining open

hallenges

31 Novel features

Groupcast communication mechanisms In most audiovideo

treaming applications a group of clients must receive the same

tream simultaneously A multicast protocol is necessary to avoid that

he same content is replicated throughout the network In wireless

etworks multicast transmission can exploit the intrinsic broadcast

ature of the wireless channel ie broadcast transmissions from an

P are physically received by all other stations in the same collision

omain However multicast and broadcast frames in IEEE 80211 net-

orks are not protected by an acknowledgement mechanism as in

he case of unicast frames Thus layer-2 multicast transmissions de-

ned by legacy IEEE 80211 standards are unreliable and not suit-

ble for streaming applications To partially address this limitation

he Direct Multicast Service (DMS) was first specified in the IEEE

0211v amendment [67] Basically DMS converts multicast streams

nto unicast streams In this way frames destined to a multicast ad-

ress are individually transmitted as unicast frames to the stations

hat joined that multicast group Obviously DMS provides the same

eliability as unicast transmission services but the consumed band-

idth increases linearly with the number of group members To ad-

ress this scalability issue IEEE 80211aa includes the Groupcast with

etries (GCR) service in addition to DMS Specifically the GCR ser-

ice defines new mechanisms and the related management frames

or group formation which allows a set of stations to agree on a

hared (non-multicast) address called the groupcast concealment ad-

ress3 Furthermore the GCR service specifies two retransmission

olicies GCR Unsolicited Retry (GCR-UR) and GCR Block Ack (GCR-

A) When using GCR-UR the AP can proactively retransmit all group-

ast frames a number of times to mitigate the impact of channel

rrors (see Fig 5a)) while receivers are not requested to send ac-

nowledgements Intuitively this approach improves transmission

eliability but it still suffers from scalability issues In contrast when

CR-BA is used the AP sends a burst of consecutive groupcast frames

nd it requests the receivers to reply with a Block ACK frame which

ontains a bitmap to positively or negatively acknowledge trans-

itted frames (see Fig 5b)) The Block ACK mechanism defined for

he GCR-BA service is quite flexible because Block ACK frames can

e requested immediately after a transmission burst or after a ran-

omised back-off delay Furthermore the AP can request the Block

B Bellalta et al Computer Communications 75 (2016) 1ndash25 9

Fig 5 GCR service with different retransmission schemes

A

r

o

l

o

e

v

e

v

d

f

d

s

v

c

s

a

a

f

d

t

F

t

d

v

a

f

c

t

t

o

t

i

c

r

v

i

r

s

o

t

a

(

p

i

f

a

o

i

t

t

c

s

fi

o

c

I

n

i

a

t

a

t

f

i

B

T

f

t

a

g

a

a

i

s

a

d

s

p

m

m

i

a

t

b

t

i

(

r

a

S

i

c

s

c

T

Q

p

a

3

g

t

4 TSPEC is a traffic specification sent from a QoS capable wireless client that requests

a certain amount of network traffic from the AP for the traffic stream it represents

CK frame to all groupcast recipients or only to a subset of them to

educe overheads and delays The advantages of the GCR methods

ver broadcast and DMS have been extensively demonstrated in the

iterature [1668]

Intra-access category prioritisation The IEEE 80211e amendment

nly allows traffic differentiation between four different access cat-

gories (ACs) that are broadly mapped to four application classes

oice (VO) video (VD) best-effort (BE) and background (BK) How-

ver there is a variety of streaming services ranging from simple

ideoconferencing to HD streaming over IPTV systems which have

ifferent QoS requirements (see Table 3) To provide the ability to dif-

erentiate among individual streams IEEE 80211aa includes an ad-

itional scheduling layer with respect to IEEE 80211e IEEE 80211aa

plits each one of the transmission queues associated with voice and

ideo ACs into a primary and an alternate queue In this way spe-

ialised scheduling rules can be applied to decide which queue to

erve when the EDCA function for inter-AC collision resolution grants

n access opportunity to voice or video ACs To facilitate the man-

gement of service level agreements IEEE 80211aa follows the de-

ault mappings between user priority values and traffic types that are

efined in the IEEE 8021D standard [69] It is then straightforward

o further map traffic types onto transmission queues and ACs (see

ig 6) Finally it is important to point out that the intra-AC differen-

iation functionality can be used to provide more sophisticated traffic

ifferentiation than simple stream prioritisation For instance most

ideo applications use Scalable Video Coding (SVC) schemes that en-

ble the partitioning of a video sequence into multiple layers with dif-

erent qualities and rates [70] Typically an SVC-based video stream

ontains a base layer which provides a basic level of quality and mul-

iple enhancement layers which can only be decoded together with

he base layer to improve the video quality Thus the different layers

f the same encoded video steam can be easily mapped to different

ransmission queues to receive differentiated QoS [71]

Stream classification service The stream classification service (SCS)

s an optional service that can be provided by an AP to the asso-

iated stations to classify multimedia streams based on arbitrary

ules that are established directly by the stations instead of the con-

entional 8021D user priorities To this end the station request-

ng the use of SCS must initiate an SCS session by sending an SCS

equest frame to the AP which contains an identifier for the SCS

tream and the descriptor of the classification rule The AP may accept

r reject the requirements specified by the station Once accepted

he AP must assign all frames that match the classification rule to

specific AC When intra-access category prioritisation is enabled

see Section 331) calternate transmit queues so that finer grained

rioritisation can be applied Finally there is also a Drop Eligibil-

ty Indicator (DEI) bit in the SCS descriptor that indicates whether

rames from this traffic stream can be dropped in the case that there

re insufficient resources Specifically frames with the DEI bit set to

ne have a higher probability of being discarded because their max-

mum number of allowed retries is smaller than the default Note

hat how to combine intra-AC queues and frame dropping settings

o achieve graceful degradation of the audiovideo stream quality in

ase of bandwidth shortage is beyond the scope of the IEEE 80211aa

pecification

Overlapping Basic Service Set (OBSS) management Network densi-

cation ie a denser deployment of wireless infrastructure nodes is

ne of the key strategies that is used nowadays to easily increase the

apacity of wireless systems even for indoor WLANs [72] However

EEE 80211 networks have a limited number of orthogonal chan-

els available and even if optimised frequency planning is applied

t might happen that neighbouring APs are mutually interfering and

station may affect multiple overlapping BSSs In this case conges-

ion not only increases but it is also likely to observe an unfair us-

ge of wireless capacity with the channel retained by one AP for long

ime intervals This is mainly due to the neighbourhood capture ef-

ect ie hidden terminal phenomena among APs [73] To address this

ssue IEEE 80211aa specifies a new functionality called Overlapping

SS (OBSS) management which is based on two new mechanisms

he first defines a set of parameters to quantify the load and inter-

erence among neighbouring BSSs such as medium occupancy frac-

ion number of admitted audiovideo streams data traffic volumes

nd the number of BSSs that are using the same channel as the tar-

et one Note that the traffic load consists of two components the

llocated traffic which is derived on the basis of the TSPEC values of

dmitted streams4 and predicted traffic which is evaluated by track-

ng the maximum value of the allocated EDCA and HCCA traffic over

even-day periods Once load measurement reports are exchanged

mong the APs a second OBSS component is responsible for coor-

inated admission control procedures on the basis of two suggested

haring schemes proportional sharing and on-demand sharing The

urpose of both schemes is to keep the total allocated traffic below a

aximum value in order to provide some QoS protection to admitted

ultimedia streams Finally IEEE 80211aa recommends implement-

ng additional OBSS management procedures for channel selection

nd cooperatively creating HCCA schedules that do not collide

Interworking with IEEE 8021AVB Audio Video Bridging (AVB) is a

erm commonly used to denote a set of technical standards developed

y IEEE to support real-time streaming services with bounded latency

hrough IEEE 802 networks [74] This objective is achieved by spec-

fying mechanisms to allow the synchronisation of multiple streams

IEEE 8021AS [75]) and traffic shaping (IEEE 8021Qav [76]) and to

eserve network resources for specific audiovideo streams traversing

bridged local area network by using a signalling protocol called the

tream Reservation Protocol (SRP) (IEEE 8021Qat [77]) IEEE 80211aa

ntegrates the SRP operations with the EDCA admission control pro-

edures Specifically the SRP RequestResponse messages are encap-

ulated in the management frames that are used to carry the traffic

haracteristics and the QoS requirements during admission control

his enables the end-to-end management of resource reservation for

oS guaranteed streams even when one or more IEEE 80211 links are

art of a path from the stream producers (called IEEE 8021Q talkers)

nd the stream consumers (called IEEE 8021Q listeners)

32 Open challenges

In recent years several MAC enhancements have been investi-

ated to improve QoS guarantees for real-time multimedia applica-

ions in IEEE 80211 networks [20] and the IEEE 80211aa standard

10 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 6 Stream classification and inter-AC traffic prioritisation

a

l

t

r

n

a

t

n

i

t

i

t

r

d

p

t

t

e

m

b

t

a

i

s

o

s

a

e

t

i

f

e

m

8

t

I

i

fi

which was finalised in 2012 included several of these proposed im-

provements Significant research efforts have focused on improving

the transmission reliability of multicasting by integrating ARQ mech-

anisms in IEEE 80211-based multicast transmissions Modifications

to the legacy MAC protocol were proposed in [78] to enable the

RTSCTS option in multicast mode and to select one or more multi-

cast receivers (called leaders) for acknowledging multicast data pack-

ets However these enhancements require changes to the standard

specifications The main problems of leader-based ARQ schemes are

leader election and the trade-off between scalability and reliability

The authors in [79] propose selecting the multicast recipient operat-

ing in the worst channel conditions as the unique leader but this ap-

proach may perform poorly in lossy environments In the Batch mode

multicast MAC (BMMM) [80] all multicast recipients are polled by

the multicast originator to send individual ACKs but this scheme is

not suitable for large multicast groups The Enhanced Leader Based

Protocol (ELBP) is proposed in [81] on the basis of multiple ACK-

leaders and block acknowledgement techniques Analytical models

are then developed to help select optimal ACK-leaders to meet ap-

plication QoS requirements However the models apply only to sat-

urated traffic while multimedia streams are typically bursty Another

class of reliable multicast protocols relies on busy tones to reduce

packet losses due to collisions [82] but the additional radio inter-

face needed for the busy tone limits the practicality of such solu-

tions An alternative approach to avoid collisions of multicast pack-

ets is the multicast collision prevention (MCP) scheme [83] which

is based on the use of a shorter waiting time for transmitting mul-

ticast packets An interesting approach is also proposed in [84] to

retransmit lost packets using an online linear XOR coding algorithm

However a modification to the standard MAC protocol is required to

enable simultaneous ACK transmissions In summary several differ-

ent methods have been proposed to improve multicast transmission

reliability by integrating ARQ schemes into the protocol architecture

but there are not conclusive results on which is the best solution The

choice of the most efficient mechanism depends on a variety of in-

terdependent factors such as loss ratios channel congestion multi-

cast group size and QoS requirements of multimedia streams A com-

prehensive analytical framework is needed to optimise the setting of

the parameters for each scheme and to dynamically select the best

one

As discussed above one main difference between unicast services

nd multicast services in the legacy IEEE 80211 standard was the

ack of acknowledgements Another critical difference is that mul-

icast frames must be transmitted using a fixed rate in the basic

ate set while the transmission rate of unicast frames can be dy-

amically adapted to the channel and traffic conditions [85] Thus

group of research papers has investigated the use of rate adap-

ation to improve the throughput of multicast services in IEEE 80211

etworks [7186ndash89] For instance the authors in [86] propose us-

ng RTS frames to allow group members to estimate channel condi-

ions Each member will then send a dummy CTS frame with a length

nversely proportional to channel quality In this way the multicast

ransmitter can use the collision duration to predict the lowest data

ate that can be used for group transmissions The overhead intro-

uced by this mechanism is quite high however The solution pro-

osed in [88] called ARSM also relies on feedback messages sent by

he multicast receivers called multicast response frames to identify

he group member exhibiting the poorest channel conditions How-

ver in this case a different back off timer is associated with each

ulticast receiver depending on the SNR of previously received feed-

ack messages in order to prevent collision An approach similar to

he one employed in the Auto Rate Fallback (ARF) protocol a rate

daptation scheme originally proposed in [90] is used in [87] Specif-

cally the number of successful consecutive transmissions and con-

ecutive transmission failures are used to decide when to increase

r decrease the transmission data rate respectively A modified ARF

cheme is also proposed in [71] which can be applied to videos that

re encoded into two layers namely the base and enhancement lay-

rs However how to integrate rate adaptation with the different re-

ransmission policies that are defined in IEEE 80211aa is still an open

ssue

One research area that is expected to be crucial in the success-

ul development of IEEE 80211aa-based products is the design of

fficient scheduling algorithms for supporting voicevideo traffic Al-

ost all research work in this field has been triggered by the IEEE

0211e amendment that enhanced the original IEEE 80211 MAC with

wo new QoS-aware access mechanisms ie EDCA and HCCA [91]

n principle with a well-designed admission control and schedul-

ng scheme HCCA is able to provide hard QoS guarantees to traf-

c flows [9293] However HCCA is rarely implemented in IEEE

B Bellalta et al Computer Communications 75 (2016) 1ndash25 11

8

c

c

a

s

b

w

o

j

t

f

a

Q

t

t

A

p

t

s

a

f

s

H

a

i

c

i

t

s

f

s

s

m

i

t

s

l

t

t

[

t

i

o

b

r

W

v

m

8

s

i

o

p

q

4

c

e

t

d

a

t

m

Fig 7 WLANs for M2M communications STAs represent sensor and actuator devices

4

t

w

s

a

c

g

e

(

p

o

t

p

9

r

o

a

p

c

o

i

p

b

f

t

s

o

i

4

8

f

t

i

s

m

d

0211e-based WLANs owing to its higher complexity and cost con-

erns Instead EDCA is widely adopted Most papers have thus fo-

used on improving EDCA performance Many papers have proposed

nalytical models for various subsets of EDCA functionalities For in-

tance a saturation-based performance analysis is conducted in [94]

y differentiating the minimum back-off window size the back-off

indow-increasing factor and the retransmission limit The authors

f [9596] also model AIFS differentiation while the model in [97]

ointly captures all the four EDCA parameters for traffic differen-

iation More recent papers have analysed the EDCA performance

or non-saturated conditions and for arbitrary buffer sizes [98] The

uthors in [99] have developed an analytical model to predict the

oS levels that can be achieved once a new voicevideo flow is in-

roduced in the WLAN A Kalman filter is proposed in [100] to ob-

ain estimates on the number of active transmission queues of each

ccess Category in EDCA These analytical models can then be ex-

loited to derive the optimal configuration of the EDCA parameters

o achieve given performance criteria or to design admission control

chemes that preserve QoS constraints For instance a scheme that

ssigns contention-window values to achieve pre-defined weighted-

airness goals is proposed in [101] A control-theoretic scheme is de-

igned in [102] with the goal of minimising the video traffic delay

owever most of these solutions rely on non-realistic assumptions

bout video traffic dynamics An alternative class of solutions dynam-

cally updates the EDCA parameters based on the observed network

onditions In [103] the EDCA parameters are optimised consider-

ng a WLAN with rigid and elastic traffic simultaneously analysing

he interactions between both types of traffic The authors in [104]

pecify several bandwidth-sharing mechanisms with guaranteed QoS

or voice and video traffic Measurement-based admission control

chemes are proposed in [105] A TXOP adaptation method is de-

cribed in [106] that takes into account video frame sizes and trans-

it queue lengths However the main drawback of these solutions

s that they are based on heuristics and hence do not ensure op-

imal and guaranteed performance Finally a third category of re-

earch papers tries to improve video performance by designing cross-

ayer scheduling approaches Specifically these works take advan-

age of multi-layer video encoding to classify the frames according

o their importance and assign them to different access categories

107] For instance the authors in [108] define classifiers and waiting

ime priority schedulers that dynamically change the packet prior-

ties according to end-to-end delay measurements A disadvantage

f this approach however is that an additional adaptation layer may

e needed to implement the complex interactions that are typically

equired between the video coding applications and the MAC layer

e conclude this section by pointing out that existing studies pro-

ide the basic design principles and techniques for improving multi-

edia streaming performance in IEEE 80211 networks Still the IEEE

0211aa standard poses new research challenges that have not been

ufficiently explored and that will require innovative solutions For

nstance scheduling between primary and alternate queues is still an

pen research area as the mapping of individual frames to multi-

le queues in order to achieve graceful degradation of voicevideo

uality [16]

Sensor Networks and machine-type communications

As discussed in Section 2 M2M communications refer to any

ommunication technology that enables sensoractuator devices to

xchange information and perform actions without the manual assis-

ance of humans This section reviews the main features currently un-

er consideration in the development of the upcoming IEEE 80211ah

mendment which targets the main challenges of those networks as

he IoT in general such as the energy consumption or the manage-

ent of many devices

1 The IEEE 80211ah amendment

The IEEE 80211ah amendment [109] aims to provide WLANs with

he ability to both manage a large number of heterogeneous STAs

ithin a single BSS and minimise the energy consumption of the

ensor-type battery-powered STAs

The initial design requirements of the IEEE 80211ah amendment

re detailed in [110] these entail the support of up to 8192 STAs asso-

iated with a single AP the adoption of efficient power saving strate-

ies a minimum data rate of 100 kbps the operation in the license-

xempt sub 1 GHz band and a coverage up to 1 km in outdoor areas

see Fig 7 for an illustrative example) A preliminary assessment of

erformance of the IEEE 80211ah technology in terms of the number

f STAs that can be effectively supported in a single WLAN as well as

heir energy consumption is presented in [111]

IEEE 80211ah operates over different sub-1 GHz ISM bands de-

ending on country regulations 863ndash868 MHz in Europe 902ndash

28 MHz in the US and 9165ndash9275 MHz in Japan China South Ko-

ea and Singapore also have specific channelisations Channel widths

f 1 MHz and 2 MHz have been adopted although 4 8 and 16 MHz

re also supported in some countries IEEE 80211ah furthermore pro-

oses new PHY and MAC layers The IEEE 80211ah PHY layer can be

onsidered to some extent a sub-1 GHz version of the IEEE 80211ac

ne At the physical layer OFDM is the chosen modulation method us-

ng 32 or 64 tonessub-carriers that are spaced by 3125 kHz The sup-

orted modulations include BPSK QPSK and from 16 to 256-QAM A

road range of antenna technologies ranging from single-user beam-

orming to MIMO and DL-MU-MIMO which was first introduced in

he IEEE 80211ac amendment are also included in the IEEE 80211ah

pecification Similarly the IEEE 80211ah MAC protocol include most

f IEEE 80211 main characteristics further extending its power sav-

ng (PS) mechanisms

11 Novel features

This section overviews the extensions introduced by the IEEE

0211ah amendment to the IEEE 80211 PS mechanisms to account

or the specific characteristics of resource-constrained sensor and ac-

uator devices A more detailed review can be found in [110] includ-

ng a performance assessment of IEEE 80211ah in several of the key

cenarios for M2M communications such as agriculture and animal

onitoring smart metering and industrial automation plants In ad-

ition a detailed survey of the IEEE 80211ah is reported in [112]

12 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 8 IEEE 80211ah PS mode for TIM-STAs non-TIM STAs and Unscheduled STAs

a

d

w

i

a

a

c

m

i

a

e

t

s

m

t

t

u

b

i

I

s

d

f

a

b

s

a

b

S

T

c

t

w

I

4

a

a

which addresses aspects not considered in [110] such as the use of

sectorisation to avoid overlapping between multiple IEEE 80211ah

WLANs and the use of relays to further extend the IEEE 80211ah cov-

erage among other aspects Finally in [113] both IEEE 80211ah PHY

and MAC characteristics are introduced with emphasis on the ben-

efits of using the sub-1 GHz ISM band in terms of link budget the

location of the OFDM pilots for outdoors operation the use of bidi-

rectional transmission opportunities and the new packet formats to

minimise overheads among others

Enhanced power saving mechanism PS mechanisms for WLANs

were already considered in the development of the first IEEE 80211

standard with the goal of improving the lifetime of battery equipped

devices [114] In PS mode STAs keep the transceiver in sleeping mode

as much time as possible They periodically wake up to listen to the

beacons transmitted by the AP Those beacons indicate whether an

STA has packets waiting for it at the AP In the positive case that STA

remains awake and requests the delivery of those packets Otherwise

given it has nothing to receive it returns to sleep mode until the next

beacon is expected

In the IEEE 80211ah amendment time is divided into pages DTIM

(Delivery Traffic Indication Map) periods TIM (Traffic Indication Map)

periods and slots DTIM and TIM periods begin with the correspond-

ing DTIM and TIM beacons sent by the AP The functions of DTIM and

TIM beacons are described below

1 DTIM beacons They inform as to which TIM Groups (ie the

group of STAs assigned to the same TIM period) have pending

packets at the AP

2 TIM beacons Each TIM message informs a TIM Group about

which specific STA has pending data in the AP Between two

consecutive DTIMs there are as many TIM beacons as TIM

Groups

Using this DTIMTIM-based approach any STA can enter into a

power saving state if it does not have packets pending for transmis-

sion and one of two conditions is met (1) it observes in the DTIM

beacon that there is no downlink traffic addressed to its TIM Group or

(2) it observes in the DTIM beacon that there is some downlink traffic

addressed to its TIM Group but that STA does not explicitly appear in

the TIM beacon Compared to the preliminary IEEE 80211 PS mecha-

nism this approach reduces the size of the Traffic Indication Map in

each TIM beacon thus reducing the overhead and the time STAs need

to listen and process them In addition TIM periods can be organised

in pages which further increases the number of TIM groups between

two DTIM beacons

The temporal organisation of pages DTIM and TIM periods is re-

ported in Fig 8 which shows the division of TIM periods in RAW (Re-

stricted Access Window) and PRAW (Periodic RAW) The RAW is a

time interval in each TIM period where TIM stations can transmit and

receive data (see below the different types of STAs defined in IEEE

80211ah) In addition it can be divided into several downlink and

uplink slots for further granularity In the downlink the slots are as-

signed to a single STA or a group of STAs while the uplink slots are

randomly selected by the STAs with packets ready for transmission

The PRAW is the period of time in each TIM where non-TIM stations

can transmit and receive data

Types of STAs IEEE 80211ah supports three types of STAs TIM

non-TIM and Unscheduled STAs A TIM station is assigned to a TIM

Group Their data transmissions must be performed within a RAW

(Restricted Access Window) period Non-TIM stations do not have to

listen to beacons to transmit data During the association process

non-TIM devices directly negotiate with the AP to obtain a trans-

mission time allocated in a PRAW The following channel access can

either be renegotiated or occurs periodically depending on the re-

quirements set by the STA Unscheduled STAs do not need to listen to

any beacons similar to non-TIM stations Even inside any restricted

access window unscheduled STAs can send a poll frame to the AP

sking for immediate access to the channel The response frame in-

icates an interval (outside both restricted access windows) during

hich unscheduled stations can access the channel This procedure

s meant for STAs that transmit data very sporadically

Hierarchical station organisation To support a large number of STAs

nd their organisation in pages DTIM and TIM periods IEEE 80211ah

ssigns to each associated STA a unique identifier of 13 bits which is

alled the Association Identifier (AID) Using this new AID the maxi-

um number of supported STAs is increased from the original 2007

n IEEE 80211 to 8191 (= 213 minus 1) in IEEE 80211ah However it also

llows categorising STAs according to the type of application they are

xecuting their power level or even their desired QoS by assigning

hem to different TIM groups

Long sleeping periods IEEE 80211ah offers TIM Non-TIM and Un-

cheduled STAs the possibility to set very long doze times (up to

onths) The corresponding clock drift produced by such long doze

imes must be taken into consideration however as the higher the

ime an STA has been asleep the further in advance it should wake

p to avoid possible synchronisation problems with the network

Efficient small data transmission Three new enhancements have

een proposed to reduce the overhead when the data packet size

s small First while IEEE 80211 contains a 28-byte MAC header

EEE 80211ah proposes a short 18-byte version by using AIDs in-

tead of MAC addresses Second IEEE 80211 has defined several null

ata packet (NDP) frames which consist only of a PHY header These

rames can be used to create short ACKs short Block ACKs short CTSs

nd short PS-Polls Finally a Fast Frame Exchange mechanism has

een developed so that if an STA has data to transmit it can notify a

uccessful reception by transmitting its data frame instead of an ACK

Sectorisation Since the PHY layer is based on the IEEE 80211ac

mendment single and multi-user beam-forming are also supported

y IEEE 80211ah This allows the transmission of data to multiple

TAs simultaneously in the downlink increasing the system capacity

he use of the beam-forming capability of IEEE 80211ah APs is also

onsidered as a means to group STAs into different independent an-

enna sectors with the main goal of reducing interference issues This

ould be particularly useful in the case of overlapping with other

EEE 80211ah WLANs or in the presence of hidden nodes

12 Open challenges

A first open challenge is to understand the coverage and achiev-

ble transmission rates in IEEE 80211ah WLANs in both indoor

nd outdoor scenarios Propagation models for WLANs working at

B Bellalta et al Computer Communications 75 (2016) 1ndash25 13

f

c

8

p

R

p

i

m

l

a

m

c

p

p

f

m

t

a

e

p

t

m

a

l

T

m

a

i

n

t

t

t

u

p

c

w

f

a

t

8

a

m

a

s

l

a

m

b

m

d

S

i

a

o

a

R

t

n

G

fi

s

t

p

t

c

u

c

w

i

o

p

g

a

c

i

f

c

(

d

m

l

i

R

t

l

t

L

m

a

m

o

e

f

a

I

i

t

w

i

t

a

c

w

p

f

t

d

5

i

s

e

s

w

5

a

c

a

n

requencies lower than 1 GHz are evaluated in [115] The authors

ompare two path loss propagation models proposed by the IEEE

0211ah Task Group (one for macro and one for picohotzone de-

loyments) [116] with Lee and HatandashOkumura propagation models

esults show that the IEEE 80211ah channel models underestimate

ath loss with respect to Lee and Hata models Moreover in a compar-

son with empirical data it is observed that the IEEE 80211ah channel

odels also underestimate the initial loss and the slope of the path-

oss function A new model parameterisation is thus proposed by the

uthors In [117] the feasibility of an IEEE 80211ah WLAN deploy-

ent is also evaluated in terms of the achievable range and bit rate

omputed on the basis of the link budget using the same path loss

ropagation models as in [115] Results show that the transmission

ower limitations in the uplink can limit the overall network per-

ormance Finally in [118] the authors evaluate the achievable trans-

ission range for the different transmission rates and the achievable

hroughput for different combination of transmission power values

nd transmission rates Further studies are required due to the het-

rogeneity of scenarios in which IEEE 80211ah WLANs can be de-

loyed to characterise the effective coverage ranges with special at-

ention to mixed indoors and outdoors scenarios and in presence of

obile nodes

Due to the wide coverage IEEE 80211ah WLANs may be severely

ffected by the presence of hidden terminals To mitigate such a prob-

em smart solutions to guarantee that the STAs assigned to the same

IM Group are inside each other transmission range or the employ-

ent of RTSCTS-based solutions may contribute to mitigate such

problem An overview of long-range scenarios for IEEE 80211ah

s presented in [119] including a detailed description of the hidden

ode problem in them

Single-hop uplink transmissions from distant STAs require high

ransmission power to reach the AP Since not all STAs may be able

o transmit at the required power a solution could be the introduc-

ion of nodes able to relay transmissions from them However the

se of relays has to be efficiently harmonised with the operation in

ower saving mode allowing temporal periods in which the relays

an gather the data from their associated STAs and then periods in

hich they forward the data to the AP which is still an open challenge

or IEEE 80211ah WLANs In case relays are used data prediction and

ggregation techniques could be implemented to make more efficient

ransmissions [120]

Another challenge is the assessment of the efficiency of IEEE

0211ah PS mechanisms including their optimisation In [111] the

uthors evaluate the capacity of the TIM and page segmentation

echanisms in terms of the maximum number of nodes supported

nd the energy consumed given a certain network traffic profile Re-

ults confirm that a large number of STAs can be supported with

ow energy consumption The impact of the number of nodes oper-

ting in PS mode on the energy consumption and the delay perfor-

ance is analysed in [121] A detailed analysis of the TIM Group-

ased channel access adopted by the IEEE 80211ah task group is

ade in [122] where the STAs assigned to each TIM are uniformly

istributed between the slots of the RAW Since only one group of

TAs is allowed to transmit in each RAW slot the channel contention

s minimised The number of downlink and uplink slots in a RAW

s well as their duration is optimised on the basis of the number

f active STAs and the network traffic profile in [123] In [124] the

uthors propose an algorithm to determine the optimal size of the

AW interval for uplink transmissions based on the estimation of

he number of STAs transmitting and the duration of the RAW Fi-

ally in [125] the authors address the contention problems in Smart

rid communication networks with many nodes and periodic traf-

c when using an IEEE 80211ah WLAN It is still a challenge to con-

ider smart systems able to adapt the IEEE 80211ah parameters to

he instantaneous system state in order to save as much energy as

ossible

Since the PS mechanisms affect only the operation of TIM STAs

he performance of non-TIM and Unscheduled STAs has not yet been

onsidered in the literature To identify the scenarios in which the

se of non-TIM and Unscheduled STAs are of interest is still an open

hallenge as further investigating how non-TIM STAs negotiate over

hich PRAWs they can use to transmit and receive data Co-existence

ssues between the three types of STAs in a single network is also an

pen challenge that requires further work to guarantee a minimum

erformance level for all of them

The design of strategies to distribute the STAs between all the TIM

roups based on their specific traffic profile position battery level

nd application priority is another important challenge that is still

ompletely open EDCA is the default channel access scheme included

n the IEEE 80211ah amendment and provides some basic traffic dif-

erentiation capabilities at the packet level Besides the different ac-

ess categories (ACs) needing to be renamed and their parameters

Arbitration Inter-Frame Spacing or AIFS CWmin TXOP duration) up-

ated to fit the specific traffic profiles of M2M communications other

echanisms can be implemented A mechanism to assign the down-

ink RAW slots when packets from multiple priority levels are wait-

ng for transmission at the AP is required for example Also uplink

AW slots are currently selected randomly by the STAs that want to

ransmit a packet A mechanism to reserve some slots at each up-

ink RAW for high priority STAs may therefore be an option although

his may severely degrade the performance of the low priority STAs

astly different repetition patterns for different priority TIM groups

ay also allow priority STAs assigned to those priority TIM groups to

ccess the channel more often We expect this challenge will receive

uch attention in the upcoming years because of the heterogeneity

f sensors that will be connected though a single IEEE 80211ah AP

specially in urban scenarios

Similar to what has been discussed for 4G cellular networks used

or M2M communications [126] the development of mechanisms to

void and resolve congestion situations is another open challenge for

EEE 80211ah WLANs with many nodes For instance a mechanism

n which STAs have to wait a random number of DTIM periods before

hey can start a transmission may be implemented This approach

ould effectively distribute the traffic over a longer period of time

n overload conditions but otherwise would unnecessarily increase

he access delay Adaptive and load-aware solutions may play again

n important role to keep the optimal WLAN operation in all possible

ircumstances

Finally since IEEE 80211ah will compete with 4G5G cellular net-

orks and WSNs to provide M2M connectivity in many different ap-

lication domains such as smart cities or e-health comparative per-

ormance studies to determine the strong and weak points of each

echnology are required including also aspects such as the cost of the

evices and the system reliability

Cognitive radio technology for TV White Spaces

This section gives an overview of the IEEE 80211af amendment

ntroducing its most relevant features and open challenges Fig 9

hows the basic components and features considered for WLANs op-

rating in the TV White Spaces (TVWS) which include local spectrum

ensing by the AP and STAs and the use of geolocation data bases

ith information on channel availability

1 The IEEE 80211af amendment

Thanks to the transition from analog TV to digital TV several VHF

nd UHF spectrum channels used for decades for analog TV broad-

asting are now unused or are under-utilised in many geographical

reas [127ndash129] This ldquodigital dividendrdquo of spectrum has suggested

ational regulators such as the FCC in the US and Ofcom in the UK

14 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 9 WLAN operating in TVWS basic elements and functionalities

Fig 10 Hidden terminal problem in TVWS when the secondary network uses only in-

stantaneous channel sensing to decide whether or not a TV channel is occupied When

the STA is not able to detect the TV signal due to the presence of obstacles it may

initiate a transmission and create interference to the primary users

s

c

5

a

i

p

t

o

o

t

e

c

u

l

m

b

a

m

t

e

t

t

t

m

D

c

I

n

b

u

8

c

a

t

i

t

w

f

to discuss how to reuse these channels for unlicensed devicesrsquo com-

munications [130131] These potentially vacant channels in the VHF

and UHF bands are referred to as TVWS [132] and include spectrum

portions like the 470ndash790 MHz in Europe and non-contiguous 54ndash72

76ndash88 MHz 174ndash216 MHz 470ndash698 MHz and 698ndash806 MHz in USA

A snapshot of the TV spectrum occupancy in the city of Barcelona in

2012 is shown in [133]

The attractive characteristics of TVWS (not only for WLANs) in-

clude the ability to penetrate through walls and other obstacles much

more effectively than other widely used spectrum bands such as the

24 and 57 GHz ISM bands [134ndash137] This fact along with a progres-

sive spectrum scarcity in ISM bands has suggested the birth of the

IEEE 80211af amendment published in February 2014 [138] which

provides the IEEE 80211 operational characteristics for TVWS access

of unlicensed White Space Devices (WSD) A good summary of IEEE

80211af can be found in [137] The main advantage of operating IEEE

80211 WLANs in the TVWS comes from an increased coverage range

which can reach up to 1 km in rural areas and open fields [139]

and less energy needed to transmit However this comes at the cost

of an increased interference risk to other WSDs which creates co-

existence problems and demands new PHY and MAC layers to ef-

ficiently support channel access and operations preserving licensed

usersrsquo devices

WLANs operated in the TVWS could cover a number of interest-

ing and emerging use cases and scenarios eg Internet access in ru-

ral or sparsely populated areas Smart Grid sensor aggregation me-

tering and control Internet of Things advanced WLAN operations

and TVWS traffic offloading in indoor environments [140141] Nev-

ertheless the concept could be realised to create or extend com-

mercial or municipality WiFi services offered to citizens with cover-

age at the whole city level realised with reasonably limited network

infrastructures [142143]

IEEE 80211af will use the PHY and MAC layers derived from IEEE

80211ac It will adopt concepts such as the OFDM multi-user beam-

forming contiguous and non-contiguous channel bonding and packet

aggregation Among the mandatory and innovative behavioural and

operational parameters the most notable one is the channel acqui-

sition support realised through remote geolocation-based spectrum

allocation databases which maintain the channelsrsquo availability infor-

mation in any given area and time of day providing upon request the

list of free channels available for use

In the following section we highlight and summarise three main

novelties that IEEE 80211af introduces In Section 511 we explain

how the access infrastructure to the remote spectrum database is de-

signed and what the requirements are we discuss the coexistence is-

sues and methodologies adopted in IEEE 80211af and we also discuss

the novel concept of non-contiguous channel bonding The following

ections also contain an illustration of related works Open research

hallenges will be summarised in Section 512

11 Novel features

This section provides an illustration of novel features introduced

nd discussed in the path to IEEE 80211af and the directions provided

n related works and standardisation initiatives to resolve classical

roblems properly characterised in the new framework of TVWS

echnologies

Channel acquisition spectrum database and channel sensing Much

f the attention in operating in the TVWS is given to the protection

f the primary licensed users in the TVWS spectrum band In general

he primary users were considered as the Digital TV (DTV) broadcast-

rs and receivers In fact receivers would suffer the wasteful effect of

ollisions during the TV broadcast reception in the case of secondary

sersrsquo (SU) transmission interference (Fig 10)

It is a well-known problem that the definition of a coherent wire-

ess channel status (idlebusy) from the viewpoint of distributed pri-

ary users (eg DTV receivers) is made difficult by the difference

etween transmitters and receivers interpretation of signals in a vari-

ble time-space collision domain resulting in hiddenexposed ter-

inals concepts Solutions proposed to reduce the hiddenexposed

erminals included the RTSCTS mechanisms and beaconing How-

ver it is difficult to implement solutions that would effectively de-

ermine the channel status of TVWS channels based on localised dis-

ributed sensing activities performed by multiple secondary users in

he same collision domain On the other hand the use of RTSCTS

echanism would not be effective in this case because most of the

TV primary users (and specifically the receivers) were not originally

onceived to transmit signals or implement the RTSCTS handshake

n conclusion the actions to identify free (and busy) TVWS chan-

els should not be based on primary usersrsquo involvement but should

e almost totally implemented by specifically designed secondary

sersrsquo methodologies

There has been much debate on the methodology that IEEE

0211af secondary devices should adopt to get knowledge of free

hannels at a given time on a given target communication area in

n effort to guarantee verified optimised and reliable TVWS spec-

rum use The three main methodologies discussed include sens-

ng solutions geolocation databases (DB) and beaconing Moreover

wo approaches are possible in general distributed and centralised

ith their well known tradeoffs in terms of effectiveness required in-

rastructure deployment and coordination overhead The distributed

B Bellalta et al Computer Communications 75 (2016) 1ndash25 15

m

b

s

c

t

u

a

t

o

l

t

w

t

t

a

o

c

c

a

(

t

o

t

t

D

t

t

o

t

f

a

t

c

i

a

c

d

t

s

t

b

I

l

(

e

a

i

R

p

e

t

c

c

i

b

i

i

t

b

t

t

t

W

a

p

c

i

t

h

p

t

n

a

t

e

t

t

t

p

t

a

r

8

w

s

w

[

i

S

o

a

t

f

i

8

i

r

a

a

i

t

c

(

I

s

a

e

t

t

c

[

i

c

i

g

s

n

s

t

i

o

o

s

c

D

o

ethodology that is receiving more attention in the literature is

ased on distributed channel sensing solutions [127144145] These

olutions include the observation of spectrum use and possible

ooperative aggregation of spectrum sensing information to increase

he accuracy of detection and reduce the vulnerability of primary

sersrsquo transmissions The spectrum sensing capability is required in

ll the secondary devices with a minus114 dBm sensitivity Dynamic

ransmission power control must be provided and the upper limit

n emissions is 100 mW EIRP (20 dBm) for portable devices (further

imited in the case of adjacent channel use to reduce out-of-band in-

erference) However database (DB) coordination of TVWS combined

ith spectrum sensing is considered the most promising and effec-

ive technique compared with spectrum sensing alone The DB spec-

rum information has more chances to be effective (ie identifying

ll free channels opportunities) and reliable (ie not prone to attacks

r erroneous interpretation of channels status) In fact regulators de-

ided to push for the conservative solution of a remotely accessible

entralised spectrum DB that maintains the information on the avail-

bility of all TVWS channels at any given point in time and location

with a target accuracy around 50 m) under their responsibility For

his reason the spectrum databases have become a mandatory part

f many spectrum sharing systems and the dominant technical solu-

ion to support TV white spaces

The IEEE 80211af standard adheres to this vision and also has

he role of a common regulatory framework for different spectrum

B implementations defining a generalised coordination architec-

ure protocols and interfaces for spectrum queries and local spec-

rum information management The protocol explicitly considers out

f scope the communication protocol and technology adopted in

he background leaving the freedom to those players responsible

or the deployment to select any suitable Internet-based and local

ccess network technology Secondary devices that need to access

he TVWS spectrum should get information on which channels they

an effectively use and which parameters to adopt without impact-

ng primary usersrsquo transmissions (and receptions) by querying the

vailable spectrum DB The query must contain the transmission

haracteristics and accurate geographical position of the secondary

evice which can be determined by means of a geo-positioning sys-

em (GPS) regularly updated in the case of movement (or manually

et for static devices) The IEEE 80211af reference system architec-

ure for DB spectrum access is composed of multiple Geolocation-

ased Spectrum information Databases (GDBs) entities connected via

nternet to Registered Location Secure Servers (RLSS) RLSSs work as

ocal proxies of the GDB for a localised group of Basic Service Sets

BSS) RLSS are connected via a secure protocol architecture with the

quivalent of multiple Access Points (AP) in different BSS realising

trusted DB infrastructure APs locally coordinate the exchange of

nformation and channel access management between the GDB (via

LSS) and the secondary usersrsquo end stations (STA)

Mode I devices are those under the control of a device that em-

loys geo-location database access while Mode II devices are those

mploying geo-location database access by themselves The informa-

ion exchange provided between the GDB and a secondary user STA

an be provided in both an open-loop (eg adopted by FCC) and a

losed-loop (eg adopted by ETSI) implementation In an open loop

mplementation daily spectrum availability information is provided

y the GDB and no feedback on the spectrum information received

s provided by STAs thus the system approach for spectrum access

s much more conservative leading to low channelsrsquo utilisation po-

ential In a closed loop implementation the STA can provide feed-

ack to the GDB and there are more communication overheads due

o the high granularity of updates however the system is more effec-

ive and reliable in the exploitation of TVWS on behalf of STAs The

ypical information provided by the GDB includes (1) the updated

hite Space Maps (WSM) of frequencies allowed for secondary use

t the timespace of the querying STA and (2) the device-dependent

ower limitations for transmission (in general conservative and ac-

urate enough to avoid relevant interference effects on primary users

dentified in the area)

The basic IEEE 80211af mechanisms regulating the communica-

ion between STA and GDB can be found in [137138]

In parallel with IEEE 80211af standards such as the IEEE 19006

ave been created that consider interfaces and data structures sup-

orting spectrum sensing information exchange applicable to spec-

rum sensing (and particularly distributed spectrum sensing) sce-

arios In particular in October 2014 the IEEE 19006 WG initiated

project for a new standard called IEEE 19006b [146] concerning

he use of spectrum sensing information to support and optimise the

ffectiveness reliability and robustness of spectrum database solu-

ions The aim is to enhance the performance and capabilities of spec-

rum databases through the use of spectrum sensing information

Co-existence Support for co-existence mechanisms so that mul-

iple technologies can effectively utilise the TVWS spectrum is im-

ortant Self co-existence between network devices of a common

echnology (eg deployed by different operators in the same area)

nd co-existence among different technologies are relevant topic of

esearch for Cognitive Radio (CR) systems and specifically for IEEE

0211af Many solutions appeared on the research scene but no one

as so far finalised as the target solution for IEEE 80211af Specific

tandardisation has been started to regulate coexistence between

ireless standards of unlicensed devices including the IEEE 802191

147] The purpose of the IEEE 802191 standard is to enable the fam-

ly of IEEE 802 Wireless Standards to most effectively use TV White

pace by providing standard coexistence methods among dissimilar

r independently operated TVWS devices Early examples of gener-

lised coexistence mechanisms included Dynamic Frequency Selec-

ion (DFS) Transmission Power Control (TPC) listen before talk (eg

or contention based IEEE 80211 80215) time division multiplex-

ng (also among different techniques such as the IEEE 80216 80220

0222) and Message-based Spectrum Contention (that is beacon-

ng messages that carry coexistence information) Opportune met-

ics must be defined to assess the measurable coexistence achieved

mong different technologies as an example the hidden node prob-

bility for a target scenario or the estimate of percentage variation

n normalised network throughput and latency (before and during

he SU transmissions) On the other hand a centralised coexistence

ontrol mechanism could be effectively realised by a central manager

or coexistence-DB like the GDB) in critical scenarios To this end

EEE 19004 [148] (a standard for heterogeneous networks in dynamic

pectrum context part of IEEE Standards Coordinating Committee 41)

ims to standardise the overall system architecture and information

xchange between the network and mobile devices which will allow

hese elements to optimally choose from available radio resources

There are three possible classification of co-existence archi-

ectures and inter-network coordination channels for CR systems

entralised coordinated and autonomous co-existence mechanisms

149] In centralised co-existence schemes the co-existence is admin-

stered by a central entity (eg like in IEEE 802191) This solution

ould be applied in both homogeneous IEEE 80211af systems (eg

ncarnated by the centralised spectrum DB entity) and in hetero-

eneous inter-networks without requiring modifications to existing

tandards under the assumption that all the involved co-existing

etwork entities would adhere to the same centralised coordination

cheme and inter-network coordination protocol On the other hand

his assumption could be hard to satisfy in some practical scenar-

os The centralised schemes are effective in providing minimisation

f inter-network interference based on the availability and quality

f centralised co-existence information In coordinated co-existence

chemes the centralised entity could be present but not taking de-

isions in general The central entity could provide a co-existence

B with required information which can be used via in-band and

ut-band signalling for co-existence decisions taken by the cluster

16 B Bellalta et al Computer Communications 75 (2016) 1ndash25

c

e

n

c

o

s

i

d

a

S

8

n

a

b

o

b

i

t

a

8

b

[

t

i

n

b

5

t

a

c

b

s

a

c

n

s

p

i

c

l

r

c

f

a

v

t

a

e

c

t

Q

a

b

s

i

c

u

n

c

i

heads of many co-existing inter-networks for implementing a proper

common coordination protocol (eg out-band busy tone signalling)

These are often hybrid solutions realising a compromise between ef-

fectiveness and suitability In autonomous co-existence schemes all the

decisions and coordination are implemented in a distributed way

by the involved network entities In these schemes many policies

can be adopted to realise a sufficient level of co-existence under a

best effort approach Solutions that could be implemented include

the use of busy tones beaconing and other signalling protocols in

dedicated control channels dynamic distributed frequency selection

schemes listen before transmit token-based and dynamic reserva-

tion schemes All the above mentioned solutions have goods and

bads and a general illustration can be found in [149] Other proac-

tive co-existence techniques try to early detect and recovermitigate

co-existence issues with relaxed inter-network coordination eg re-

alised via spectrum sensing and interference avoidancesuppression

techniques

It must be clear that different aims exist regarding the co-

existence for primary TVWS users protection against secondary

users and for secondary devices mutual interference avoidance

Many co-existence problems have been analysed in the literature

for TVWS technologies in particular between the primary (licensed)

and secondary (unlicensed) devices The coexistence management in

TVWS is complex because primary users of DVB transmissions are in-

tended to be the pure receivers rather than the broadcasters In other

words potential protected users could be everywhere and the hid-

den terminal problem could arise for these systems The co-existence

problem is also complicated by many factors that cause asymme-

try and dynamicity in the TVWS such as the mobility variable den-

sity power asymmetry and heterogeneous MACPHY layers TVWS

co-existence problems originate between the IEEE 80211af technolo-

gies and the IEEE 80222 technologies for wireless regional area net-

works (WRAN) Results of analysis of TVWS usage in Europe show

that white spaces are typically present and fragmented They are typ-

ically more abundant in rural areas where larger contiguous blocks

of unused channels are available due to broadcast network planning

giving priorities linked to population density However the exploita-

tion of TVWS in urban areas is possible eg some recent research

was realised under the assumption of exploiting shadowing effects

created in the communication environment (eg by buildings obsta-

cles) in favour of frequency reuse [150]

Since different standards for opportunistic communication in the

TV White Spaces have now been published such as IEEE 80222 [151]

IEEE 802154m [152] Weightless [153] and of course IEEE 80211af

improved methods to guarantee the coexistence of different devices

operating on several protocols in the same bands must be deployed

We highlight here some work eg [150154ndash157] that focuses on

the coexistence between different technologies operating in the same

bands In particular [155] studies the performance degradation of

IEEE 80222 when an IEEE 80211af network is operated in the same

area The problem becomes even worse when the IEEE 80211af net-

work is located near 80222 user equipment causing both networks

to perceive a strong interference from each other One proposed solu-

tion is the Coexistence Beacon Protocol [158] studied for 80222 net-

works which foresees the exchange of a periodic beacon to identify

the neighbouring and possibly interfering networks

For what concerns the co-existence problems research has ad-

dressed other important related aspects of new WLANs on TV White

Spaces For instance the higher coverage range makes them attrac-

tive for applications in the smart grid and also for M2M on TVWS

[134150159] Here the increased range compared to standard tech-

nologies in the ISM bands can have beneficial effects for indoor

mobile devices [150] and M2M [159] Specifically indoor devices

communicating on TV White Spaces offer better propagation char-

acteristics and penetration through obstacles making it easier to

realise scalable home connectivity [135] although interference and

o-existence problems could be exacerbated and must be resolved

g see [139] However studies in the literature show how the sig-

al coming from an indoor transmitter on TV White Spaces remains

onstrained inside the house making it difficult to deploy indoor-to-

utdoor networks On the other hand this shadowing limit can be

een as an opportunity lowering the interference outside the build-

ng in which the TVWS network is locally operated [134150] Un-

er this approach HDTV streaming has recently been considered as

possible use case of WLAN networks operating in the TV White

pace [160]

Non-contiguous channel bonding A relevant novel feature of IEEE

0211af is the potential for contiguous and non-contiguous chan-

el bonding which permits aggregating basic channels (also non-

djacent ones) and leveraging the possible large frequency spread

etween multiple available channels Due to the rather static nature

f primary DTV transmissions where a busy channel is unlikely to

ecome free in the near future and state changes are coarse-grained

n general it is crucial to exploit the time-locality effect and exploit

he maximum physical channel availability that could be aggregated

t any given location With the methodology inherited from IEEE

0211ac IEEE 80211af is capable of bonding together two up to four

asic channels grouped in up to two different non-contiguous chunks

137] The spectrum bandwidth of a DVB-T basic channel can be ei-

her 6 7 or 8 MHz depending on the country in which the service

s operated As an example this creates a 144 to 168 OFDM chan-

elsrsquo bandwidth potential when up to four 6ndash7ndash8 MHz channels are

onded [137]

12 Open challenges

The deployment of WLANs in the TVWS at their maximum poten-

ial still requires the resolution of open problems Generally speaking

s mentioned in previous sections the challenges for such networks

an be divided into three main categories (1) spectrum sharing

etween opportunistic (secondary) and heterogeneous devices (2)

pectrum sensingmanagement and maximum exploitation of avail-

ble spectrum (potentially enabling the spectrum-on-demand con-

ept) and (3) co-existence and interference mitigation to the primary

etwork (primary user protection) and secondary networks [154]

The overall aim of spectrum sharing techniques is to maximise

eparation to avoid overlapping or contiguous channels operations

rovided that sufficient channels are available in the TVWS spectrum

n a given spacetime scenario [154] In general important research

ontributions have to be realised for the design of proper spectrum al-

ocation methodologies and techniques satisfying multi-factorial QoS

equirements at the system-level and the user-level Novel ideas in-

lude the spectrum sharing and spectrum sensing techniques being

urther divided into cooperative and non-cooperative

Non-cooperative solutions attempt to realise spectrum sharing

nd sensing on a local basis without direct cooperation between de-

ices Examples include the ldquolisten-before-transmitrdquo approach and

ransmission power control to limit the interference and spectrum

llocation policies that are based on local node feedback only Coop-

rative solutions rely on the existence of a common communication

hannel through which devices can tentatively agree on the spec-

rum allocation that provides the desired spectrum separation and

oS Major issues include the definition of a methodology to identify

common channel (or multiple channels mutually shared in space

y pairs of heterogeneous and distributed devices) and to efficiently

hare common information For the access to the spectrum DBs the

dentification of proper common access channels is still under dis-

ussion out-of-TVWS-band cellular communications can be widely

sed given their high coverage also indoor

Trying to avoid the use of complex and resource-hungry coordi-

ation schemes and dedicated control channels rendezvous proto-

ols play a significant role both for co-existence and for establish-

ng dynamic communication channels based on spectrum sharing

B Bellalta et al Computer Communications 75 (2016) 1ndash25 17

R

n

c

t

l

o

m

e

n

i

n

w

n

a

u

s

p

b

s

m

s

a

d

t

b

a

a

m

a

t

u

8

n

t

d

d

m

f

t

n

d

n

e

r

a

h

b

f

a

c

t

u

i

f

t

d

t

s

i

f

I

l

i

a

n

c

d

I

n

o

t

i

t

e

w

t

c

m

t

a

i

t

q

b

c

a

i

l

i

a

v

t

a

a

c

t

i

t

b

t

t

a

D

p

c

s

o

q

r

p

a

g

t

e

u

d

t

m

n

t

s

f

i

1

8

f

T

endezvous is considered a fundamental problem in generalised CR

etworks at the basis of many communication processes eg in-

luding neighbour discovery routing and broadcast Rendezvous pro-

ocols attempt to establish a new link for communication on a se-

ected and agreed frequency band (channel) identified within a set

f available resources (channels) thus creating the basis for com-

unication between two or more SUs This problem is often exac-

rbated by the high (and unknown) numbers of contending SUs high

umber of available channels and their ambiguous boundaries and

dentifiers and tough co-existence requirements satisfaction in cog-

itive radio systems [161] Preliminary approaches for rendezvous

ere based on the existence of a dedicated Common Control Chan-

el (CCC) agreed among all the SUs or the existence of a centralised

gent (decision maker) assigning channels to any pairs of requesting

sers This approach could be considered in the case of IEEE 80211af

olutions based on the spectrum DB implementation however the

roblem is complicated by the fact that the common channel must

e identified and made available for all the requesting SUs at the

ame time even when these are spread over multiple collision do-

ains in space On the other hand the scalability and reliability is-

ues which are caused by the congestion of CCCs and the vulner-

bility risks of centralised DBs motivate the research of alternative

istributed solutions called blind rendezvous protocols [162] Dis-

ributed solutions have been mainly based on beaconing andor slot-

ased channel hopping algorithms exploiting randomisation sensing

nd discoverysynchronisation messages spread in slotted time over

set of candidate available channels [161] These protocols aim to

inimise the convergence time to maximise the use of resources

nd to avoid collisions among SUs by identifying and agreeing on

he use of a given common channel among N non-overlapping and

nambiguously labelled channels The main complication in IEEE

0211af systems is given by the wide availability of many different

on-contiguous spectrum bands which cannot be uniformly quan-

ised in single channels and unambiguously labelled in space in or-

er to realise a common reference domain for all the IEEE 80211af

evices In fact the knowledge of the number of channels their com-

on labels the unique IDs of SUs and the number of SUs contending

or the channel assignment are the key factors determining the effec-

iveness and good properties of blind rendezvous protocols Recently

ew classes of rendezvous protocols have been proposed which are

istributed blind and oblivious Oblivious means that available chan-

els may be identified with different labels on behalf of the differ-

nt involved SUs A number of these distributed blind and oblivious

endezvous protocols have been proposed and have been referenced

nd analysed in [162] To conclude solutions for dynamic spectrum

andoff issues (eg in case PUs suddenly appear in a region) must

e resolved in advance to minimise the latency of re-establishing a

unctional communication channel eg by implementing a proactive

lternative rendezvous channel selection in background to ongoing

ommunication processes [163]

Another challenge is to provide cooperative techniques in a dis-

ributed vs centralised implementation by analysing their convergence

nder variable conditions overheads and tradeoffs In some cases as

n vehicular networks the exploitation of scenario characteristics and

actors such as the constrained mobility could help to realise effec-

ive dissemination of sensing information over an extended and pre-

ictable horizon under a cooperative approach [164ndash166] Coopera-

ion could be helpful in the realisation of the cooperative spectrum

ensing in combination with the Geolocation DB approach Another

nteresting direction of research is the investigation of mutual ef-

ects of coexisting cooperation-based vs non-cooperative techniques

n cooperation-based devices the exchange of information could al-

ow a quick convergence to a solution and avoidance of further

nterference [145] Certainly cooperative techniques have desirable

dvantages over non-cooperative ones However tight time synchro-

isation and high coordination overheads are required The common

hannel for cooperation purposes could become a bottleneck and re-

uce the potential advantages of dynamic spectrum allocation The

EEE 80219 foresees the presence of a shared common control chan-

el (CCC) to which devices need to tune in order to gather the state

f the network Solutions such as [141] propose the use of TVWS for

he CCC

A promising direction for research is the adoption of cluster-

ng schemes for secondary devices associated with a hybrid dis-

ributedcooperative vs centralisednon-cooperative approach in an

ffort to reduce overheads while maximising advantages In this

ay secondary nodes could implement cooperation for sensing func-

ions using properly identified common channels and delegate the

entralised decisions and spectrum-DB management to well instru-

ented cluster-leader nodes This concept is incarnated to some ex-

ent by RLSS in IEEE 80211af More recently there have been propos-

ls to build distributed databases that refresh their contents period-

cally by querying the remote spectrum database [167] Here Mas-

er devices (according to Ofcom terminology) periodically cache the

uery replies from the spectrum database in order to reply to and

roadcast the spectrum availability to neighbour Slave devices

Another challenging direction for research is the definition of ac-

urate models and efficient simulation tools enabling dynamic spectrum

nalysis for both rural and urban areas Recent developments of dig-

tal maps simulation tools and propagation models theoretically al-

ow improving the capability of predicting radio propagation effects

n complex scenarios with an acceptable computation time This en-

bling technology could be used in parallel to spectrum DBs to pro-

ide more accurate forecasting of dynamic frequency allocation in

imespace scenarios Another issue to be taken into account is the

ccuracy of the remote spectrum database regarding the channel

vailability estimation Several works have already shown the inac-

uracy of propagation models due to the complexity of parameters

o be considered and differences between modelled and real scenar-

os [132134168169] Better propagation models will make it possible

o more efficiently estimate the interference between devices and

uild Radio Environment Maps (REM) or White Space Maps (WSM)

hat will possibly lead to a more accurate sharing of the radio spec-

rum [144] To this end current research could focus on how to lever-

ge simulation models in order to build more accurate frequency

Bs

In general the adoption of remote spectrum DBs methodology

ushed by national regulators has made the spectrum sensing pro-

ess optional and conservatively realised (eg power control for

ensing-only devices is limited to below 17 dBm EIRP) Thus much

f the research has focused on techniques and solutions to successfully

uery the remote database also in challenging environments such as

ural areas and indoor However since the queries must provide the

osition of the mobile device the accuracy of positioning for mobile

nd handheld devices still needs more appropriate solutions Trian-

ulation can be used to estimate the position of mobile devices with

he accuracy needed by the current regulations outdoor (50 m) How-

ver a low GPS accuracy is possible indoors [170] Certified man-

al positioning can be provided by technicians for static and indoor

evices although limiting mobility Alternatives to costly infrastruc-

ures for positioning (eg based on short range beaconing devices)

ust be deployed in indoor scenarios

Finally regarding spectrum sensing and interference mitigation

ew challenges come from the heterogeneity and different charac-

eristics that networks operating in the TV White Spaces can have

uch as different transmitting power (4 W for static devices 100 mW

or mobile and portable devices and 40 mW for communication

n channels adjacent to occupied ones) different bandwidths (5

0 15 and 20 MHz are currently foreseen for IEEE 80211af IEEE

02154m can have narrower bandwidths or wider ones) and dif-

erent medium access schemes (protocols that use either CSMA or

DMA are required to co-exist in the TV White Spaces) With all this

18 B Bellalta et al Computer Communications 75 (2016) 1ndash25

d

a

T

i

a

c

t

l

b

o

t

a

d

p

a

r

p

n

a

m

t

p

p

d

d

o

t

i

D

w

p

i

a

a

d

b

fi

d

o

t

l

b

t

b

a

i

e

o

o

c

t

t

w

A

c

M

s

c

l

O

w

g

v

t

in place the problem of determining whether another secondary de-

vice is currently transmitting in the same channel is still an open

issue

Spectrum sensing is a very hot topic in the literature with many

proposals that leverage the cooperation between devices to raise the

accuracy of the detection process eg [144171] The current litera-

ture also offers room for proposals that suggest exploiting sensing to

build more accurate spectrum DBs [164172173] Lastly regarding the

interference mitigation to the primary network regulators provide a

conservative approach based on a remote spectrum database to avoid

high risk for the primary users However in particular for the indoor

scenario the concept of secondary usersrsquo transmission in TV Grey

Spaces has been proposed TV Grey Space identifies busy channels

that are formally used by primary users in the area of interest (eg

at the rooftop level of a building for DVB-T) but which could be con-

sidered usable indoors (eg at the basement level) without causing

effective interference to the primary users [134150174] This concept

is based on the generalised spectrum underlay paradigm [175] The

remaining research challenge is to provide TV Grey space access guar-

anteeing a sufficient amount of protection to the primary user while

providing a satisfactory QoS to opportunistic devices eg through

the use of customised and validated propagation models for indoor

TVWS networks and feedback mechanisms to constantly monitor the

interference generated to the primary network

6 Emerging new trends and technologies

In this last section we review three emerging trends that in our

opinion will have a large influence in the conception of future WLANs

as they change the way WLAN protocols and functionalities are de-

veloped implemented tested and integrated with other wireless

networks

61 Programmable Wireless LANs

Especially in the enterprise environment WLAN deployments

need to support a wide range of functionalities and services This is

intrinsically difficult because of the large number of APs that must be

managed which calls for scalable solutions Typical services include

channel assignment load balancing among APs authentication au-

thorisation and accounting (AAA) policy management support for

client mobility and interference coordination Another problem is the

fact that WLAN clients autonomously take several decisions such as

which APs to associate with when to hand-over etc Therefore sup-

porting roaming clients requires the management of a large number

of association states across several APs which is a challenge if sup-

port for real-time hand-over is desired Typically such management

schemes are centralised and most of them are proprietary such as

WLAN controller solutions from Aruba [176] or Cisco [177] although

the IEEE 80211u amendment has been released to allow mobile users

to seamless roam between WiFi networks with automatic authenti-

cation and handoff [178] For example Dyson [179] enables STAs to

send information such as radio channel conditions to a centralised

controller based on a custom API (eg based on Python) As the con-

troller has a centralised view of the network it can enforce a rich

set of policies to control the network also using historical informa-

tion A demo system has been implemented along with applications

such as airtime reservations for specific clients or optimised hand-

offs However Dyson requires STAs to be modified in order to use

those new services offered by the centralised controller TRANTOR

[180] is another example of a centralised management system that

requires changes of clients in order for the infrastructure to gather

information from them in terms of eg interference measurements

In addition control commands enable the infrastructure to exercise

control such as modifying the transmit power or influencing the as-

sociation procedure Clients still use standard CSMA MAC layer for

ata transmission In contrast to DenseAP [181] or DIRAC [182] which

re also based on a centralised controller but do not modify clients

RANTOR exercises much more control on clients by the use of a ded-

cated API which enables a significantly higher gain in coordination

nd thus capacity CENTAUR [183] is another example of a centralised

ontroller that centrally schedules hidden and exposed terminals in

he downlink while using standard CSMA MAC for uplink traffic and

egacy downlink traffic To achieve good performance it uses a fixed

ack-off packet staggering techniques and a hybrid data path that

nly schedules downlink transmissions towards hidden and exposed

erminals centrally All other traffic is sent using standard DCF with

standard back-off procedure In contrast to previous approaches it

oes not require any modifications to the clients but does require data

lane centralisation and it remains questionable how scalable such

solution is for large WLAN deployments with high PHY layer data

ates beyond 100 Mbps However new approaches for fast data plane

rocessing that are currently explored in the virtualisation commu-

ity which move all the packet processing into user space may be

n option to significantly increase speed Designing such manage-

ent systems raises several interesting research questions (a) Cen-

ralised infrastructures may incur high latency but are more simple to

rogram and maintain while designing a completely distributed ap-

roach that operates close to WLAN devices may result in significant

istributed coordination and consistency problems (b) what level of

istributed control is required in order to support the flexibility needs

f future WiFi based networks supporting a high level of mobility

Traditionally WLAN APs are built on proprietary operating sys-

ems that are tightly coupled with the hardware This design makes

t hard to create new applications on top of such networking devices

espite the fact that protocols and mechanisms are available that

ould greatly improve the utility of existing networks those new

rotocols are not deployed because the closed system design makes

t very difficult to extend their functionality An important aspect for

ll such new approaches is therefore how to provide open interfaces

nd open source to speed up innovation As an example Linux-based

evices are completely open source but in order to increase flexi-

ility much more work is needed in the area of open drivers and

rmware In fact the Atheros based ath9k driver has been the main

river of innovation in the open source community because it is the

nly driver that interworks with open firmware In general solutions

hat modify the OS driver focus mainly on programmable network

evel solutions for WLANs such as channel switching or handovers

etween APs While the ath9k driver is an excellent example in how

o speed up innovation it is still very cumbersome to support flexi-

le MAC engine reprogrammability with ath9k In contrast MAClets

nd Wireless MAC processors (WMP) [184] allow a much more flex-

ble reprogramming of MAC functionalities A MAC processor is an

ntity able to execute general MAC commands that specify the MAC

perations through a software-defined state machine The behaviour

f the MAC protocol can therefore be updated at run-time by simply

hanging the sequence in which those commands are executed (ie

he MAClets) As a proof-of-concept the authors of [184] implement

he proposed MAC processor solution in a commodity WLAN hard-

are card extending the basic DCF in three directions piggy-backing

CKs a pseudo TDMA and the use of multiple channels In [185] a

ontrol framework for this WMP system is also proposed to support

AClet code mobility ie for moving loading and activating MAC

oftware programs embedded into ordinary data packets (akin to the

apsule model of traditional active networks)

The difficulty in re-programming networking hardware has also

ed to the concept of Software Defined Networks (SDNs) based on the

penFlow protocol [186] The main idea of SDNs is to extend net-

orking devices with standardised APIs that allow third-party pro-

rammers to flexibly control the data path In addition SDNs pro-

ide higher level abstractions to network designers and programmers

hrough the use of a centralised control plane offered by a network

B Bellalta et al Computer Communications 75 (2016) 1ndash25 19

Fig 11 The CLOUDMAC architecture VAP stands for Virtual Access Point and WT

stands for Wireless Termination

c

a

c

t

m

W

n

t

t

c

p

t

u

h

a

i

e

s

o

c

a

a

f

T

r

s

a

s

p

o

M

C

i

T

t

fl

A

a

a

i

c

t

s

t

e

I

t

w

H

e

b

I

l

d

o

6

r

W

s

l

a

e

e

a

s

b

c

t

(

S

p

c

F

B

t

p

T

e

t

s

u

m

i

e

o

a

A

p

d

t

t

n

p

i

t

t

f

i

ontroller such as NOX [187] which allows reuse of components such

s topology discovery or network access control for different appli-

ations The SDN concept has recently been applied to WLAN archi-

ecture in order to enable fine grained control over mobility manage-

ent and data forwarding focusing on a programmable enterprise

LAN architecture An important part of the SDN architecture is the

etwork controller which provides a centralised view of (parts of)

he SDN enabled network and uses the OpenFlow protocol to install

he forwarding rules on SDN-enabled devices (routers switches ac-

ess points cellular base stations etc)

ODIN [188] is designed to support programmability in enter-

rise WLAN architecture by separating the association state from

he physical AP They implement Light Virtual Access Points (LVAP)

sing a Split-MAC approach where the infrastructure controls the

andover procedure among different WLAN APs By managing

ssociations through SDN controllers ODIN enables proactive mobil-

ty management and load balancing within the SDN enabled WLAN

nterprise network without the need for changes in the client WLAN

tack or IEEE 80211 MAC layer While ODIN requires agents to reside

n the APs that communicate with the Odin Master within the SDN

ontroller CLOUDMAC [189] is based purely on the concept of SDN

nd virtual APs Similar to ODIN CLOUDMAC implements a Split-MAC

pproach but in addition enables the processing of WLAN MAC layer

rames within a co-located Cloud using so-called Virtual APs (VAPs)

he physical APs in CLOUDMAC are lightweight WLAN APs that are

esponsible only for sending out IEEE 80211 based MAC layer ACKs to

tandard WLAN clients and tunnel WLAN MAC layer frames through

n SDN-enabled enterprise WLAN towards the VAPs As association

tates are kept in the VAPs fast mobility is supported using sim-

le OpenFlow forwarding rules Because of the additional processing

f IEEE 80211 WLAN MAC frames in the co-located Cloud CLOUD-

AC has higher latency than a standard WLAN deployment However

LOUDMAC offers a Webservice based API to third party applications

n order to program the enterprise WLAN and enable new services

he architecture of CLOUDMAC is depicted in Fig 11 It has been ex-

ended in [190] to support QoS management and in [191] to support

exible MAC management frame prioritisation based on IEEE 80211

system based on OpenFlow has been recently proposed in [192] to

llow the station to be associated with multiple APs simultaneously

nd to switch between APs with low overhead

An important aspect to consider in order to enable programmabil-

ty is backwards compatibility Several approaches (eg [180]) require

lients to be modified to utilise the features provided This is difficult

o do in practice because it requires changes in the operating system

oftware of all clients If not all clients can utilise those APIs it is ques-

ionable how usable the new architecture will be and how much ben-

fit in terms of aggregate performance such architecture will allow

n contrast the SDN based approaches are interesting in the sense

hat they do not require changes in the client WLAN stack and work

ith the standard IEEE 80211 MAC layer deployed within the clients

owever it remains questionable how scalable such solutions are For

xample an interesting open research topic is to evaluate the scala-

ility of approaches such as CLOUDMAC [189] In addition processing

EEE 80211 MAC frames within co-located private Cloud requires low

atency support from local Cloud solutions (such as OpenStack) in or-

er to reduce the IEEE 80211 MAC processing time which is an area

f active research

2 Prototyping and testing IEEE 80211 enhancements

Most of the new proposals for next-generation WLANs are cur-

ently only evaluated using mathematical analysis and simulation

hile both analysis and simulation are necessary to characterise and

tudy those enhancements in the initial design phase or to consider

arge-scale scenarios it is difficult to consider all practical aspects of

real-world scenario This can sometimes cause significant differ-

nces between what simulations and real experiments show How-

ver real experiments are challenging because of the high complexity

nd costs of building the new hardware and software for each specific

olution to test

To mitigate implementation complexity there are several flexi-

le hardware platforms where low-level MAC mechanisms can be

ompletely implemented in software Examples based on FPGAs are

he USRP (Universal Software Radio Peripheral) [193] and the WARP

Wireless Open-Access Research Platform) [194] As an alternative

ORA (Microsoft Research Software Radio) [195] works on general

urpose computers by taking some advantage from modern multiple

ore systems

A different approach is provided by OpenFWWF [196] Open-

WWF provides an open CSMACA firmware for specific models of

roadcom chipsets so the resulting firmware can be uploaded and

ested in real commercial hardware OpenFWWF implements a sim-

le State Machine (SM) for controlling the hardware in real time

he SM evolution is driven by a main loop that reacts to events by

xecuting specific handlers When a packet originally prepared by

he Linux kernel is ready in the NIC memory handler Packet_Ready

ets up the radio hardware according to the packet metadata sched-

les the transmission and jumps back to the main loop The Trans-

ission Engine (TXE) then takes care of accessing the channel ie

t decrements the back-off counter according to the DCF rules and

ventually starts the actual transmission This triggers the execution

f the TX_frame_now event that prepares the ACK time-out clock

nd finalises the MAC header If the ACK-frame is received or if the

CK time-out expires and the maximum number of attempts for this

acket is reached handler Update_Params resets the contention win-

ow to its minimum value or otherwise doubles it Finally it loads

he back-off counter with a fresh value

In the following we will review a list of selected papers that use

he OpenFWWF firmware and the WARP or USRP platforms to test

ew proposals for WLANs in real scenarios as examples A first im-

lementation of the groupcast mechanisms defined in IEEE 80211aa

s presented in [197] The authors modify the OpenFWWF firmware

o include those functionalities and evaluate them using a 30 STAs

estbed A second example of the use of OpenFWWF is the collision-

ree MAC protocol presented in [198] In [198] the authors design and

mplement in OpenFWWF a MAC protocol that is able to achieve a

20 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 12 LTE to WIFI offloading

a

n

c

t

[

d

a

W

v

t

t

o

b

f

S

g

l

i

I

s

I

b

n

s

m

a

p

d

c

N

o

v

t

t

n

l

i

a

t

a

collision-free operation by waiting a deterministic time after success-

ful transmissions before to start a new transmission This work shows

the experimental performance of a collision-free MAC (CF-MAC) pro-

tocol for WLANs using commercial hardware The results show that

the proposed CF-MAC protocol leads to a better distribution of the

available bandwidth among users a higher throughput and lower

losses than DCF

In [199] the authors use the WARP platform to test a variant of the

DCF ndash called IEEE 80211ec ndash that substitutes control packets such as

the RTS CTS and ACK with short detectable sequences Since those se-

quences are shorter than the control packets and can be detected cor-

rectly even at lower SNRs values a significant gain in performance is

achieved MIDAS (Multiple-Input Distributed Antenna Systems) [200]

implementation using the WARP platform shows the benefits of dis-

tributing the antennas over the area to cover instead of co-locating

them at the AP in terms of capacity when MU-MIMO is employed

The authors also propose a new MAC protocol to benefit from the

spatial reuse that the DAS allows taking as a basis the IEEE 80211ac

Experimental results show that their proposal is able to achieve up

to 200 gains versus the traditional approach where all antennas are

co-located at the AP

The USRP platform has recently been proposed to develop a first

implementation of IEEE 80211ah in order to obtain a preliminary ex-

perimental performance assessment of such a technology because

no commercial off-the-shelf hardware is yet available as the amend-

ment is still in progress [201] The USRP platform is also used in

[202] to evaluate TIMO (Technology Independent Multi-Output) a so-

lution to deal with high-power non-IEEE 80211 interferers in ISM

bands

To conclude this section it is worth mentioning here also to

MAClets and WMP as given such flexibility to implement and dis-

tribute new MAC protocols and other IEEE 80211 functionalities we

believe that if a use-friendly implementation of such a MAC proces-

sors framework is provided it would significantly contribute to the

development and testing of new MAC enhancements by the research

community

63 CellularWLAN interworking

Public hotspots that offer Internet access over a WLAN using IEEE

80211 technology are now nearly ubiquitous It is forecasted that the

cumulative installed base of WiFi hotspots worldwide will amount

to 551 million by 2018 excluding private hotspots (eg WiFi access

points deployed at home) [203] The sharp increase in the availabil-

ity of public WiFi was initially perceived by mobile cellular operators

as a threat due to the additional competition from wireline Internet

service providers or emerging crowdsourced WiFi networks such as

FON5 However as cellular operators are fighting to cope with the ex-

plosion of mobile data traffic created by the rising use of multimedia

content traffic over mobile devices [15] they are also starting to use

WLANs based on the IEEE 80211 technology to offload data from their

core and access networks In general mobile data offloading refers to

the use of complementary network technologies (in licensed or un-

licensed spectrum) for delivering data originally targeted to cellular

networks Intuitively the simplest type of offloading consists of ex-

ploiting connectivity to existing co-located WiFi networks and trans-

ferring data without any delay (Fig 12) Thus this offloading tech-

nique is known as on-the-spot offloading As a consequence of this

new trend the seamless integration of cellular (eg 3GLTE) and WiFi

technologies has attracted significant research interest in recent years

(see [204] for a survey) a few solutions have already been standard-

ised [205206] and roaming between cellular and WiFi is becoming

increasingly transparent to end users CellularWLAN interworking is

5 httpscorpfoncomen

fl

t

f

lso fostered by the support in the evolving 4G standards of heteroge-

eous network deployments (HetNets) in which the existing macro

ells are complemented with a number of small low-power base sta-

ions with the goal of increasing capacity in highly congested areas

207208] It is envisaged that small cells will be based on 4G stan-

ards (eg pico and femto cells) as well as IEEE 80211 technologies

nd multimode base stations that work simultaneously with LTE and

iFi are already entering the market

It is important to point out that LTE standards already support a

ariety of mechanisms that enable data offloading However most of

he existing solutions such as Local IP Access (LIPA) and Selected In-

ernet IP Traffic Offload (SIPTO) focus on data offloading at the core

f the network [205] For instance LIPA allows an IP-enabled mo-

ile device to transfer data to another device in the same pico or

emto cell without passing through the cellular access network while

IPTO enables the routing of selected IP data flows through different

ateways The only LTE offloading mechanism that supports seam-

ess interworking with IEEE 80211-based WLANs is IP Flow Mobil-

ty and Seamless Offload (IFOM) Specifically IFOM relies on Mobile

Pv6 technologies to allow a user terminal to simultaneously route

elected IP flows over different radio access technologies [209210]

n this way a user terminal can offload selected flows to a WLAN

ased on some operator-defined policy while keeping the LTE con-

ection running However to enable such an approach it is neces-

ary to support an entity in the cellular core network that can com-

unicate with the user terminal to exchange information about the

vailability and quality of neighbouring access networks as well as to

rovide the user terminal with predefined rules to manage the han-

over process This entity in current LTE standards is called the Ac-

ess Network Discovery and Selection Function (ANDSF) server [211]

ote that the problem of selecting the best communication technol-

gy in an heterogeneous wireless network has been extensively in-

estigated in the literature both with centralised as well as decen-

ralised approaches (see [212] for a survey) However the use of mul-

iple interfaces in parallel as well as per-flow offloading are relatively

ew concepts Thus the design of scalable and efficient network se-

ection strategies for the ANDSF framework is still an open issue For

nstance three offloading methods suitable for the ANDSF framework

re proposed in [213] which are based on coverage SNR and sys-

em load A reinforcement learning approach is designed in [214] that

llows multimode base stations to autonomously steer their traffic

ows across different access technologies depending on the traffic

ype the usersrsquo QoS requirements the network load and the inter-

erence levels A number of studies have also tried to quantify the

B Bellalta et al Computer Communications 75 (2016) 1ndash25 7

d

e

I

w

f

t

s

3

P

s

G

t

8

p

C

c

m

c

i

t

a

s

w

f

t

t

e

l

p

e

a

H

d

t

c

o

s

p

n

t

b

a

w

c

t

w

n

8

a

c

w

d

H

n

a

h

p

w

i

t

fi

e

c

u

t

m

p

p

b

p

i

t

8

d

I

m

M

u

t

m

i

t

D

W

I

u

M

i

w

p

h

t

n

o

p

t

s

u

s

r

b

A

t

O

t

t

w

s

p

o

i

W

t

t

a

[

m

ense scenarios can contribute to notably improving the user experi-

nce We further discuss this last topic in Section 6

Different from the other amendments covered in this survey the

EEE 80211ax amendment is just in its initial stages of development

ith only very few technical aspects consolidated at this stage There-

ore in the following subsection we will overview both the new fea-

ures and open challenges of the IEEE 80211ax amendment at the

ame time

21 Novel features and open challenges

The IEEE 80111ax Task Group is currently working in four areas

HY MAC Multi-user and Spatial Reuse [49] Next we will overview

ome of the topics currently under discussion in the IEEE 80211 Task

roup in each category

PHY layer The IEEE 80211ax PHY layer will be an evolution of

he IEEE 80211ac one The challenges in the design of the IEEE

0211ax PHY layer are related with the extensions required to sup-

ort multi-user MU-MIMO and OFDMA transmissions and Dynamic

CA Also improvements in the supported modulation and channel

oding techniques will be likely considered to allow for higher trans-

ission rates at lower SNR values For example IEEE 80211ax may

onsider LDPC (Low-Density Parity Check) coding which is optional

n IEEE 80211ac instead of the traditional convolutional codes as

hey provide a coding gain of 1ndash2 dB Moreover the PHY layer may

lso include some functionalities to support the use of Hybrid ARQ

chemes to improve the efficiency of packet retransmissions

Medium acess control In order to keep backward compatibility

ith previous IEEE 80211 amendments besides a common PHY

rame preamble compatible MAC protocols are required This means

hat it is likely that EDCA will be kept as the main channel access

echnique in the IEEE 80211ax amendment Therefore the most rel-

vant open challenges are related to EDCA extensions to support a

arge number of STAs improve traffic differentiation capabilities im-

rove the energy consumption and provide mechanisms to fairly co-

xist with neighboring wireless networks

To support a large number of contenders with a low collision prob-

bility a simple solution is to use large backoff contention windows

owever it would also increase the time a STA is in backoff so re-

ucing the number of packets it can transmit A solution to reduce

he backoff duration when using large contention windows is to de-

rease the duration of the backoff slots Latest technological advances

n electronics should require less time to perform a CCA check to

witch between reception and transmission modes and to process a

acket thus enabling such a possibility

Another approach to enhance the underlying CSMACA mecha-

ism in EDCA and achieve a higher efficiency is to consider decen-

ralised collision-free MAC strategies Those MAC protocols are able to

uild collision-free schedules thus improving the network efficiency

s collisions are reduced while preserving backward compatibility

ith the default EDCA implementation An overview of several de-

entralised collision-free MAC protocols can be found in [55] Among

hem CSMAECA [56] is especially relevant since it is fully compatible

ith EDCA and latest upgrades made it adaptive to the instantaneous

umber of contenders in a single WLAN

In addition to EDCA IEEE 80211ax WLANs can rely on the IEEE

0211aa amendment to further improve the WLAN traffic differenti-

tion capabilities with intra-AC traffic differentiation and groupcast

ommunication mechanisms among other features In Section 33

e will provide further details about the IEEE 80211aa amendment

IEEE 80211ax will likely keep the same channel widths that were

efined in the IEEE 80211ac amendment ie 20 40 80 and 160 MHz

owever it is expected that IEEE 80211ax will extend current chan-

el bonding mechanisms to further improve the spectrum utilisation

nd the coexistence between neighbouring WLANs For example it

as been shown in [32] that the use of dynamic channel bonding

rovides significant throughput gains in dense scenarios compared

ith the static approach while minimising the inter-WLAN negative

nteractions [27] Furthermore additional mechanisms are required

o fully exploit the use of channel bonding such as the use of ef-

cient algorithms to select the position of the primary channel or

ven to consider the existence of multiple primary channels to in-

rease the number of bonded channel combinations that a node can

se for transmission

The MAC layer in IEEE 80211ax may work with the PHY layer

o implement an efficient Hybrid ARQ mechanism able to retrans-

it only incremental redundancy bits in short packets Opportunistic

iggy backing of data packets in ACKs and vice versa may further im-

rove the efficiency of IEEE 80211ax WLANs by reducing the num-

er of transmissions in a bidirectional data exchange [57] Finally

acket headers can be reduced if shorter STA identificators are used

nstead of MAC addresses and unnecessary fields are avoided in all

ransmissions

Finally in terms of power saving mechanisms it is expected IEEE

0211ax will follow the same trends as in previous IEEE 80211 stan-

ards including the extensions currently under development for the

EEE 80211ah amendment that are discussed in Section 4

Multi-user Multi-user communications will likely be one of the

ain characteristics of IEEE 80211ax as both uplink and downlink

U-MUMO and OFDMA are under consideration The use of multi-

ser communication techniques does not necessarily represent a sys-

em capacity increase because the available transmission resources

ay be the same as in the single-user communication case However

n WLANs the simultaneous transmission from different users is able

o parallelise the large temporal overheads of each transmission (ie

IFS SIFS ACKs packet headers etc) which can notably improve the

LAN efficiency

IEEE 80211ax will further develop the MU-MIMO capabilities of

EEE 80211ac by allowing multiple simultaneous transmissions in the

plink which is known as uplink (UL-) MU-MIMO [58] Similar to DL-

U-MIMO transmissions an open challenge to enable UL-MU-MIMO

s to design a mechanism able to efficiently schedule the users that

ill transmit at the same time In one hand a pure decentralised ap-

roach would be easy to implement with minimal signalling over-

eads However since it requires that all STAs finish their backoff at

he same time it may be very inefficient besides that those STAs may

ot be compatible in terms of their respective spatial channels In the

ther hand a pure centralised approach requires that the AP has com-

lete CSI and buffer occupancy information from all STAs to select

he most suitable group to perform a multi-user transmission Once a

uitable group of STAs is selected by the AP a ldquoTriggerrdquo frame may be

sed to notify the group of selected users that can initiate a transmis-

ion This approach guarantees efficient multi-user transmissions but

equires some extra overheads to collect all the required information

y the AP and signal the selected STAs In both cases new multi-user

CKs will be likely introduced by IEEE 80211ax to acknowledge all

ransmissions with a single control packet

Multi-user OFDMA is also in the agenda for IEEE 80211ax Using

FDMA a channel can be split in several sub-channels and assigned

o different users Likely OFDMA will be implemented in combina-

ion with channel bonding where each of the 20 MHz subchannels

ill be assigned to a different user in both downlink and uplink Be-

ides that a similar operation as in the multi-user MIMO case is ex-

ected as there are almost the same challenges to solve A survey

f current OFDMA proposals for WLANs is presented in [59] show-

ng also how the use of OFDMA is able to significantly improve the

LAN efficiency Fig 4 illustrates an uplink OFDMA and MU-MIMO

ransmission

In addition to Multi-user MIMO and OFDMA the use of Simul-

aneous Transmit and Receive (STR) techniques commonly known

s full-duplex transmission have been suggested for IEEE 80211ax

5051] Using STR a pair of nodes is able to transmit and receive si-

ultaneously [60] which theoretically doubles the channel capacity

8 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 4 Multiuser uplink transmission using two different OFDM subchannels Trans-

missions from STA B and C over the same subchannel represent an uplink MU-MIMO

transmission All transmissions are acknowledged with a single Multiuser ACK The

trigger frame is used to signal the selected STAs

t

t

t

a

s

t

f

f

c

3

s

s

t

n

n

A

d

w

t

fi

a

t

8

i

d

t

r

w

d

R

v

f

s

d

p

B

c

e

k

r

G

a

c

m

t

b

d

3 The concealment address protects legacy stations ie GCR-incapable stations

from receiving duplicated group-addressed frames

The challenge is that both the AP and the STA involved in a full-duplex

transmission have to start to transmit at the same time To achieve

that information about full duplex transmission capabilities can be

included in RTSCTS control packets to set-up a full-duplex transmis-

sion between the transmission initiator and its destination

Spatial reuse Dense WLAN deployments are necessary to offer a

continuous coverage with high transmission rates To improve both

the co-existence with those neighboring networks and the spatial

reuse of the spectrum a WLAN has two options (i) minimise its

area of influence by reducing its transmit power and (ii) accept

higher interference levels by increasing the Clear Channel Assess-

ment (CCA) level Use of both techniques may increase the number of

concurrent transmissions between neighbouring WLANs and there-

fore their capacity although it may also result in the opposite effect

since the achievable transmission rates may be negatively affected by

the higher interference levels observed which is the main challenge

to solve

Due the high WLAN dynamics the use of adaptive systems is cru-

cial but challenging as adaptivity requires extra complexity in terms

of computing and memory resources and there are no guarantees

that the implemented solution converges due to the decentralised

operation of each WLAN The use of DSC (Dynamic Sensitivity Con-

trol) to dynamically adjust the CCA level is one of the aspects cur-

rently under discussion in the IEEE 80211ax Task Group First studies

evaluating the performance of DSC for IEEE 80211ax WLANs show

a clear improvement on the spatial reuse and the area throughput

[61] Another example of the achievable throughput gains obtained

by adapting the CCA level can be found in [62] where the authors

show that gains of up to 100 can be achieved Moreover trans-

mit Power Control (TPC) to mitigate interference between WLANs in

dense scenarios is studied in [63] showing the need of jointly opti-

mising both TPC and CCA to maximise the network performance

Finally sectorisation by using beamforming is also under consid-

eration for the development of the IEEE 80211ax amendment as a

potential solution to improve spatial reuse [64] Using sectorisation

only the nodes of a given area are allowed to receive or transmit data

hence reducing the contention between different networks whenever

they activate non-overlapping sectors A challenge here is to coor-

dinate the different neighboring APs when they belong to different

administration domains Decentralised learning approaches may be

implemented to find feasible temporal patterns of non-overlapping

sectors

33 The IEEE 80211aa amendment

As discussed above legacy IEEE 80211 standards do not provide

robust and efficient delivery of audiovideo streaming services Thus

he IEEE 80211aa amendment was developed to include new fea-

ures and additional mechanisms to improve the performance of real-

ime multi-media content delivery [65] Specifically IEEE 80211aa

ddresses the following five shortcomings of previous 80211

tandards [1666]

(i) the lack of reliable and efficient support for multicast and

group communications

(ii) the incapacity of applying traffic prioritisation to different

multimedia streams or different types of frames from the same

stream

(iii) the absence of methods for cooperative resource sharing

among neighbouring APs

(iv) the lack of mechanisms for graceful degradation of audiovideo

streaming quality

(v) the non-interoperability with existing IEEE 8021 standards for

Audio Video Bridging (AVB)

In the following sections we present in detail the solutions to

hose problems introduced in the IEEE 80211aa amendment We

urther discuss the research studies that have provided the basis

or the IEEE 80211aa design and we identify the remaining open

hallenges

31 Novel features

Groupcast communication mechanisms In most audiovideo

treaming applications a group of clients must receive the same

tream simultaneously A multicast protocol is necessary to avoid that

he same content is replicated throughout the network In wireless

etworks multicast transmission can exploit the intrinsic broadcast

ature of the wireless channel ie broadcast transmissions from an

P are physically received by all other stations in the same collision

omain However multicast and broadcast frames in IEEE 80211 net-

orks are not protected by an acknowledgement mechanism as in

he case of unicast frames Thus layer-2 multicast transmissions de-

ned by legacy IEEE 80211 standards are unreliable and not suit-

ble for streaming applications To partially address this limitation

he Direct Multicast Service (DMS) was first specified in the IEEE

0211v amendment [67] Basically DMS converts multicast streams

nto unicast streams In this way frames destined to a multicast ad-

ress are individually transmitted as unicast frames to the stations

hat joined that multicast group Obviously DMS provides the same

eliability as unicast transmission services but the consumed band-

idth increases linearly with the number of group members To ad-

ress this scalability issue IEEE 80211aa includes the Groupcast with

etries (GCR) service in addition to DMS Specifically the GCR ser-

ice defines new mechanisms and the related management frames

or group formation which allows a set of stations to agree on a

hared (non-multicast) address called the groupcast concealment ad-

ress3 Furthermore the GCR service specifies two retransmission

olicies GCR Unsolicited Retry (GCR-UR) and GCR Block Ack (GCR-

A) When using GCR-UR the AP can proactively retransmit all group-

ast frames a number of times to mitigate the impact of channel

rrors (see Fig 5a)) while receivers are not requested to send ac-

nowledgements Intuitively this approach improves transmission

eliability but it still suffers from scalability issues In contrast when

CR-BA is used the AP sends a burst of consecutive groupcast frames

nd it requests the receivers to reply with a Block ACK frame which

ontains a bitmap to positively or negatively acknowledge trans-

itted frames (see Fig 5b)) The Block ACK mechanism defined for

he GCR-BA service is quite flexible because Block ACK frames can

e requested immediately after a transmission burst or after a ran-

omised back-off delay Furthermore the AP can request the Block

B Bellalta et al Computer Communications 75 (2016) 1ndash25 9

Fig 5 GCR service with different retransmission schemes

A

r

o

l

o

e

v

e

v

d

f

d

s

v

c

s

a

a

f

d

t

F

t

d

v

a

f

c

t

t

o

t

i

c

r

v

i

r

s

o

t

a

(

p

i

f

a

o

i

t

t

c

s

fi

o

c

I

n

i

a

t

a

t

f

i

B

T

f

t

a

g

a

a

i

s

a

d

s

p

m

m

i

a

t

b

t

i

(

r

a

S

i

c

s

c

T

Q

p

a

3

g

t

4 TSPEC is a traffic specification sent from a QoS capable wireless client that requests

a certain amount of network traffic from the AP for the traffic stream it represents

CK frame to all groupcast recipients or only to a subset of them to

educe overheads and delays The advantages of the GCR methods

ver broadcast and DMS have been extensively demonstrated in the

iterature [1668]

Intra-access category prioritisation The IEEE 80211e amendment

nly allows traffic differentiation between four different access cat-

gories (ACs) that are broadly mapped to four application classes

oice (VO) video (VD) best-effort (BE) and background (BK) How-

ver there is a variety of streaming services ranging from simple

ideoconferencing to HD streaming over IPTV systems which have

ifferent QoS requirements (see Table 3) To provide the ability to dif-

erentiate among individual streams IEEE 80211aa includes an ad-

itional scheduling layer with respect to IEEE 80211e IEEE 80211aa

plits each one of the transmission queues associated with voice and

ideo ACs into a primary and an alternate queue In this way spe-

ialised scheduling rules can be applied to decide which queue to

erve when the EDCA function for inter-AC collision resolution grants

n access opportunity to voice or video ACs To facilitate the man-

gement of service level agreements IEEE 80211aa follows the de-

ault mappings between user priority values and traffic types that are

efined in the IEEE 8021D standard [69] It is then straightforward

o further map traffic types onto transmission queues and ACs (see

ig 6) Finally it is important to point out that the intra-AC differen-

iation functionality can be used to provide more sophisticated traffic

ifferentiation than simple stream prioritisation For instance most

ideo applications use Scalable Video Coding (SVC) schemes that en-

ble the partitioning of a video sequence into multiple layers with dif-

erent qualities and rates [70] Typically an SVC-based video stream

ontains a base layer which provides a basic level of quality and mul-

iple enhancement layers which can only be decoded together with

he base layer to improve the video quality Thus the different layers

f the same encoded video steam can be easily mapped to different

ransmission queues to receive differentiated QoS [71]

Stream classification service The stream classification service (SCS)

s an optional service that can be provided by an AP to the asso-

iated stations to classify multimedia streams based on arbitrary

ules that are established directly by the stations instead of the con-

entional 8021D user priorities To this end the station request-

ng the use of SCS must initiate an SCS session by sending an SCS

equest frame to the AP which contains an identifier for the SCS

tream and the descriptor of the classification rule The AP may accept

r reject the requirements specified by the station Once accepted

he AP must assign all frames that match the classification rule to

specific AC When intra-access category prioritisation is enabled

see Section 331) calternate transmit queues so that finer grained

rioritisation can be applied Finally there is also a Drop Eligibil-

ty Indicator (DEI) bit in the SCS descriptor that indicates whether

rames from this traffic stream can be dropped in the case that there

re insufficient resources Specifically frames with the DEI bit set to

ne have a higher probability of being discarded because their max-

mum number of allowed retries is smaller than the default Note

hat how to combine intra-AC queues and frame dropping settings

o achieve graceful degradation of the audiovideo stream quality in

ase of bandwidth shortage is beyond the scope of the IEEE 80211aa

pecification

Overlapping Basic Service Set (OBSS) management Network densi-

cation ie a denser deployment of wireless infrastructure nodes is

ne of the key strategies that is used nowadays to easily increase the

apacity of wireless systems even for indoor WLANs [72] However

EEE 80211 networks have a limited number of orthogonal chan-

els available and even if optimised frequency planning is applied

t might happen that neighbouring APs are mutually interfering and

station may affect multiple overlapping BSSs In this case conges-

ion not only increases but it is also likely to observe an unfair us-

ge of wireless capacity with the channel retained by one AP for long

ime intervals This is mainly due to the neighbourhood capture ef-

ect ie hidden terminal phenomena among APs [73] To address this

ssue IEEE 80211aa specifies a new functionality called Overlapping

SS (OBSS) management which is based on two new mechanisms

he first defines a set of parameters to quantify the load and inter-

erence among neighbouring BSSs such as medium occupancy frac-

ion number of admitted audiovideo streams data traffic volumes

nd the number of BSSs that are using the same channel as the tar-

et one Note that the traffic load consists of two components the

llocated traffic which is derived on the basis of the TSPEC values of

dmitted streams4 and predicted traffic which is evaluated by track-

ng the maximum value of the allocated EDCA and HCCA traffic over

even-day periods Once load measurement reports are exchanged

mong the APs a second OBSS component is responsible for coor-

inated admission control procedures on the basis of two suggested

haring schemes proportional sharing and on-demand sharing The

urpose of both schemes is to keep the total allocated traffic below a

aximum value in order to provide some QoS protection to admitted

ultimedia streams Finally IEEE 80211aa recommends implement-

ng additional OBSS management procedures for channel selection

nd cooperatively creating HCCA schedules that do not collide

Interworking with IEEE 8021AVB Audio Video Bridging (AVB) is a

erm commonly used to denote a set of technical standards developed

y IEEE to support real-time streaming services with bounded latency

hrough IEEE 802 networks [74] This objective is achieved by spec-

fying mechanisms to allow the synchronisation of multiple streams

IEEE 8021AS [75]) and traffic shaping (IEEE 8021Qav [76]) and to

eserve network resources for specific audiovideo streams traversing

bridged local area network by using a signalling protocol called the

tream Reservation Protocol (SRP) (IEEE 8021Qat [77]) IEEE 80211aa

ntegrates the SRP operations with the EDCA admission control pro-

edures Specifically the SRP RequestResponse messages are encap-

ulated in the management frames that are used to carry the traffic

haracteristics and the QoS requirements during admission control

his enables the end-to-end management of resource reservation for

oS guaranteed streams even when one or more IEEE 80211 links are

art of a path from the stream producers (called IEEE 8021Q talkers)

nd the stream consumers (called IEEE 8021Q listeners)

32 Open challenges

In recent years several MAC enhancements have been investi-

ated to improve QoS guarantees for real-time multimedia applica-

ions in IEEE 80211 networks [20] and the IEEE 80211aa standard

10 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 6 Stream classification and inter-AC traffic prioritisation

a

l

t

r

n

a

t

n

i

t

i

t

r

d

p

t

t

e

m

b

t

a

i

s

o

s

a

e

t

i

f

e

m

8

t

I

i

fi

which was finalised in 2012 included several of these proposed im-

provements Significant research efforts have focused on improving

the transmission reliability of multicasting by integrating ARQ mech-

anisms in IEEE 80211-based multicast transmissions Modifications

to the legacy MAC protocol were proposed in [78] to enable the

RTSCTS option in multicast mode and to select one or more multi-

cast receivers (called leaders) for acknowledging multicast data pack-

ets However these enhancements require changes to the standard

specifications The main problems of leader-based ARQ schemes are

leader election and the trade-off between scalability and reliability

The authors in [79] propose selecting the multicast recipient operat-

ing in the worst channel conditions as the unique leader but this ap-

proach may perform poorly in lossy environments In the Batch mode

multicast MAC (BMMM) [80] all multicast recipients are polled by

the multicast originator to send individual ACKs but this scheme is

not suitable for large multicast groups The Enhanced Leader Based

Protocol (ELBP) is proposed in [81] on the basis of multiple ACK-

leaders and block acknowledgement techniques Analytical models

are then developed to help select optimal ACK-leaders to meet ap-

plication QoS requirements However the models apply only to sat-

urated traffic while multimedia streams are typically bursty Another

class of reliable multicast protocols relies on busy tones to reduce

packet losses due to collisions [82] but the additional radio inter-

face needed for the busy tone limits the practicality of such solu-

tions An alternative approach to avoid collisions of multicast pack-

ets is the multicast collision prevention (MCP) scheme [83] which

is based on the use of a shorter waiting time for transmitting mul-

ticast packets An interesting approach is also proposed in [84] to

retransmit lost packets using an online linear XOR coding algorithm

However a modification to the standard MAC protocol is required to

enable simultaneous ACK transmissions In summary several differ-

ent methods have been proposed to improve multicast transmission

reliability by integrating ARQ schemes into the protocol architecture

but there are not conclusive results on which is the best solution The

choice of the most efficient mechanism depends on a variety of in-

terdependent factors such as loss ratios channel congestion multi-

cast group size and QoS requirements of multimedia streams A com-

prehensive analytical framework is needed to optimise the setting of

the parameters for each scheme and to dynamically select the best

one

As discussed above one main difference between unicast services

nd multicast services in the legacy IEEE 80211 standard was the

ack of acknowledgements Another critical difference is that mul-

icast frames must be transmitted using a fixed rate in the basic

ate set while the transmission rate of unicast frames can be dy-

amically adapted to the channel and traffic conditions [85] Thus

group of research papers has investigated the use of rate adap-

ation to improve the throughput of multicast services in IEEE 80211

etworks [7186ndash89] For instance the authors in [86] propose us-

ng RTS frames to allow group members to estimate channel condi-

ions Each member will then send a dummy CTS frame with a length

nversely proportional to channel quality In this way the multicast

ransmitter can use the collision duration to predict the lowest data

ate that can be used for group transmissions The overhead intro-

uced by this mechanism is quite high however The solution pro-

osed in [88] called ARSM also relies on feedback messages sent by

he multicast receivers called multicast response frames to identify

he group member exhibiting the poorest channel conditions How-

ver in this case a different back off timer is associated with each

ulticast receiver depending on the SNR of previously received feed-

ack messages in order to prevent collision An approach similar to

he one employed in the Auto Rate Fallback (ARF) protocol a rate

daptation scheme originally proposed in [90] is used in [87] Specif-

cally the number of successful consecutive transmissions and con-

ecutive transmission failures are used to decide when to increase

r decrease the transmission data rate respectively A modified ARF

cheme is also proposed in [71] which can be applied to videos that

re encoded into two layers namely the base and enhancement lay-

rs However how to integrate rate adaptation with the different re-

ransmission policies that are defined in IEEE 80211aa is still an open

ssue

One research area that is expected to be crucial in the success-

ul development of IEEE 80211aa-based products is the design of

fficient scheduling algorithms for supporting voicevideo traffic Al-

ost all research work in this field has been triggered by the IEEE

0211e amendment that enhanced the original IEEE 80211 MAC with

wo new QoS-aware access mechanisms ie EDCA and HCCA [91]

n principle with a well-designed admission control and schedul-

ng scheme HCCA is able to provide hard QoS guarantees to traf-

c flows [9293] However HCCA is rarely implemented in IEEE

B Bellalta et al Computer Communications 75 (2016) 1ndash25 11

8

c

c

a

s

b

w

o

j

t

f

a

Q

t

t

A

p

t

s

a

f

s

H

a

i

c

i

t

s

f

s

s

m

i

t

s

l

t

t

[

t

i

o

b

r

W

v

m

8

s

i

o

p

q

4

c

e

t

d

a

t

m

Fig 7 WLANs for M2M communications STAs represent sensor and actuator devices

4

t

w

s

a

c

g

e

(

p

o

t

p

9

r

o

a

p

c

o

i

p

b

f

t

s

o

i

4

8

f

t

i

s

m

d

0211e-based WLANs owing to its higher complexity and cost con-

erns Instead EDCA is widely adopted Most papers have thus fo-

used on improving EDCA performance Many papers have proposed

nalytical models for various subsets of EDCA functionalities For in-

tance a saturation-based performance analysis is conducted in [94]

y differentiating the minimum back-off window size the back-off

indow-increasing factor and the retransmission limit The authors

f [9596] also model AIFS differentiation while the model in [97]

ointly captures all the four EDCA parameters for traffic differen-

iation More recent papers have analysed the EDCA performance

or non-saturated conditions and for arbitrary buffer sizes [98] The

uthors in [99] have developed an analytical model to predict the

oS levels that can be achieved once a new voicevideo flow is in-

roduced in the WLAN A Kalman filter is proposed in [100] to ob-

ain estimates on the number of active transmission queues of each

ccess Category in EDCA These analytical models can then be ex-

loited to derive the optimal configuration of the EDCA parameters

o achieve given performance criteria or to design admission control

chemes that preserve QoS constraints For instance a scheme that

ssigns contention-window values to achieve pre-defined weighted-

airness goals is proposed in [101] A control-theoretic scheme is de-

igned in [102] with the goal of minimising the video traffic delay

owever most of these solutions rely on non-realistic assumptions

bout video traffic dynamics An alternative class of solutions dynam-

cally updates the EDCA parameters based on the observed network

onditions In [103] the EDCA parameters are optimised consider-

ng a WLAN with rigid and elastic traffic simultaneously analysing

he interactions between both types of traffic The authors in [104]

pecify several bandwidth-sharing mechanisms with guaranteed QoS

or voice and video traffic Measurement-based admission control

chemes are proposed in [105] A TXOP adaptation method is de-

cribed in [106] that takes into account video frame sizes and trans-

it queue lengths However the main drawback of these solutions

s that they are based on heuristics and hence do not ensure op-

imal and guaranteed performance Finally a third category of re-

earch papers tries to improve video performance by designing cross-

ayer scheduling approaches Specifically these works take advan-

age of multi-layer video encoding to classify the frames according

o their importance and assign them to different access categories

107] For instance the authors in [108] define classifiers and waiting

ime priority schedulers that dynamically change the packet prior-

ties according to end-to-end delay measurements A disadvantage

f this approach however is that an additional adaptation layer may

e needed to implement the complex interactions that are typically

equired between the video coding applications and the MAC layer

e conclude this section by pointing out that existing studies pro-

ide the basic design principles and techniques for improving multi-

edia streaming performance in IEEE 80211 networks Still the IEEE

0211aa standard poses new research challenges that have not been

ufficiently explored and that will require innovative solutions For

nstance scheduling between primary and alternate queues is still an

pen research area as the mapping of individual frames to multi-

le queues in order to achieve graceful degradation of voicevideo

uality [16]

Sensor Networks and machine-type communications

As discussed in Section 2 M2M communications refer to any

ommunication technology that enables sensoractuator devices to

xchange information and perform actions without the manual assis-

ance of humans This section reviews the main features currently un-

er consideration in the development of the upcoming IEEE 80211ah

mendment which targets the main challenges of those networks as

he IoT in general such as the energy consumption or the manage-

ent of many devices

1 The IEEE 80211ah amendment

The IEEE 80211ah amendment [109] aims to provide WLANs with

he ability to both manage a large number of heterogeneous STAs

ithin a single BSS and minimise the energy consumption of the

ensor-type battery-powered STAs

The initial design requirements of the IEEE 80211ah amendment

re detailed in [110] these entail the support of up to 8192 STAs asso-

iated with a single AP the adoption of efficient power saving strate-

ies a minimum data rate of 100 kbps the operation in the license-

xempt sub 1 GHz band and a coverage up to 1 km in outdoor areas

see Fig 7 for an illustrative example) A preliminary assessment of

erformance of the IEEE 80211ah technology in terms of the number

f STAs that can be effectively supported in a single WLAN as well as

heir energy consumption is presented in [111]

IEEE 80211ah operates over different sub-1 GHz ISM bands de-

ending on country regulations 863ndash868 MHz in Europe 902ndash

28 MHz in the US and 9165ndash9275 MHz in Japan China South Ko-

ea and Singapore also have specific channelisations Channel widths

f 1 MHz and 2 MHz have been adopted although 4 8 and 16 MHz

re also supported in some countries IEEE 80211ah furthermore pro-

oses new PHY and MAC layers The IEEE 80211ah PHY layer can be

onsidered to some extent a sub-1 GHz version of the IEEE 80211ac

ne At the physical layer OFDM is the chosen modulation method us-

ng 32 or 64 tonessub-carriers that are spaced by 3125 kHz The sup-

orted modulations include BPSK QPSK and from 16 to 256-QAM A

road range of antenna technologies ranging from single-user beam-

orming to MIMO and DL-MU-MIMO which was first introduced in

he IEEE 80211ac amendment are also included in the IEEE 80211ah

pecification Similarly the IEEE 80211ah MAC protocol include most

f IEEE 80211 main characteristics further extending its power sav-

ng (PS) mechanisms

11 Novel features

This section overviews the extensions introduced by the IEEE

0211ah amendment to the IEEE 80211 PS mechanisms to account

or the specific characteristics of resource-constrained sensor and ac-

uator devices A more detailed review can be found in [110] includ-

ng a performance assessment of IEEE 80211ah in several of the key

cenarios for M2M communications such as agriculture and animal

onitoring smart metering and industrial automation plants In ad-

ition a detailed survey of the IEEE 80211ah is reported in [112]

12 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 8 IEEE 80211ah PS mode for TIM-STAs non-TIM STAs and Unscheduled STAs

a

d

w

i

a

a

c

m

i

a

e

t

s

m

t

t

u

b

i

I

s

d

f

a

b

s

a

b

S

T

c

t

w

I

4

a

a

which addresses aspects not considered in [110] such as the use of

sectorisation to avoid overlapping between multiple IEEE 80211ah

WLANs and the use of relays to further extend the IEEE 80211ah cov-

erage among other aspects Finally in [113] both IEEE 80211ah PHY

and MAC characteristics are introduced with emphasis on the ben-

efits of using the sub-1 GHz ISM band in terms of link budget the

location of the OFDM pilots for outdoors operation the use of bidi-

rectional transmission opportunities and the new packet formats to

minimise overheads among others

Enhanced power saving mechanism PS mechanisms for WLANs

were already considered in the development of the first IEEE 80211

standard with the goal of improving the lifetime of battery equipped

devices [114] In PS mode STAs keep the transceiver in sleeping mode

as much time as possible They periodically wake up to listen to the

beacons transmitted by the AP Those beacons indicate whether an

STA has packets waiting for it at the AP In the positive case that STA

remains awake and requests the delivery of those packets Otherwise

given it has nothing to receive it returns to sleep mode until the next

beacon is expected

In the IEEE 80211ah amendment time is divided into pages DTIM

(Delivery Traffic Indication Map) periods TIM (Traffic Indication Map)

periods and slots DTIM and TIM periods begin with the correspond-

ing DTIM and TIM beacons sent by the AP The functions of DTIM and

TIM beacons are described below

1 DTIM beacons They inform as to which TIM Groups (ie the

group of STAs assigned to the same TIM period) have pending

packets at the AP

2 TIM beacons Each TIM message informs a TIM Group about

which specific STA has pending data in the AP Between two

consecutive DTIMs there are as many TIM beacons as TIM

Groups

Using this DTIMTIM-based approach any STA can enter into a

power saving state if it does not have packets pending for transmis-

sion and one of two conditions is met (1) it observes in the DTIM

beacon that there is no downlink traffic addressed to its TIM Group or

(2) it observes in the DTIM beacon that there is some downlink traffic

addressed to its TIM Group but that STA does not explicitly appear in

the TIM beacon Compared to the preliminary IEEE 80211 PS mecha-

nism this approach reduces the size of the Traffic Indication Map in

each TIM beacon thus reducing the overhead and the time STAs need

to listen and process them In addition TIM periods can be organised

in pages which further increases the number of TIM groups between

two DTIM beacons

The temporal organisation of pages DTIM and TIM periods is re-

ported in Fig 8 which shows the division of TIM periods in RAW (Re-

stricted Access Window) and PRAW (Periodic RAW) The RAW is a

time interval in each TIM period where TIM stations can transmit and

receive data (see below the different types of STAs defined in IEEE

80211ah) In addition it can be divided into several downlink and

uplink slots for further granularity In the downlink the slots are as-

signed to a single STA or a group of STAs while the uplink slots are

randomly selected by the STAs with packets ready for transmission

The PRAW is the period of time in each TIM where non-TIM stations

can transmit and receive data

Types of STAs IEEE 80211ah supports three types of STAs TIM

non-TIM and Unscheduled STAs A TIM station is assigned to a TIM

Group Their data transmissions must be performed within a RAW

(Restricted Access Window) period Non-TIM stations do not have to

listen to beacons to transmit data During the association process

non-TIM devices directly negotiate with the AP to obtain a trans-

mission time allocated in a PRAW The following channel access can

either be renegotiated or occurs periodically depending on the re-

quirements set by the STA Unscheduled STAs do not need to listen to

any beacons similar to non-TIM stations Even inside any restricted

access window unscheduled STAs can send a poll frame to the AP

sking for immediate access to the channel The response frame in-

icates an interval (outside both restricted access windows) during

hich unscheduled stations can access the channel This procedure

s meant for STAs that transmit data very sporadically

Hierarchical station organisation To support a large number of STAs

nd their organisation in pages DTIM and TIM periods IEEE 80211ah

ssigns to each associated STA a unique identifier of 13 bits which is

alled the Association Identifier (AID) Using this new AID the maxi-

um number of supported STAs is increased from the original 2007

n IEEE 80211 to 8191 (= 213 minus 1) in IEEE 80211ah However it also

llows categorising STAs according to the type of application they are

xecuting their power level or even their desired QoS by assigning

hem to different TIM groups

Long sleeping periods IEEE 80211ah offers TIM Non-TIM and Un-

cheduled STAs the possibility to set very long doze times (up to

onths) The corresponding clock drift produced by such long doze

imes must be taken into consideration however as the higher the

ime an STA has been asleep the further in advance it should wake

p to avoid possible synchronisation problems with the network

Efficient small data transmission Three new enhancements have

een proposed to reduce the overhead when the data packet size

s small First while IEEE 80211 contains a 28-byte MAC header

EEE 80211ah proposes a short 18-byte version by using AIDs in-

tead of MAC addresses Second IEEE 80211 has defined several null

ata packet (NDP) frames which consist only of a PHY header These

rames can be used to create short ACKs short Block ACKs short CTSs

nd short PS-Polls Finally a Fast Frame Exchange mechanism has

een developed so that if an STA has data to transmit it can notify a

uccessful reception by transmitting its data frame instead of an ACK

Sectorisation Since the PHY layer is based on the IEEE 80211ac

mendment single and multi-user beam-forming are also supported

y IEEE 80211ah This allows the transmission of data to multiple

TAs simultaneously in the downlink increasing the system capacity

he use of the beam-forming capability of IEEE 80211ah APs is also

onsidered as a means to group STAs into different independent an-

enna sectors with the main goal of reducing interference issues This

ould be particularly useful in the case of overlapping with other

EEE 80211ah WLANs or in the presence of hidden nodes

12 Open challenges

A first open challenge is to understand the coverage and achiev-

ble transmission rates in IEEE 80211ah WLANs in both indoor

nd outdoor scenarios Propagation models for WLANs working at

B Bellalta et al Computer Communications 75 (2016) 1ndash25 13

f

c

8

p

R

p

i

m

l

a

m

c

p

p

f

m

t

a

e

p

t

m

a

l

T

m

a

i

n

t

t

t

u

p

c

w

f

a

t

8

a

m

a

s

l

a

m

b

m

d

S

i

a

o

a

R

t

n

G

fi

s

t

p

t

c

u

c

w

i

o

p

g

a

c

i

f

c

(

d

m

l

i

R

t

l

t

L

m

a

m

o

e

f

a

I

i

t

w

i

t

a

c

w

p

f

t

d

5

i

s

e

s

w

5

a

c

a

n

requencies lower than 1 GHz are evaluated in [115] The authors

ompare two path loss propagation models proposed by the IEEE

0211ah Task Group (one for macro and one for picohotzone de-

loyments) [116] with Lee and HatandashOkumura propagation models

esults show that the IEEE 80211ah channel models underestimate

ath loss with respect to Lee and Hata models Moreover in a compar-

son with empirical data it is observed that the IEEE 80211ah channel

odels also underestimate the initial loss and the slope of the path-

oss function A new model parameterisation is thus proposed by the

uthors In [117] the feasibility of an IEEE 80211ah WLAN deploy-

ent is also evaluated in terms of the achievable range and bit rate

omputed on the basis of the link budget using the same path loss

ropagation models as in [115] Results show that the transmission

ower limitations in the uplink can limit the overall network per-

ormance Finally in [118] the authors evaluate the achievable trans-

ission range for the different transmission rates and the achievable

hroughput for different combination of transmission power values

nd transmission rates Further studies are required due to the het-

rogeneity of scenarios in which IEEE 80211ah WLANs can be de-

loyed to characterise the effective coverage ranges with special at-

ention to mixed indoors and outdoors scenarios and in presence of

obile nodes

Due to the wide coverage IEEE 80211ah WLANs may be severely

ffected by the presence of hidden terminals To mitigate such a prob-

em smart solutions to guarantee that the STAs assigned to the same

IM Group are inside each other transmission range or the employ-

ent of RTSCTS-based solutions may contribute to mitigate such

problem An overview of long-range scenarios for IEEE 80211ah

s presented in [119] including a detailed description of the hidden

ode problem in them

Single-hop uplink transmissions from distant STAs require high

ransmission power to reach the AP Since not all STAs may be able

o transmit at the required power a solution could be the introduc-

ion of nodes able to relay transmissions from them However the

se of relays has to be efficiently harmonised with the operation in

ower saving mode allowing temporal periods in which the relays

an gather the data from their associated STAs and then periods in

hich they forward the data to the AP which is still an open challenge

or IEEE 80211ah WLANs In case relays are used data prediction and

ggregation techniques could be implemented to make more efficient

ransmissions [120]

Another challenge is the assessment of the efficiency of IEEE

0211ah PS mechanisms including their optimisation In [111] the

uthors evaluate the capacity of the TIM and page segmentation

echanisms in terms of the maximum number of nodes supported

nd the energy consumed given a certain network traffic profile Re-

ults confirm that a large number of STAs can be supported with

ow energy consumption The impact of the number of nodes oper-

ting in PS mode on the energy consumption and the delay perfor-

ance is analysed in [121] A detailed analysis of the TIM Group-

ased channel access adopted by the IEEE 80211ah task group is

ade in [122] where the STAs assigned to each TIM are uniformly

istributed between the slots of the RAW Since only one group of

TAs is allowed to transmit in each RAW slot the channel contention

s minimised The number of downlink and uplink slots in a RAW

s well as their duration is optimised on the basis of the number

f active STAs and the network traffic profile in [123] In [124] the

uthors propose an algorithm to determine the optimal size of the

AW interval for uplink transmissions based on the estimation of

he number of STAs transmitting and the duration of the RAW Fi-

ally in [125] the authors address the contention problems in Smart

rid communication networks with many nodes and periodic traf-

c when using an IEEE 80211ah WLAN It is still a challenge to con-

ider smart systems able to adapt the IEEE 80211ah parameters to

he instantaneous system state in order to save as much energy as

ossible

Since the PS mechanisms affect only the operation of TIM STAs

he performance of non-TIM and Unscheduled STAs has not yet been

onsidered in the literature To identify the scenarios in which the

se of non-TIM and Unscheduled STAs are of interest is still an open

hallenge as further investigating how non-TIM STAs negotiate over

hich PRAWs they can use to transmit and receive data Co-existence

ssues between the three types of STAs in a single network is also an

pen challenge that requires further work to guarantee a minimum

erformance level for all of them

The design of strategies to distribute the STAs between all the TIM

roups based on their specific traffic profile position battery level

nd application priority is another important challenge that is still

ompletely open EDCA is the default channel access scheme included

n the IEEE 80211ah amendment and provides some basic traffic dif-

erentiation capabilities at the packet level Besides the different ac-

ess categories (ACs) needing to be renamed and their parameters

Arbitration Inter-Frame Spacing or AIFS CWmin TXOP duration) up-

ated to fit the specific traffic profiles of M2M communications other

echanisms can be implemented A mechanism to assign the down-

ink RAW slots when packets from multiple priority levels are wait-

ng for transmission at the AP is required for example Also uplink

AW slots are currently selected randomly by the STAs that want to

ransmit a packet A mechanism to reserve some slots at each up-

ink RAW for high priority STAs may therefore be an option although

his may severely degrade the performance of the low priority STAs

astly different repetition patterns for different priority TIM groups

ay also allow priority STAs assigned to those priority TIM groups to

ccess the channel more often We expect this challenge will receive

uch attention in the upcoming years because of the heterogeneity

f sensors that will be connected though a single IEEE 80211ah AP

specially in urban scenarios

Similar to what has been discussed for 4G cellular networks used

or M2M communications [126] the development of mechanisms to

void and resolve congestion situations is another open challenge for

EEE 80211ah WLANs with many nodes For instance a mechanism

n which STAs have to wait a random number of DTIM periods before

hey can start a transmission may be implemented This approach

ould effectively distribute the traffic over a longer period of time

n overload conditions but otherwise would unnecessarily increase

he access delay Adaptive and load-aware solutions may play again

n important role to keep the optimal WLAN operation in all possible

ircumstances

Finally since IEEE 80211ah will compete with 4G5G cellular net-

orks and WSNs to provide M2M connectivity in many different ap-

lication domains such as smart cities or e-health comparative per-

ormance studies to determine the strong and weak points of each

echnology are required including also aspects such as the cost of the

evices and the system reliability

Cognitive radio technology for TV White Spaces

This section gives an overview of the IEEE 80211af amendment

ntroducing its most relevant features and open challenges Fig 9

hows the basic components and features considered for WLANs op-

rating in the TV White Spaces (TVWS) which include local spectrum

ensing by the AP and STAs and the use of geolocation data bases

ith information on channel availability

1 The IEEE 80211af amendment

Thanks to the transition from analog TV to digital TV several VHF

nd UHF spectrum channels used for decades for analog TV broad-

asting are now unused or are under-utilised in many geographical

reas [127ndash129] This ldquodigital dividendrdquo of spectrum has suggested

ational regulators such as the FCC in the US and Ofcom in the UK

14 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 9 WLAN operating in TVWS basic elements and functionalities

Fig 10 Hidden terminal problem in TVWS when the secondary network uses only in-

stantaneous channel sensing to decide whether or not a TV channel is occupied When

the STA is not able to detect the TV signal due to the presence of obstacles it may

initiate a transmission and create interference to the primary users

s

c

5

a

i

p

t

o

o

t

e

c

u

l

m

b

a

m

t

e

t

t

t

m

D

c

I

n

b

u

8

c

a

t

i

t

w

f

to discuss how to reuse these channels for unlicensed devicesrsquo com-

munications [130131] These potentially vacant channels in the VHF

and UHF bands are referred to as TVWS [132] and include spectrum

portions like the 470ndash790 MHz in Europe and non-contiguous 54ndash72

76ndash88 MHz 174ndash216 MHz 470ndash698 MHz and 698ndash806 MHz in USA

A snapshot of the TV spectrum occupancy in the city of Barcelona in

2012 is shown in [133]

The attractive characteristics of TVWS (not only for WLANs) in-

clude the ability to penetrate through walls and other obstacles much

more effectively than other widely used spectrum bands such as the

24 and 57 GHz ISM bands [134ndash137] This fact along with a progres-

sive spectrum scarcity in ISM bands has suggested the birth of the

IEEE 80211af amendment published in February 2014 [138] which

provides the IEEE 80211 operational characteristics for TVWS access

of unlicensed White Space Devices (WSD) A good summary of IEEE

80211af can be found in [137] The main advantage of operating IEEE

80211 WLANs in the TVWS comes from an increased coverage range

which can reach up to 1 km in rural areas and open fields [139]

and less energy needed to transmit However this comes at the cost

of an increased interference risk to other WSDs which creates co-

existence problems and demands new PHY and MAC layers to ef-

ficiently support channel access and operations preserving licensed

usersrsquo devices

WLANs operated in the TVWS could cover a number of interest-

ing and emerging use cases and scenarios eg Internet access in ru-

ral or sparsely populated areas Smart Grid sensor aggregation me-

tering and control Internet of Things advanced WLAN operations

and TVWS traffic offloading in indoor environments [140141] Nev-

ertheless the concept could be realised to create or extend com-

mercial or municipality WiFi services offered to citizens with cover-

age at the whole city level realised with reasonably limited network

infrastructures [142143]

IEEE 80211af will use the PHY and MAC layers derived from IEEE

80211ac It will adopt concepts such as the OFDM multi-user beam-

forming contiguous and non-contiguous channel bonding and packet

aggregation Among the mandatory and innovative behavioural and

operational parameters the most notable one is the channel acqui-

sition support realised through remote geolocation-based spectrum

allocation databases which maintain the channelsrsquo availability infor-

mation in any given area and time of day providing upon request the

list of free channels available for use

In the following section we highlight and summarise three main

novelties that IEEE 80211af introduces In Section 511 we explain

how the access infrastructure to the remote spectrum database is de-

signed and what the requirements are we discuss the coexistence is-

sues and methodologies adopted in IEEE 80211af and we also discuss

the novel concept of non-contiguous channel bonding The following

ections also contain an illustration of related works Open research

hallenges will be summarised in Section 512

11 Novel features

This section provides an illustration of novel features introduced

nd discussed in the path to IEEE 80211af and the directions provided

n related works and standardisation initiatives to resolve classical

roblems properly characterised in the new framework of TVWS

echnologies

Channel acquisition spectrum database and channel sensing Much

f the attention in operating in the TVWS is given to the protection

f the primary licensed users in the TVWS spectrum band In general

he primary users were considered as the Digital TV (DTV) broadcast-

rs and receivers In fact receivers would suffer the wasteful effect of

ollisions during the TV broadcast reception in the case of secondary

sersrsquo (SU) transmission interference (Fig 10)

It is a well-known problem that the definition of a coherent wire-

ess channel status (idlebusy) from the viewpoint of distributed pri-

ary users (eg DTV receivers) is made difficult by the difference

etween transmitters and receivers interpretation of signals in a vari-

ble time-space collision domain resulting in hiddenexposed ter-

inals concepts Solutions proposed to reduce the hiddenexposed

erminals included the RTSCTS mechanisms and beaconing How-

ver it is difficult to implement solutions that would effectively de-

ermine the channel status of TVWS channels based on localised dis-

ributed sensing activities performed by multiple secondary users in

he same collision domain On the other hand the use of RTSCTS

echanism would not be effective in this case because most of the

TV primary users (and specifically the receivers) were not originally

onceived to transmit signals or implement the RTSCTS handshake

n conclusion the actions to identify free (and busy) TVWS chan-

els should not be based on primary usersrsquo involvement but should

e almost totally implemented by specifically designed secondary

sersrsquo methodologies

There has been much debate on the methodology that IEEE

0211af secondary devices should adopt to get knowledge of free

hannels at a given time on a given target communication area in

n effort to guarantee verified optimised and reliable TVWS spec-

rum use The three main methodologies discussed include sens-

ng solutions geolocation databases (DB) and beaconing Moreover

wo approaches are possible in general distributed and centralised

ith their well known tradeoffs in terms of effectiveness required in-

rastructure deployment and coordination overhead The distributed

B Bellalta et al Computer Communications 75 (2016) 1ndash25 15

m

b

s

c

t

u

a

t

o

l

t

w

t

t

a

o

c

c

a

(

t

o

t

t

D

t

t

o

t

f

a

t

c

i

a

c

d

t

s

t

b

I

l

(

e

a

i

R

p

e

t

c

c

i

b

i

i

t

b

t

t

t

W

a

p

c

i

t

h

p

t

n

a

t

e

t

t

t

p

t

a

r

8

w

s

w

[

i

S

o

a

t

f

i

8

i

r

a

a

i

t

c

(

I

s

a

e

t

t

c

[

i

c

i

g

s

n

s

t

i

o

o

s

c

D

o

ethodology that is receiving more attention in the literature is

ased on distributed channel sensing solutions [127144145] These

olutions include the observation of spectrum use and possible

ooperative aggregation of spectrum sensing information to increase

he accuracy of detection and reduce the vulnerability of primary

sersrsquo transmissions The spectrum sensing capability is required in

ll the secondary devices with a minus114 dBm sensitivity Dynamic

ransmission power control must be provided and the upper limit

n emissions is 100 mW EIRP (20 dBm) for portable devices (further

imited in the case of adjacent channel use to reduce out-of-band in-

erference) However database (DB) coordination of TVWS combined

ith spectrum sensing is considered the most promising and effec-

ive technique compared with spectrum sensing alone The DB spec-

rum information has more chances to be effective (ie identifying

ll free channels opportunities) and reliable (ie not prone to attacks

r erroneous interpretation of channels status) In fact regulators de-

ided to push for the conservative solution of a remotely accessible

entralised spectrum DB that maintains the information on the avail-

bility of all TVWS channels at any given point in time and location

with a target accuracy around 50 m) under their responsibility For

his reason the spectrum databases have become a mandatory part

f many spectrum sharing systems and the dominant technical solu-

ion to support TV white spaces

The IEEE 80211af standard adheres to this vision and also has

he role of a common regulatory framework for different spectrum

B implementations defining a generalised coordination architec-

ure protocols and interfaces for spectrum queries and local spec-

rum information management The protocol explicitly considers out

f scope the communication protocol and technology adopted in

he background leaving the freedom to those players responsible

or the deployment to select any suitable Internet-based and local

ccess network technology Secondary devices that need to access

he TVWS spectrum should get information on which channels they

an effectively use and which parameters to adopt without impact-

ng primary usersrsquo transmissions (and receptions) by querying the

vailable spectrum DB The query must contain the transmission

haracteristics and accurate geographical position of the secondary

evice which can be determined by means of a geo-positioning sys-

em (GPS) regularly updated in the case of movement (or manually

et for static devices) The IEEE 80211af reference system architec-

ure for DB spectrum access is composed of multiple Geolocation-

ased Spectrum information Databases (GDBs) entities connected via

nternet to Registered Location Secure Servers (RLSS) RLSSs work as

ocal proxies of the GDB for a localised group of Basic Service Sets

BSS) RLSS are connected via a secure protocol architecture with the

quivalent of multiple Access Points (AP) in different BSS realising

trusted DB infrastructure APs locally coordinate the exchange of

nformation and channel access management between the GDB (via

LSS) and the secondary usersrsquo end stations (STA)

Mode I devices are those under the control of a device that em-

loys geo-location database access while Mode II devices are those

mploying geo-location database access by themselves The informa-

ion exchange provided between the GDB and a secondary user STA

an be provided in both an open-loop (eg adopted by FCC) and a

losed-loop (eg adopted by ETSI) implementation In an open loop

mplementation daily spectrum availability information is provided

y the GDB and no feedback on the spectrum information received

s provided by STAs thus the system approach for spectrum access

s much more conservative leading to low channelsrsquo utilisation po-

ential In a closed loop implementation the STA can provide feed-

ack to the GDB and there are more communication overheads due

o the high granularity of updates however the system is more effec-

ive and reliable in the exploitation of TVWS on behalf of STAs The

ypical information provided by the GDB includes (1) the updated

hite Space Maps (WSM) of frequencies allowed for secondary use

t the timespace of the querying STA and (2) the device-dependent

ower limitations for transmission (in general conservative and ac-

urate enough to avoid relevant interference effects on primary users

dentified in the area)

The basic IEEE 80211af mechanisms regulating the communica-

ion between STA and GDB can be found in [137138]

In parallel with IEEE 80211af standards such as the IEEE 19006

ave been created that consider interfaces and data structures sup-

orting spectrum sensing information exchange applicable to spec-

rum sensing (and particularly distributed spectrum sensing) sce-

arios In particular in October 2014 the IEEE 19006 WG initiated

project for a new standard called IEEE 19006b [146] concerning

he use of spectrum sensing information to support and optimise the

ffectiveness reliability and robustness of spectrum database solu-

ions The aim is to enhance the performance and capabilities of spec-

rum databases through the use of spectrum sensing information

Co-existence Support for co-existence mechanisms so that mul-

iple technologies can effectively utilise the TVWS spectrum is im-

ortant Self co-existence between network devices of a common

echnology (eg deployed by different operators in the same area)

nd co-existence among different technologies are relevant topic of

esearch for Cognitive Radio (CR) systems and specifically for IEEE

0211af Many solutions appeared on the research scene but no one

as so far finalised as the target solution for IEEE 80211af Specific

tandardisation has been started to regulate coexistence between

ireless standards of unlicensed devices including the IEEE 802191

147] The purpose of the IEEE 802191 standard is to enable the fam-

ly of IEEE 802 Wireless Standards to most effectively use TV White

pace by providing standard coexistence methods among dissimilar

r independently operated TVWS devices Early examples of gener-

lised coexistence mechanisms included Dynamic Frequency Selec-

ion (DFS) Transmission Power Control (TPC) listen before talk (eg

or contention based IEEE 80211 80215) time division multiplex-

ng (also among different techniques such as the IEEE 80216 80220

0222) and Message-based Spectrum Contention (that is beacon-

ng messages that carry coexistence information) Opportune met-

ics must be defined to assess the measurable coexistence achieved

mong different technologies as an example the hidden node prob-

bility for a target scenario or the estimate of percentage variation

n normalised network throughput and latency (before and during

he SU transmissions) On the other hand a centralised coexistence

ontrol mechanism could be effectively realised by a central manager

or coexistence-DB like the GDB) in critical scenarios To this end

EEE 19004 [148] (a standard for heterogeneous networks in dynamic

pectrum context part of IEEE Standards Coordinating Committee 41)

ims to standardise the overall system architecture and information

xchange between the network and mobile devices which will allow

hese elements to optimally choose from available radio resources

There are three possible classification of co-existence archi-

ectures and inter-network coordination channels for CR systems

entralised coordinated and autonomous co-existence mechanisms

149] In centralised co-existence schemes the co-existence is admin-

stered by a central entity (eg like in IEEE 802191) This solution

ould be applied in both homogeneous IEEE 80211af systems (eg

ncarnated by the centralised spectrum DB entity) and in hetero-

eneous inter-networks without requiring modifications to existing

tandards under the assumption that all the involved co-existing

etwork entities would adhere to the same centralised coordination

cheme and inter-network coordination protocol On the other hand

his assumption could be hard to satisfy in some practical scenar-

os The centralised schemes are effective in providing minimisation

f inter-network interference based on the availability and quality

f centralised co-existence information In coordinated co-existence

chemes the centralised entity could be present but not taking de-

isions in general The central entity could provide a co-existence

B with required information which can be used via in-band and

ut-band signalling for co-existence decisions taken by the cluster

16 B Bellalta et al Computer Communications 75 (2016) 1ndash25

c

e

n

c

o

s

i

d

a

S

8

n

a

b

o

b

i

t

a

8

b

[

t

i

n

b

5

t

a

c

b

s

a

c

n

s

p

i

c

l

r

c

f

a

v

t

a

e

c

t

Q

a

b

s

i

c

u

n

c

i

heads of many co-existing inter-networks for implementing a proper

common coordination protocol (eg out-band busy tone signalling)

These are often hybrid solutions realising a compromise between ef-

fectiveness and suitability In autonomous co-existence schemes all the

decisions and coordination are implemented in a distributed way

by the involved network entities In these schemes many policies

can be adopted to realise a sufficient level of co-existence under a

best effort approach Solutions that could be implemented include

the use of busy tones beaconing and other signalling protocols in

dedicated control channels dynamic distributed frequency selection

schemes listen before transmit token-based and dynamic reserva-

tion schemes All the above mentioned solutions have goods and

bads and a general illustration can be found in [149] Other proac-

tive co-existence techniques try to early detect and recovermitigate

co-existence issues with relaxed inter-network coordination eg re-

alised via spectrum sensing and interference avoidancesuppression

techniques

It must be clear that different aims exist regarding the co-

existence for primary TVWS users protection against secondary

users and for secondary devices mutual interference avoidance

Many co-existence problems have been analysed in the literature

for TVWS technologies in particular between the primary (licensed)

and secondary (unlicensed) devices The coexistence management in

TVWS is complex because primary users of DVB transmissions are in-

tended to be the pure receivers rather than the broadcasters In other

words potential protected users could be everywhere and the hid-

den terminal problem could arise for these systems The co-existence

problem is also complicated by many factors that cause asymme-

try and dynamicity in the TVWS such as the mobility variable den-

sity power asymmetry and heterogeneous MACPHY layers TVWS

co-existence problems originate between the IEEE 80211af technolo-

gies and the IEEE 80222 technologies for wireless regional area net-

works (WRAN) Results of analysis of TVWS usage in Europe show

that white spaces are typically present and fragmented They are typ-

ically more abundant in rural areas where larger contiguous blocks

of unused channels are available due to broadcast network planning

giving priorities linked to population density However the exploita-

tion of TVWS in urban areas is possible eg some recent research

was realised under the assumption of exploiting shadowing effects

created in the communication environment (eg by buildings obsta-

cles) in favour of frequency reuse [150]

Since different standards for opportunistic communication in the

TV White Spaces have now been published such as IEEE 80222 [151]

IEEE 802154m [152] Weightless [153] and of course IEEE 80211af

improved methods to guarantee the coexistence of different devices

operating on several protocols in the same bands must be deployed

We highlight here some work eg [150154ndash157] that focuses on

the coexistence between different technologies operating in the same

bands In particular [155] studies the performance degradation of

IEEE 80222 when an IEEE 80211af network is operated in the same

area The problem becomes even worse when the IEEE 80211af net-

work is located near 80222 user equipment causing both networks

to perceive a strong interference from each other One proposed solu-

tion is the Coexistence Beacon Protocol [158] studied for 80222 net-

works which foresees the exchange of a periodic beacon to identify

the neighbouring and possibly interfering networks

For what concerns the co-existence problems research has ad-

dressed other important related aspects of new WLANs on TV White

Spaces For instance the higher coverage range makes them attrac-

tive for applications in the smart grid and also for M2M on TVWS

[134150159] Here the increased range compared to standard tech-

nologies in the ISM bands can have beneficial effects for indoor

mobile devices [150] and M2M [159] Specifically indoor devices

communicating on TV White Spaces offer better propagation char-

acteristics and penetration through obstacles making it easier to

realise scalable home connectivity [135] although interference and

o-existence problems could be exacerbated and must be resolved

g see [139] However studies in the literature show how the sig-

al coming from an indoor transmitter on TV White Spaces remains

onstrained inside the house making it difficult to deploy indoor-to-

utdoor networks On the other hand this shadowing limit can be

een as an opportunity lowering the interference outside the build-

ng in which the TVWS network is locally operated [134150] Un-

er this approach HDTV streaming has recently been considered as

possible use case of WLAN networks operating in the TV White

pace [160]

Non-contiguous channel bonding A relevant novel feature of IEEE

0211af is the potential for contiguous and non-contiguous chan-

el bonding which permits aggregating basic channels (also non-

djacent ones) and leveraging the possible large frequency spread

etween multiple available channels Due to the rather static nature

f primary DTV transmissions where a busy channel is unlikely to

ecome free in the near future and state changes are coarse-grained

n general it is crucial to exploit the time-locality effect and exploit

he maximum physical channel availability that could be aggregated

t any given location With the methodology inherited from IEEE

0211ac IEEE 80211af is capable of bonding together two up to four

asic channels grouped in up to two different non-contiguous chunks

137] The spectrum bandwidth of a DVB-T basic channel can be ei-

her 6 7 or 8 MHz depending on the country in which the service

s operated As an example this creates a 144 to 168 OFDM chan-

elsrsquo bandwidth potential when up to four 6ndash7ndash8 MHz channels are

onded [137]

12 Open challenges

The deployment of WLANs in the TVWS at their maximum poten-

ial still requires the resolution of open problems Generally speaking

s mentioned in previous sections the challenges for such networks

an be divided into three main categories (1) spectrum sharing

etween opportunistic (secondary) and heterogeneous devices (2)

pectrum sensingmanagement and maximum exploitation of avail-

ble spectrum (potentially enabling the spectrum-on-demand con-

ept) and (3) co-existence and interference mitigation to the primary

etwork (primary user protection) and secondary networks [154]

The overall aim of spectrum sharing techniques is to maximise

eparation to avoid overlapping or contiguous channels operations

rovided that sufficient channels are available in the TVWS spectrum

n a given spacetime scenario [154] In general important research

ontributions have to be realised for the design of proper spectrum al-

ocation methodologies and techniques satisfying multi-factorial QoS

equirements at the system-level and the user-level Novel ideas in-

lude the spectrum sharing and spectrum sensing techniques being

urther divided into cooperative and non-cooperative

Non-cooperative solutions attempt to realise spectrum sharing

nd sensing on a local basis without direct cooperation between de-

ices Examples include the ldquolisten-before-transmitrdquo approach and

ransmission power control to limit the interference and spectrum

llocation policies that are based on local node feedback only Coop-

rative solutions rely on the existence of a common communication

hannel through which devices can tentatively agree on the spec-

rum allocation that provides the desired spectrum separation and

oS Major issues include the definition of a methodology to identify

common channel (or multiple channels mutually shared in space

y pairs of heterogeneous and distributed devices) and to efficiently

hare common information For the access to the spectrum DBs the

dentification of proper common access channels is still under dis-

ussion out-of-TVWS-band cellular communications can be widely

sed given their high coverage also indoor

Trying to avoid the use of complex and resource-hungry coordi-

ation schemes and dedicated control channels rendezvous proto-

ols play a significant role both for co-existence and for establish-

ng dynamic communication channels based on spectrum sharing

B Bellalta et al Computer Communications 75 (2016) 1ndash25 17

R

n

c

t

l

o

m

e

n

i

n

w

n

a

u

s

p

b

s

m

s

a

d

t

b

a

a

m

a

t

u

8

n

t

d

d

m

f

t

n

d

n

e

r

a

h

b

f

a

c

t

u

i

f

t

d

t

s

i

f

I

l

i

a

n

c

d

I

n

o

t

i

t

e

w

t

c

m

t

a

i

t

q

b

c

a

i

l

i

a

v

t

a

a

c

t

i

t

b

t

t

a

D

p

c

s

o

q

r

p

a

g

t

e

u

d

t

m

n

t

s

f

i

1

8

f

T

endezvous is considered a fundamental problem in generalised CR

etworks at the basis of many communication processes eg in-

luding neighbour discovery routing and broadcast Rendezvous pro-

ocols attempt to establish a new link for communication on a se-

ected and agreed frequency band (channel) identified within a set

f available resources (channels) thus creating the basis for com-

unication between two or more SUs This problem is often exac-

rbated by the high (and unknown) numbers of contending SUs high

umber of available channels and their ambiguous boundaries and

dentifiers and tough co-existence requirements satisfaction in cog-

itive radio systems [161] Preliminary approaches for rendezvous

ere based on the existence of a dedicated Common Control Chan-

el (CCC) agreed among all the SUs or the existence of a centralised

gent (decision maker) assigning channels to any pairs of requesting

sers This approach could be considered in the case of IEEE 80211af

olutions based on the spectrum DB implementation however the

roblem is complicated by the fact that the common channel must

e identified and made available for all the requesting SUs at the

ame time even when these are spread over multiple collision do-

ains in space On the other hand the scalability and reliability is-

ues which are caused by the congestion of CCCs and the vulner-

bility risks of centralised DBs motivate the research of alternative

istributed solutions called blind rendezvous protocols [162] Dis-

ributed solutions have been mainly based on beaconing andor slot-

ased channel hopping algorithms exploiting randomisation sensing

nd discoverysynchronisation messages spread in slotted time over

set of candidate available channels [161] These protocols aim to

inimise the convergence time to maximise the use of resources

nd to avoid collisions among SUs by identifying and agreeing on

he use of a given common channel among N non-overlapping and

nambiguously labelled channels The main complication in IEEE

0211af systems is given by the wide availability of many different

on-contiguous spectrum bands which cannot be uniformly quan-

ised in single channels and unambiguously labelled in space in or-

er to realise a common reference domain for all the IEEE 80211af

evices In fact the knowledge of the number of channels their com-

on labels the unique IDs of SUs and the number of SUs contending

or the channel assignment are the key factors determining the effec-

iveness and good properties of blind rendezvous protocols Recently

ew classes of rendezvous protocols have been proposed which are

istributed blind and oblivious Oblivious means that available chan-

els may be identified with different labels on behalf of the differ-

nt involved SUs A number of these distributed blind and oblivious

endezvous protocols have been proposed and have been referenced

nd analysed in [162] To conclude solutions for dynamic spectrum

andoff issues (eg in case PUs suddenly appear in a region) must

e resolved in advance to minimise the latency of re-establishing a

unctional communication channel eg by implementing a proactive

lternative rendezvous channel selection in background to ongoing

ommunication processes [163]

Another challenge is to provide cooperative techniques in a dis-

ributed vs centralised implementation by analysing their convergence

nder variable conditions overheads and tradeoffs In some cases as

n vehicular networks the exploitation of scenario characteristics and

actors such as the constrained mobility could help to realise effec-

ive dissemination of sensing information over an extended and pre-

ictable horizon under a cooperative approach [164ndash166] Coopera-

ion could be helpful in the realisation of the cooperative spectrum

ensing in combination with the Geolocation DB approach Another

nteresting direction of research is the investigation of mutual ef-

ects of coexisting cooperation-based vs non-cooperative techniques

n cooperation-based devices the exchange of information could al-

ow a quick convergence to a solution and avoidance of further

nterference [145] Certainly cooperative techniques have desirable

dvantages over non-cooperative ones However tight time synchro-

isation and high coordination overheads are required The common

hannel for cooperation purposes could become a bottleneck and re-

uce the potential advantages of dynamic spectrum allocation The

EEE 80219 foresees the presence of a shared common control chan-

el (CCC) to which devices need to tune in order to gather the state

f the network Solutions such as [141] propose the use of TVWS for

he CCC

A promising direction for research is the adoption of cluster-

ng schemes for secondary devices associated with a hybrid dis-

ributedcooperative vs centralisednon-cooperative approach in an

ffort to reduce overheads while maximising advantages In this

ay secondary nodes could implement cooperation for sensing func-

ions using properly identified common channels and delegate the

entralised decisions and spectrum-DB management to well instru-

ented cluster-leader nodes This concept is incarnated to some ex-

ent by RLSS in IEEE 80211af More recently there have been propos-

ls to build distributed databases that refresh their contents period-

cally by querying the remote spectrum database [167] Here Mas-

er devices (according to Ofcom terminology) periodically cache the

uery replies from the spectrum database in order to reply to and

roadcast the spectrum availability to neighbour Slave devices

Another challenging direction for research is the definition of ac-

urate models and efficient simulation tools enabling dynamic spectrum

nalysis for both rural and urban areas Recent developments of dig-

tal maps simulation tools and propagation models theoretically al-

ow improving the capability of predicting radio propagation effects

n complex scenarios with an acceptable computation time This en-

bling technology could be used in parallel to spectrum DBs to pro-

ide more accurate forecasting of dynamic frequency allocation in

imespace scenarios Another issue to be taken into account is the

ccuracy of the remote spectrum database regarding the channel

vailability estimation Several works have already shown the inac-

uracy of propagation models due to the complexity of parameters

o be considered and differences between modelled and real scenar-

os [132134168169] Better propagation models will make it possible

o more efficiently estimate the interference between devices and

uild Radio Environment Maps (REM) or White Space Maps (WSM)

hat will possibly lead to a more accurate sharing of the radio spec-

rum [144] To this end current research could focus on how to lever-

ge simulation models in order to build more accurate frequency

Bs

In general the adoption of remote spectrum DBs methodology

ushed by national regulators has made the spectrum sensing pro-

ess optional and conservatively realised (eg power control for

ensing-only devices is limited to below 17 dBm EIRP) Thus much

f the research has focused on techniques and solutions to successfully

uery the remote database also in challenging environments such as

ural areas and indoor However since the queries must provide the

osition of the mobile device the accuracy of positioning for mobile

nd handheld devices still needs more appropriate solutions Trian-

ulation can be used to estimate the position of mobile devices with

he accuracy needed by the current regulations outdoor (50 m) How-

ver a low GPS accuracy is possible indoors [170] Certified man-

al positioning can be provided by technicians for static and indoor

evices although limiting mobility Alternatives to costly infrastruc-

ures for positioning (eg based on short range beaconing devices)

ust be deployed in indoor scenarios

Finally regarding spectrum sensing and interference mitigation

ew challenges come from the heterogeneity and different charac-

eristics that networks operating in the TV White Spaces can have

uch as different transmitting power (4 W for static devices 100 mW

or mobile and portable devices and 40 mW for communication

n channels adjacent to occupied ones) different bandwidths (5

0 15 and 20 MHz are currently foreseen for IEEE 80211af IEEE

02154m can have narrower bandwidths or wider ones) and dif-

erent medium access schemes (protocols that use either CSMA or

DMA are required to co-exist in the TV White Spaces) With all this

18 B Bellalta et al Computer Communications 75 (2016) 1ndash25

d

a

T

i

a

c

t

l

b

o

t

a

d

p

a

r

p

n

a

m

t

p

p

d

d

o

t

i

D

w

p

i

a

a

d

b

fi

d

o

t

l

b

t

b

a

i

e

o

o

c

t

t

w

A

c

M

s

c

l

O

w

g

v

t

in place the problem of determining whether another secondary de-

vice is currently transmitting in the same channel is still an open

issue

Spectrum sensing is a very hot topic in the literature with many

proposals that leverage the cooperation between devices to raise the

accuracy of the detection process eg [144171] The current litera-

ture also offers room for proposals that suggest exploiting sensing to

build more accurate spectrum DBs [164172173] Lastly regarding the

interference mitigation to the primary network regulators provide a

conservative approach based on a remote spectrum database to avoid

high risk for the primary users However in particular for the indoor

scenario the concept of secondary usersrsquo transmission in TV Grey

Spaces has been proposed TV Grey Space identifies busy channels

that are formally used by primary users in the area of interest (eg

at the rooftop level of a building for DVB-T) but which could be con-

sidered usable indoors (eg at the basement level) without causing

effective interference to the primary users [134150174] This concept

is based on the generalised spectrum underlay paradigm [175] The

remaining research challenge is to provide TV Grey space access guar-

anteeing a sufficient amount of protection to the primary user while

providing a satisfactory QoS to opportunistic devices eg through

the use of customised and validated propagation models for indoor

TVWS networks and feedback mechanisms to constantly monitor the

interference generated to the primary network

6 Emerging new trends and technologies

In this last section we review three emerging trends that in our

opinion will have a large influence in the conception of future WLANs

as they change the way WLAN protocols and functionalities are de-

veloped implemented tested and integrated with other wireless

networks

61 Programmable Wireless LANs

Especially in the enterprise environment WLAN deployments

need to support a wide range of functionalities and services This is

intrinsically difficult because of the large number of APs that must be

managed which calls for scalable solutions Typical services include

channel assignment load balancing among APs authentication au-

thorisation and accounting (AAA) policy management support for

client mobility and interference coordination Another problem is the

fact that WLAN clients autonomously take several decisions such as

which APs to associate with when to hand-over etc Therefore sup-

porting roaming clients requires the management of a large number

of association states across several APs which is a challenge if sup-

port for real-time hand-over is desired Typically such management

schemes are centralised and most of them are proprietary such as

WLAN controller solutions from Aruba [176] or Cisco [177] although

the IEEE 80211u amendment has been released to allow mobile users

to seamless roam between WiFi networks with automatic authenti-

cation and handoff [178] For example Dyson [179] enables STAs to

send information such as radio channel conditions to a centralised

controller based on a custom API (eg based on Python) As the con-

troller has a centralised view of the network it can enforce a rich

set of policies to control the network also using historical informa-

tion A demo system has been implemented along with applications

such as airtime reservations for specific clients or optimised hand-

offs However Dyson requires STAs to be modified in order to use

those new services offered by the centralised controller TRANTOR

[180] is another example of a centralised management system that

requires changes of clients in order for the infrastructure to gather

information from them in terms of eg interference measurements

In addition control commands enable the infrastructure to exercise

control such as modifying the transmit power or influencing the as-

sociation procedure Clients still use standard CSMA MAC layer for

ata transmission In contrast to DenseAP [181] or DIRAC [182] which

re also based on a centralised controller but do not modify clients

RANTOR exercises much more control on clients by the use of a ded-

cated API which enables a significantly higher gain in coordination

nd thus capacity CENTAUR [183] is another example of a centralised

ontroller that centrally schedules hidden and exposed terminals in

he downlink while using standard CSMA MAC for uplink traffic and

egacy downlink traffic To achieve good performance it uses a fixed

ack-off packet staggering techniques and a hybrid data path that

nly schedules downlink transmissions towards hidden and exposed

erminals centrally All other traffic is sent using standard DCF with

standard back-off procedure In contrast to previous approaches it

oes not require any modifications to the clients but does require data

lane centralisation and it remains questionable how scalable such

solution is for large WLAN deployments with high PHY layer data

ates beyond 100 Mbps However new approaches for fast data plane

rocessing that are currently explored in the virtualisation commu-

ity which move all the packet processing into user space may be

n option to significantly increase speed Designing such manage-

ent systems raises several interesting research questions (a) Cen-

ralised infrastructures may incur high latency but are more simple to

rogram and maintain while designing a completely distributed ap-

roach that operates close to WLAN devices may result in significant

istributed coordination and consistency problems (b) what level of

istributed control is required in order to support the flexibility needs

f future WiFi based networks supporting a high level of mobility

Traditionally WLAN APs are built on proprietary operating sys-

ems that are tightly coupled with the hardware This design makes

t hard to create new applications on top of such networking devices

espite the fact that protocols and mechanisms are available that

ould greatly improve the utility of existing networks those new

rotocols are not deployed because the closed system design makes

t very difficult to extend their functionality An important aspect for

ll such new approaches is therefore how to provide open interfaces

nd open source to speed up innovation As an example Linux-based

evices are completely open source but in order to increase flexi-

ility much more work is needed in the area of open drivers and

rmware In fact the Atheros based ath9k driver has been the main

river of innovation in the open source community because it is the

nly driver that interworks with open firmware In general solutions

hat modify the OS driver focus mainly on programmable network

evel solutions for WLANs such as channel switching or handovers

etween APs While the ath9k driver is an excellent example in how

o speed up innovation it is still very cumbersome to support flexi-

le MAC engine reprogrammability with ath9k In contrast MAClets

nd Wireless MAC processors (WMP) [184] allow a much more flex-

ble reprogramming of MAC functionalities A MAC processor is an

ntity able to execute general MAC commands that specify the MAC

perations through a software-defined state machine The behaviour

f the MAC protocol can therefore be updated at run-time by simply

hanging the sequence in which those commands are executed (ie

he MAClets) As a proof-of-concept the authors of [184] implement

he proposed MAC processor solution in a commodity WLAN hard-

are card extending the basic DCF in three directions piggy-backing

CKs a pseudo TDMA and the use of multiple channels In [185] a

ontrol framework for this WMP system is also proposed to support

AClet code mobility ie for moving loading and activating MAC

oftware programs embedded into ordinary data packets (akin to the

apsule model of traditional active networks)

The difficulty in re-programming networking hardware has also

ed to the concept of Software Defined Networks (SDNs) based on the

penFlow protocol [186] The main idea of SDNs is to extend net-

orking devices with standardised APIs that allow third-party pro-

rammers to flexibly control the data path In addition SDNs pro-

ide higher level abstractions to network designers and programmers

hrough the use of a centralised control plane offered by a network

B Bellalta et al Computer Communications 75 (2016) 1ndash25 19

Fig 11 The CLOUDMAC architecture VAP stands for Virtual Access Point and WT

stands for Wireless Termination

c

a

c

t

m

W

n

t

t

c

p

t

u

h

a

i

e

s

o

c

a

a

f

T

r

s

a

s

p

o

M

C

i

T

t

fl

A

a

a

i

c

t

s

t

e

I

t

w

H

e

b

I

l

d

o

6

r

W

s

l

a

e

e

a

s

b

c

t

(

S

p

c

F

B

t

p

T

e

t

s

u

m

i

e

o

a

A

p

d

t

t

n

p

i

t

t

f

i

ontroller such as NOX [187] which allows reuse of components such

s topology discovery or network access control for different appli-

ations The SDN concept has recently been applied to WLAN archi-

ecture in order to enable fine grained control over mobility manage-

ent and data forwarding focusing on a programmable enterprise

LAN architecture An important part of the SDN architecture is the

etwork controller which provides a centralised view of (parts of)

he SDN enabled network and uses the OpenFlow protocol to install

he forwarding rules on SDN-enabled devices (routers switches ac-

ess points cellular base stations etc)

ODIN [188] is designed to support programmability in enter-

rise WLAN architecture by separating the association state from

he physical AP They implement Light Virtual Access Points (LVAP)

sing a Split-MAC approach where the infrastructure controls the

andover procedure among different WLAN APs By managing

ssociations through SDN controllers ODIN enables proactive mobil-

ty management and load balancing within the SDN enabled WLAN

nterprise network without the need for changes in the client WLAN

tack or IEEE 80211 MAC layer While ODIN requires agents to reside

n the APs that communicate with the Odin Master within the SDN

ontroller CLOUDMAC [189] is based purely on the concept of SDN

nd virtual APs Similar to ODIN CLOUDMAC implements a Split-MAC

pproach but in addition enables the processing of WLAN MAC layer

rames within a co-located Cloud using so-called Virtual APs (VAPs)

he physical APs in CLOUDMAC are lightweight WLAN APs that are

esponsible only for sending out IEEE 80211 based MAC layer ACKs to

tandard WLAN clients and tunnel WLAN MAC layer frames through

n SDN-enabled enterprise WLAN towards the VAPs As association

tates are kept in the VAPs fast mobility is supported using sim-

le OpenFlow forwarding rules Because of the additional processing

f IEEE 80211 WLAN MAC frames in the co-located Cloud CLOUD-

AC has higher latency than a standard WLAN deployment However

LOUDMAC offers a Webservice based API to third party applications

n order to program the enterprise WLAN and enable new services

he architecture of CLOUDMAC is depicted in Fig 11 It has been ex-

ended in [190] to support QoS management and in [191] to support

exible MAC management frame prioritisation based on IEEE 80211

system based on OpenFlow has been recently proposed in [192] to

llow the station to be associated with multiple APs simultaneously

nd to switch between APs with low overhead

An important aspect to consider in order to enable programmabil-

ty is backwards compatibility Several approaches (eg [180]) require

lients to be modified to utilise the features provided This is difficult

o do in practice because it requires changes in the operating system

oftware of all clients If not all clients can utilise those APIs it is ques-

ionable how usable the new architecture will be and how much ben-

fit in terms of aggregate performance such architecture will allow

n contrast the SDN based approaches are interesting in the sense

hat they do not require changes in the client WLAN stack and work

ith the standard IEEE 80211 MAC layer deployed within the clients

owever it remains questionable how scalable such solutions are For

xample an interesting open research topic is to evaluate the scala-

ility of approaches such as CLOUDMAC [189] In addition processing

EEE 80211 MAC frames within co-located private Cloud requires low

atency support from local Cloud solutions (such as OpenStack) in or-

er to reduce the IEEE 80211 MAC processing time which is an area

f active research

2 Prototyping and testing IEEE 80211 enhancements

Most of the new proposals for next-generation WLANs are cur-

ently only evaluated using mathematical analysis and simulation

hile both analysis and simulation are necessary to characterise and

tudy those enhancements in the initial design phase or to consider

arge-scale scenarios it is difficult to consider all practical aspects of

real-world scenario This can sometimes cause significant differ-

nces between what simulations and real experiments show How-

ver real experiments are challenging because of the high complexity

nd costs of building the new hardware and software for each specific

olution to test

To mitigate implementation complexity there are several flexi-

le hardware platforms where low-level MAC mechanisms can be

ompletely implemented in software Examples based on FPGAs are

he USRP (Universal Software Radio Peripheral) [193] and the WARP

Wireless Open-Access Research Platform) [194] As an alternative

ORA (Microsoft Research Software Radio) [195] works on general

urpose computers by taking some advantage from modern multiple

ore systems

A different approach is provided by OpenFWWF [196] Open-

WWF provides an open CSMACA firmware for specific models of

roadcom chipsets so the resulting firmware can be uploaded and

ested in real commercial hardware OpenFWWF implements a sim-

le State Machine (SM) for controlling the hardware in real time

he SM evolution is driven by a main loop that reacts to events by

xecuting specific handlers When a packet originally prepared by

he Linux kernel is ready in the NIC memory handler Packet_Ready

ets up the radio hardware according to the packet metadata sched-

les the transmission and jumps back to the main loop The Trans-

ission Engine (TXE) then takes care of accessing the channel ie

t decrements the back-off counter according to the DCF rules and

ventually starts the actual transmission This triggers the execution

f the TX_frame_now event that prepares the ACK time-out clock

nd finalises the MAC header If the ACK-frame is received or if the

CK time-out expires and the maximum number of attempts for this

acket is reached handler Update_Params resets the contention win-

ow to its minimum value or otherwise doubles it Finally it loads

he back-off counter with a fresh value

In the following we will review a list of selected papers that use

he OpenFWWF firmware and the WARP or USRP platforms to test

ew proposals for WLANs in real scenarios as examples A first im-

lementation of the groupcast mechanisms defined in IEEE 80211aa

s presented in [197] The authors modify the OpenFWWF firmware

o include those functionalities and evaluate them using a 30 STAs

estbed A second example of the use of OpenFWWF is the collision-

ree MAC protocol presented in [198] In [198] the authors design and

mplement in OpenFWWF a MAC protocol that is able to achieve a

20 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 12 LTE to WIFI offloading

a

n

c

t

[

d

a

W

v

t

t

o

b

f

S

g

l

i

I

s

I

b

n

s

m

a

p

d

c

N

o

v

t

t

n

l

i

a

t

a

collision-free operation by waiting a deterministic time after success-

ful transmissions before to start a new transmission This work shows

the experimental performance of a collision-free MAC (CF-MAC) pro-

tocol for WLANs using commercial hardware The results show that

the proposed CF-MAC protocol leads to a better distribution of the

available bandwidth among users a higher throughput and lower

losses than DCF

In [199] the authors use the WARP platform to test a variant of the

DCF ndash called IEEE 80211ec ndash that substitutes control packets such as

the RTS CTS and ACK with short detectable sequences Since those se-

quences are shorter than the control packets and can be detected cor-

rectly even at lower SNRs values a significant gain in performance is

achieved MIDAS (Multiple-Input Distributed Antenna Systems) [200]

implementation using the WARP platform shows the benefits of dis-

tributing the antennas over the area to cover instead of co-locating

them at the AP in terms of capacity when MU-MIMO is employed

The authors also propose a new MAC protocol to benefit from the

spatial reuse that the DAS allows taking as a basis the IEEE 80211ac

Experimental results show that their proposal is able to achieve up

to 200 gains versus the traditional approach where all antennas are

co-located at the AP

The USRP platform has recently been proposed to develop a first

implementation of IEEE 80211ah in order to obtain a preliminary ex-

perimental performance assessment of such a technology because

no commercial off-the-shelf hardware is yet available as the amend-

ment is still in progress [201] The USRP platform is also used in

[202] to evaluate TIMO (Technology Independent Multi-Output) a so-

lution to deal with high-power non-IEEE 80211 interferers in ISM

bands

To conclude this section it is worth mentioning here also to

MAClets and WMP as given such flexibility to implement and dis-

tribute new MAC protocols and other IEEE 80211 functionalities we

believe that if a use-friendly implementation of such a MAC proces-

sors framework is provided it would significantly contribute to the

development and testing of new MAC enhancements by the research

community

63 CellularWLAN interworking

Public hotspots that offer Internet access over a WLAN using IEEE

80211 technology are now nearly ubiquitous It is forecasted that the

cumulative installed base of WiFi hotspots worldwide will amount

to 551 million by 2018 excluding private hotspots (eg WiFi access

points deployed at home) [203] The sharp increase in the availabil-

ity of public WiFi was initially perceived by mobile cellular operators

as a threat due to the additional competition from wireline Internet

service providers or emerging crowdsourced WiFi networks such as

FON5 However as cellular operators are fighting to cope with the ex-

plosion of mobile data traffic created by the rising use of multimedia

content traffic over mobile devices [15] they are also starting to use

WLANs based on the IEEE 80211 technology to offload data from their

core and access networks In general mobile data offloading refers to

the use of complementary network technologies (in licensed or un-

licensed spectrum) for delivering data originally targeted to cellular

networks Intuitively the simplest type of offloading consists of ex-

ploiting connectivity to existing co-located WiFi networks and trans-

ferring data without any delay (Fig 12) Thus this offloading tech-

nique is known as on-the-spot offloading As a consequence of this

new trend the seamless integration of cellular (eg 3GLTE) and WiFi

technologies has attracted significant research interest in recent years

(see [204] for a survey) a few solutions have already been standard-

ised [205206] and roaming between cellular and WiFi is becoming

increasingly transparent to end users CellularWLAN interworking is

5 httpscorpfoncomen

fl

t

f

lso fostered by the support in the evolving 4G standards of heteroge-

eous network deployments (HetNets) in which the existing macro

ells are complemented with a number of small low-power base sta-

ions with the goal of increasing capacity in highly congested areas

207208] It is envisaged that small cells will be based on 4G stan-

ards (eg pico and femto cells) as well as IEEE 80211 technologies

nd multimode base stations that work simultaneously with LTE and

iFi are already entering the market

It is important to point out that LTE standards already support a

ariety of mechanisms that enable data offloading However most of

he existing solutions such as Local IP Access (LIPA) and Selected In-

ernet IP Traffic Offload (SIPTO) focus on data offloading at the core

f the network [205] For instance LIPA allows an IP-enabled mo-

ile device to transfer data to another device in the same pico or

emto cell without passing through the cellular access network while

IPTO enables the routing of selected IP data flows through different

ateways The only LTE offloading mechanism that supports seam-

ess interworking with IEEE 80211-based WLANs is IP Flow Mobil-

ty and Seamless Offload (IFOM) Specifically IFOM relies on Mobile

Pv6 technologies to allow a user terminal to simultaneously route

elected IP flows over different radio access technologies [209210]

n this way a user terminal can offload selected flows to a WLAN

ased on some operator-defined policy while keeping the LTE con-

ection running However to enable such an approach it is neces-

ary to support an entity in the cellular core network that can com-

unicate with the user terminal to exchange information about the

vailability and quality of neighbouring access networks as well as to

rovide the user terminal with predefined rules to manage the han-

over process This entity in current LTE standards is called the Ac-

ess Network Discovery and Selection Function (ANDSF) server [211]

ote that the problem of selecting the best communication technol-

gy in an heterogeneous wireless network has been extensively in-

estigated in the literature both with centralised as well as decen-

ralised approaches (see [212] for a survey) However the use of mul-

iple interfaces in parallel as well as per-flow offloading are relatively

ew concepts Thus the design of scalable and efficient network se-

ection strategies for the ANDSF framework is still an open issue For

nstance three offloading methods suitable for the ANDSF framework

re proposed in [213] which are based on coverage SNR and sys-

em load A reinforcement learning approach is designed in [214] that

llows multimode base stations to autonomously steer their traffic

ows across different access technologies depending on the traffic

ype the usersrsquo QoS requirements the network load and the inter-

erence levels A number of studies have also tried to quantify the

8 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 4 Multiuser uplink transmission using two different OFDM subchannels Trans-

missions from STA B and C over the same subchannel represent an uplink MU-MIMO

transmission All transmissions are acknowledged with a single Multiuser ACK The

trigger frame is used to signal the selected STAs

t

t

t

a

s

t

f

f

c

3

s

s

t

n

n

A

d

w

t

fi

a

t

8

i

d

t

r

w

d

R

v

f

s

d

p

B

c

e

k

r

G

a

c

m

t

b

d

3 The concealment address protects legacy stations ie GCR-incapable stations

from receiving duplicated group-addressed frames

The challenge is that both the AP and the STA involved in a full-duplex

transmission have to start to transmit at the same time To achieve

that information about full duplex transmission capabilities can be

included in RTSCTS control packets to set-up a full-duplex transmis-

sion between the transmission initiator and its destination

Spatial reuse Dense WLAN deployments are necessary to offer a

continuous coverage with high transmission rates To improve both

the co-existence with those neighboring networks and the spatial

reuse of the spectrum a WLAN has two options (i) minimise its

area of influence by reducing its transmit power and (ii) accept

higher interference levels by increasing the Clear Channel Assess-

ment (CCA) level Use of both techniques may increase the number of

concurrent transmissions between neighbouring WLANs and there-

fore their capacity although it may also result in the opposite effect

since the achievable transmission rates may be negatively affected by

the higher interference levels observed which is the main challenge

to solve

Due the high WLAN dynamics the use of adaptive systems is cru-

cial but challenging as adaptivity requires extra complexity in terms

of computing and memory resources and there are no guarantees

that the implemented solution converges due to the decentralised

operation of each WLAN The use of DSC (Dynamic Sensitivity Con-

trol) to dynamically adjust the CCA level is one of the aspects cur-

rently under discussion in the IEEE 80211ax Task Group First studies

evaluating the performance of DSC for IEEE 80211ax WLANs show

a clear improvement on the spatial reuse and the area throughput

[61] Another example of the achievable throughput gains obtained

by adapting the CCA level can be found in [62] where the authors

show that gains of up to 100 can be achieved Moreover trans-

mit Power Control (TPC) to mitigate interference between WLANs in

dense scenarios is studied in [63] showing the need of jointly opti-

mising both TPC and CCA to maximise the network performance

Finally sectorisation by using beamforming is also under consid-

eration for the development of the IEEE 80211ax amendment as a

potential solution to improve spatial reuse [64] Using sectorisation

only the nodes of a given area are allowed to receive or transmit data

hence reducing the contention between different networks whenever

they activate non-overlapping sectors A challenge here is to coor-

dinate the different neighboring APs when they belong to different

administration domains Decentralised learning approaches may be

implemented to find feasible temporal patterns of non-overlapping

sectors

33 The IEEE 80211aa amendment

As discussed above legacy IEEE 80211 standards do not provide

robust and efficient delivery of audiovideo streaming services Thus

he IEEE 80211aa amendment was developed to include new fea-

ures and additional mechanisms to improve the performance of real-

ime multi-media content delivery [65] Specifically IEEE 80211aa

ddresses the following five shortcomings of previous 80211

tandards [1666]

(i) the lack of reliable and efficient support for multicast and

group communications

(ii) the incapacity of applying traffic prioritisation to different

multimedia streams or different types of frames from the same

stream

(iii) the absence of methods for cooperative resource sharing

among neighbouring APs

(iv) the lack of mechanisms for graceful degradation of audiovideo

streaming quality

(v) the non-interoperability with existing IEEE 8021 standards for

Audio Video Bridging (AVB)

In the following sections we present in detail the solutions to

hose problems introduced in the IEEE 80211aa amendment We

urther discuss the research studies that have provided the basis

or the IEEE 80211aa design and we identify the remaining open

hallenges

31 Novel features

Groupcast communication mechanisms In most audiovideo

treaming applications a group of clients must receive the same

tream simultaneously A multicast protocol is necessary to avoid that

he same content is replicated throughout the network In wireless

etworks multicast transmission can exploit the intrinsic broadcast

ature of the wireless channel ie broadcast transmissions from an

P are physically received by all other stations in the same collision

omain However multicast and broadcast frames in IEEE 80211 net-

orks are not protected by an acknowledgement mechanism as in

he case of unicast frames Thus layer-2 multicast transmissions de-

ned by legacy IEEE 80211 standards are unreliable and not suit-

ble for streaming applications To partially address this limitation

he Direct Multicast Service (DMS) was first specified in the IEEE

0211v amendment [67] Basically DMS converts multicast streams

nto unicast streams In this way frames destined to a multicast ad-

ress are individually transmitted as unicast frames to the stations

hat joined that multicast group Obviously DMS provides the same

eliability as unicast transmission services but the consumed band-

idth increases linearly with the number of group members To ad-

ress this scalability issue IEEE 80211aa includes the Groupcast with

etries (GCR) service in addition to DMS Specifically the GCR ser-

ice defines new mechanisms and the related management frames

or group formation which allows a set of stations to agree on a

hared (non-multicast) address called the groupcast concealment ad-

ress3 Furthermore the GCR service specifies two retransmission

olicies GCR Unsolicited Retry (GCR-UR) and GCR Block Ack (GCR-

A) When using GCR-UR the AP can proactively retransmit all group-

ast frames a number of times to mitigate the impact of channel

rrors (see Fig 5a)) while receivers are not requested to send ac-

nowledgements Intuitively this approach improves transmission

eliability but it still suffers from scalability issues In contrast when

CR-BA is used the AP sends a burst of consecutive groupcast frames

nd it requests the receivers to reply with a Block ACK frame which

ontains a bitmap to positively or negatively acknowledge trans-

itted frames (see Fig 5b)) The Block ACK mechanism defined for

he GCR-BA service is quite flexible because Block ACK frames can

e requested immediately after a transmission burst or after a ran-

omised back-off delay Furthermore the AP can request the Block

B Bellalta et al Computer Communications 75 (2016) 1ndash25 9

Fig 5 GCR service with different retransmission schemes

A

r

o

l

o

e

v

e

v

d

f

d

s

v

c

s

a

a

f

d

t

F

t

d

v

a

f

c

t

t

o

t

i

c

r

v

i

r

s

o

t

a

(

p

i

f

a

o

i

t

t

c

s

fi

o

c

I

n

i

a

t

a

t

f

i

B

T

f

t

a

g

a

a

i

s

a

d

s

p

m

m

i

a

t

b

t

i

(

r

a

S

i

c

s

c

T

Q

p

a

3

g

t

4 TSPEC is a traffic specification sent from a QoS capable wireless client that requests

a certain amount of network traffic from the AP for the traffic stream it represents

CK frame to all groupcast recipients or only to a subset of them to

educe overheads and delays The advantages of the GCR methods

ver broadcast and DMS have been extensively demonstrated in the

iterature [1668]

Intra-access category prioritisation The IEEE 80211e amendment

nly allows traffic differentiation between four different access cat-

gories (ACs) that are broadly mapped to four application classes

oice (VO) video (VD) best-effort (BE) and background (BK) How-

ver there is a variety of streaming services ranging from simple

ideoconferencing to HD streaming over IPTV systems which have

ifferent QoS requirements (see Table 3) To provide the ability to dif-

erentiate among individual streams IEEE 80211aa includes an ad-

itional scheduling layer with respect to IEEE 80211e IEEE 80211aa

plits each one of the transmission queues associated with voice and

ideo ACs into a primary and an alternate queue In this way spe-

ialised scheduling rules can be applied to decide which queue to

erve when the EDCA function for inter-AC collision resolution grants

n access opportunity to voice or video ACs To facilitate the man-

gement of service level agreements IEEE 80211aa follows the de-

ault mappings between user priority values and traffic types that are

efined in the IEEE 8021D standard [69] It is then straightforward

o further map traffic types onto transmission queues and ACs (see

ig 6) Finally it is important to point out that the intra-AC differen-

iation functionality can be used to provide more sophisticated traffic

ifferentiation than simple stream prioritisation For instance most

ideo applications use Scalable Video Coding (SVC) schemes that en-

ble the partitioning of a video sequence into multiple layers with dif-

erent qualities and rates [70] Typically an SVC-based video stream

ontains a base layer which provides a basic level of quality and mul-

iple enhancement layers which can only be decoded together with

he base layer to improve the video quality Thus the different layers

f the same encoded video steam can be easily mapped to different

ransmission queues to receive differentiated QoS [71]

Stream classification service The stream classification service (SCS)

s an optional service that can be provided by an AP to the asso-

iated stations to classify multimedia streams based on arbitrary

ules that are established directly by the stations instead of the con-

entional 8021D user priorities To this end the station request-

ng the use of SCS must initiate an SCS session by sending an SCS

equest frame to the AP which contains an identifier for the SCS

tream and the descriptor of the classification rule The AP may accept

r reject the requirements specified by the station Once accepted

he AP must assign all frames that match the classification rule to

specific AC When intra-access category prioritisation is enabled

see Section 331) calternate transmit queues so that finer grained

rioritisation can be applied Finally there is also a Drop Eligibil-

ty Indicator (DEI) bit in the SCS descriptor that indicates whether

rames from this traffic stream can be dropped in the case that there

re insufficient resources Specifically frames with the DEI bit set to

ne have a higher probability of being discarded because their max-

mum number of allowed retries is smaller than the default Note

hat how to combine intra-AC queues and frame dropping settings

o achieve graceful degradation of the audiovideo stream quality in

ase of bandwidth shortage is beyond the scope of the IEEE 80211aa

pecification

Overlapping Basic Service Set (OBSS) management Network densi-

cation ie a denser deployment of wireless infrastructure nodes is

ne of the key strategies that is used nowadays to easily increase the

apacity of wireless systems even for indoor WLANs [72] However

EEE 80211 networks have a limited number of orthogonal chan-

els available and even if optimised frequency planning is applied

t might happen that neighbouring APs are mutually interfering and

station may affect multiple overlapping BSSs In this case conges-

ion not only increases but it is also likely to observe an unfair us-

ge of wireless capacity with the channel retained by one AP for long

ime intervals This is mainly due to the neighbourhood capture ef-

ect ie hidden terminal phenomena among APs [73] To address this

ssue IEEE 80211aa specifies a new functionality called Overlapping

SS (OBSS) management which is based on two new mechanisms

he first defines a set of parameters to quantify the load and inter-

erence among neighbouring BSSs such as medium occupancy frac-

ion number of admitted audiovideo streams data traffic volumes

nd the number of BSSs that are using the same channel as the tar-

et one Note that the traffic load consists of two components the

llocated traffic which is derived on the basis of the TSPEC values of

dmitted streams4 and predicted traffic which is evaluated by track-

ng the maximum value of the allocated EDCA and HCCA traffic over

even-day periods Once load measurement reports are exchanged

mong the APs a second OBSS component is responsible for coor-

inated admission control procedures on the basis of two suggested

haring schemes proportional sharing and on-demand sharing The

urpose of both schemes is to keep the total allocated traffic below a

aximum value in order to provide some QoS protection to admitted

ultimedia streams Finally IEEE 80211aa recommends implement-

ng additional OBSS management procedures for channel selection

nd cooperatively creating HCCA schedules that do not collide

Interworking with IEEE 8021AVB Audio Video Bridging (AVB) is a

erm commonly used to denote a set of technical standards developed

y IEEE to support real-time streaming services with bounded latency

hrough IEEE 802 networks [74] This objective is achieved by spec-

fying mechanisms to allow the synchronisation of multiple streams

IEEE 8021AS [75]) and traffic shaping (IEEE 8021Qav [76]) and to

eserve network resources for specific audiovideo streams traversing

bridged local area network by using a signalling protocol called the

tream Reservation Protocol (SRP) (IEEE 8021Qat [77]) IEEE 80211aa

ntegrates the SRP operations with the EDCA admission control pro-

edures Specifically the SRP RequestResponse messages are encap-

ulated in the management frames that are used to carry the traffic

haracteristics and the QoS requirements during admission control

his enables the end-to-end management of resource reservation for

oS guaranteed streams even when one or more IEEE 80211 links are

art of a path from the stream producers (called IEEE 8021Q talkers)

nd the stream consumers (called IEEE 8021Q listeners)

32 Open challenges

In recent years several MAC enhancements have been investi-

ated to improve QoS guarantees for real-time multimedia applica-

ions in IEEE 80211 networks [20] and the IEEE 80211aa standard

10 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 6 Stream classification and inter-AC traffic prioritisation

a

l

t

r

n

a

t

n

i

t

i

t

r

d

p

t

t

e

m

b

t

a

i

s

o

s

a

e

t

i

f

e

m

8

t

I

i

fi

which was finalised in 2012 included several of these proposed im-

provements Significant research efforts have focused on improving

the transmission reliability of multicasting by integrating ARQ mech-

anisms in IEEE 80211-based multicast transmissions Modifications

to the legacy MAC protocol were proposed in [78] to enable the

RTSCTS option in multicast mode and to select one or more multi-

cast receivers (called leaders) for acknowledging multicast data pack-

ets However these enhancements require changes to the standard

specifications The main problems of leader-based ARQ schemes are

leader election and the trade-off between scalability and reliability

The authors in [79] propose selecting the multicast recipient operat-

ing in the worst channel conditions as the unique leader but this ap-

proach may perform poorly in lossy environments In the Batch mode

multicast MAC (BMMM) [80] all multicast recipients are polled by

the multicast originator to send individual ACKs but this scheme is

not suitable for large multicast groups The Enhanced Leader Based

Protocol (ELBP) is proposed in [81] on the basis of multiple ACK-

leaders and block acknowledgement techniques Analytical models

are then developed to help select optimal ACK-leaders to meet ap-

plication QoS requirements However the models apply only to sat-

urated traffic while multimedia streams are typically bursty Another

class of reliable multicast protocols relies on busy tones to reduce

packet losses due to collisions [82] but the additional radio inter-

face needed for the busy tone limits the practicality of such solu-

tions An alternative approach to avoid collisions of multicast pack-

ets is the multicast collision prevention (MCP) scheme [83] which

is based on the use of a shorter waiting time for transmitting mul-

ticast packets An interesting approach is also proposed in [84] to

retransmit lost packets using an online linear XOR coding algorithm

However a modification to the standard MAC protocol is required to

enable simultaneous ACK transmissions In summary several differ-

ent methods have been proposed to improve multicast transmission

reliability by integrating ARQ schemes into the protocol architecture

but there are not conclusive results on which is the best solution The

choice of the most efficient mechanism depends on a variety of in-

terdependent factors such as loss ratios channel congestion multi-

cast group size and QoS requirements of multimedia streams A com-

prehensive analytical framework is needed to optimise the setting of

the parameters for each scheme and to dynamically select the best

one

As discussed above one main difference between unicast services

nd multicast services in the legacy IEEE 80211 standard was the

ack of acknowledgements Another critical difference is that mul-

icast frames must be transmitted using a fixed rate in the basic

ate set while the transmission rate of unicast frames can be dy-

amically adapted to the channel and traffic conditions [85] Thus

group of research papers has investigated the use of rate adap-

ation to improve the throughput of multicast services in IEEE 80211

etworks [7186ndash89] For instance the authors in [86] propose us-

ng RTS frames to allow group members to estimate channel condi-

ions Each member will then send a dummy CTS frame with a length

nversely proportional to channel quality In this way the multicast

ransmitter can use the collision duration to predict the lowest data

ate that can be used for group transmissions The overhead intro-

uced by this mechanism is quite high however The solution pro-

osed in [88] called ARSM also relies on feedback messages sent by

he multicast receivers called multicast response frames to identify

he group member exhibiting the poorest channel conditions How-

ver in this case a different back off timer is associated with each

ulticast receiver depending on the SNR of previously received feed-

ack messages in order to prevent collision An approach similar to

he one employed in the Auto Rate Fallback (ARF) protocol a rate

daptation scheme originally proposed in [90] is used in [87] Specif-

cally the number of successful consecutive transmissions and con-

ecutive transmission failures are used to decide when to increase

r decrease the transmission data rate respectively A modified ARF

cheme is also proposed in [71] which can be applied to videos that

re encoded into two layers namely the base and enhancement lay-

rs However how to integrate rate adaptation with the different re-

ransmission policies that are defined in IEEE 80211aa is still an open

ssue

One research area that is expected to be crucial in the success-

ul development of IEEE 80211aa-based products is the design of

fficient scheduling algorithms for supporting voicevideo traffic Al-

ost all research work in this field has been triggered by the IEEE

0211e amendment that enhanced the original IEEE 80211 MAC with

wo new QoS-aware access mechanisms ie EDCA and HCCA [91]

n principle with a well-designed admission control and schedul-

ng scheme HCCA is able to provide hard QoS guarantees to traf-

c flows [9293] However HCCA is rarely implemented in IEEE

B Bellalta et al Computer Communications 75 (2016) 1ndash25 11

8

c

c

a

s

b

w

o

j

t

f

a

Q

t

t

A

p

t

s

a

f

s

H

a

i

c

i

t

s

f

s

s

m

i

t

s

l

t

t

[

t

i

o

b

r

W

v

m

8

s

i

o

p

q

4

c

e

t

d

a

t

m

Fig 7 WLANs for M2M communications STAs represent sensor and actuator devices

4

t

w

s

a

c

g

e

(

p

o

t

p

9

r

o

a

p

c

o

i

p

b

f

t

s

o

i

4

8

f

t

i

s

m

d

0211e-based WLANs owing to its higher complexity and cost con-

erns Instead EDCA is widely adopted Most papers have thus fo-

used on improving EDCA performance Many papers have proposed

nalytical models for various subsets of EDCA functionalities For in-

tance a saturation-based performance analysis is conducted in [94]

y differentiating the minimum back-off window size the back-off

indow-increasing factor and the retransmission limit The authors

f [9596] also model AIFS differentiation while the model in [97]

ointly captures all the four EDCA parameters for traffic differen-

iation More recent papers have analysed the EDCA performance

or non-saturated conditions and for arbitrary buffer sizes [98] The

uthors in [99] have developed an analytical model to predict the

oS levels that can be achieved once a new voicevideo flow is in-

roduced in the WLAN A Kalman filter is proposed in [100] to ob-

ain estimates on the number of active transmission queues of each

ccess Category in EDCA These analytical models can then be ex-

loited to derive the optimal configuration of the EDCA parameters

o achieve given performance criteria or to design admission control

chemes that preserve QoS constraints For instance a scheme that

ssigns contention-window values to achieve pre-defined weighted-

airness goals is proposed in [101] A control-theoretic scheme is de-

igned in [102] with the goal of minimising the video traffic delay

owever most of these solutions rely on non-realistic assumptions

bout video traffic dynamics An alternative class of solutions dynam-

cally updates the EDCA parameters based on the observed network

onditions In [103] the EDCA parameters are optimised consider-

ng a WLAN with rigid and elastic traffic simultaneously analysing

he interactions between both types of traffic The authors in [104]

pecify several bandwidth-sharing mechanisms with guaranteed QoS

or voice and video traffic Measurement-based admission control

chemes are proposed in [105] A TXOP adaptation method is de-

cribed in [106] that takes into account video frame sizes and trans-

it queue lengths However the main drawback of these solutions

s that they are based on heuristics and hence do not ensure op-

imal and guaranteed performance Finally a third category of re-

earch papers tries to improve video performance by designing cross-

ayer scheduling approaches Specifically these works take advan-

age of multi-layer video encoding to classify the frames according

o their importance and assign them to different access categories

107] For instance the authors in [108] define classifiers and waiting

ime priority schedulers that dynamically change the packet prior-

ties according to end-to-end delay measurements A disadvantage

f this approach however is that an additional adaptation layer may

e needed to implement the complex interactions that are typically

equired between the video coding applications and the MAC layer

e conclude this section by pointing out that existing studies pro-

ide the basic design principles and techniques for improving multi-

edia streaming performance in IEEE 80211 networks Still the IEEE

0211aa standard poses new research challenges that have not been

ufficiently explored and that will require innovative solutions For

nstance scheduling between primary and alternate queues is still an

pen research area as the mapping of individual frames to multi-

le queues in order to achieve graceful degradation of voicevideo

uality [16]

Sensor Networks and machine-type communications

As discussed in Section 2 M2M communications refer to any

ommunication technology that enables sensoractuator devices to

xchange information and perform actions without the manual assis-

ance of humans This section reviews the main features currently un-

er consideration in the development of the upcoming IEEE 80211ah

mendment which targets the main challenges of those networks as

he IoT in general such as the energy consumption or the manage-

ent of many devices

1 The IEEE 80211ah amendment

The IEEE 80211ah amendment [109] aims to provide WLANs with

he ability to both manage a large number of heterogeneous STAs

ithin a single BSS and minimise the energy consumption of the

ensor-type battery-powered STAs

The initial design requirements of the IEEE 80211ah amendment

re detailed in [110] these entail the support of up to 8192 STAs asso-

iated with a single AP the adoption of efficient power saving strate-

ies a minimum data rate of 100 kbps the operation in the license-

xempt sub 1 GHz band and a coverage up to 1 km in outdoor areas

see Fig 7 for an illustrative example) A preliminary assessment of

erformance of the IEEE 80211ah technology in terms of the number

f STAs that can be effectively supported in a single WLAN as well as

heir energy consumption is presented in [111]

IEEE 80211ah operates over different sub-1 GHz ISM bands de-

ending on country regulations 863ndash868 MHz in Europe 902ndash

28 MHz in the US and 9165ndash9275 MHz in Japan China South Ko-

ea and Singapore also have specific channelisations Channel widths

f 1 MHz and 2 MHz have been adopted although 4 8 and 16 MHz

re also supported in some countries IEEE 80211ah furthermore pro-

oses new PHY and MAC layers The IEEE 80211ah PHY layer can be

onsidered to some extent a sub-1 GHz version of the IEEE 80211ac

ne At the physical layer OFDM is the chosen modulation method us-

ng 32 or 64 tonessub-carriers that are spaced by 3125 kHz The sup-

orted modulations include BPSK QPSK and from 16 to 256-QAM A

road range of antenna technologies ranging from single-user beam-

orming to MIMO and DL-MU-MIMO which was first introduced in

he IEEE 80211ac amendment are also included in the IEEE 80211ah

pecification Similarly the IEEE 80211ah MAC protocol include most

f IEEE 80211 main characteristics further extending its power sav-

ng (PS) mechanisms

11 Novel features

This section overviews the extensions introduced by the IEEE

0211ah amendment to the IEEE 80211 PS mechanisms to account

or the specific characteristics of resource-constrained sensor and ac-

uator devices A more detailed review can be found in [110] includ-

ng a performance assessment of IEEE 80211ah in several of the key

cenarios for M2M communications such as agriculture and animal

onitoring smart metering and industrial automation plants In ad-

ition a detailed survey of the IEEE 80211ah is reported in [112]

12 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 8 IEEE 80211ah PS mode for TIM-STAs non-TIM STAs and Unscheduled STAs

a

d

w

i

a

a

c

m

i

a

e

t

s

m

t

t

u

b

i

I

s

d

f

a

b

s

a

b

S

T

c

t

w

I

4

a

a

which addresses aspects not considered in [110] such as the use of

sectorisation to avoid overlapping between multiple IEEE 80211ah

WLANs and the use of relays to further extend the IEEE 80211ah cov-

erage among other aspects Finally in [113] both IEEE 80211ah PHY

and MAC characteristics are introduced with emphasis on the ben-

efits of using the sub-1 GHz ISM band in terms of link budget the

location of the OFDM pilots for outdoors operation the use of bidi-

rectional transmission opportunities and the new packet formats to

minimise overheads among others

Enhanced power saving mechanism PS mechanisms for WLANs

were already considered in the development of the first IEEE 80211

standard with the goal of improving the lifetime of battery equipped

devices [114] In PS mode STAs keep the transceiver in sleeping mode

as much time as possible They periodically wake up to listen to the

beacons transmitted by the AP Those beacons indicate whether an

STA has packets waiting for it at the AP In the positive case that STA

remains awake and requests the delivery of those packets Otherwise

given it has nothing to receive it returns to sleep mode until the next

beacon is expected

In the IEEE 80211ah amendment time is divided into pages DTIM

(Delivery Traffic Indication Map) periods TIM (Traffic Indication Map)

periods and slots DTIM and TIM periods begin with the correspond-

ing DTIM and TIM beacons sent by the AP The functions of DTIM and

TIM beacons are described below

1 DTIM beacons They inform as to which TIM Groups (ie the

group of STAs assigned to the same TIM period) have pending

packets at the AP

2 TIM beacons Each TIM message informs a TIM Group about

which specific STA has pending data in the AP Between two

consecutive DTIMs there are as many TIM beacons as TIM

Groups

Using this DTIMTIM-based approach any STA can enter into a

power saving state if it does not have packets pending for transmis-

sion and one of two conditions is met (1) it observes in the DTIM

beacon that there is no downlink traffic addressed to its TIM Group or

(2) it observes in the DTIM beacon that there is some downlink traffic

addressed to its TIM Group but that STA does not explicitly appear in

the TIM beacon Compared to the preliminary IEEE 80211 PS mecha-

nism this approach reduces the size of the Traffic Indication Map in

each TIM beacon thus reducing the overhead and the time STAs need

to listen and process them In addition TIM periods can be organised

in pages which further increases the number of TIM groups between

two DTIM beacons

The temporal organisation of pages DTIM and TIM periods is re-

ported in Fig 8 which shows the division of TIM periods in RAW (Re-

stricted Access Window) and PRAW (Periodic RAW) The RAW is a

time interval in each TIM period where TIM stations can transmit and

receive data (see below the different types of STAs defined in IEEE

80211ah) In addition it can be divided into several downlink and

uplink slots for further granularity In the downlink the slots are as-

signed to a single STA or a group of STAs while the uplink slots are

randomly selected by the STAs with packets ready for transmission

The PRAW is the period of time in each TIM where non-TIM stations

can transmit and receive data

Types of STAs IEEE 80211ah supports three types of STAs TIM

non-TIM and Unscheduled STAs A TIM station is assigned to a TIM

Group Their data transmissions must be performed within a RAW

(Restricted Access Window) period Non-TIM stations do not have to

listen to beacons to transmit data During the association process

non-TIM devices directly negotiate with the AP to obtain a trans-

mission time allocated in a PRAW The following channel access can

either be renegotiated or occurs periodically depending on the re-

quirements set by the STA Unscheduled STAs do not need to listen to

any beacons similar to non-TIM stations Even inside any restricted

access window unscheduled STAs can send a poll frame to the AP

sking for immediate access to the channel The response frame in-

icates an interval (outside both restricted access windows) during

hich unscheduled stations can access the channel This procedure

s meant for STAs that transmit data very sporadically

Hierarchical station organisation To support a large number of STAs

nd their organisation in pages DTIM and TIM periods IEEE 80211ah

ssigns to each associated STA a unique identifier of 13 bits which is

alled the Association Identifier (AID) Using this new AID the maxi-

um number of supported STAs is increased from the original 2007

n IEEE 80211 to 8191 (= 213 minus 1) in IEEE 80211ah However it also

llows categorising STAs according to the type of application they are

xecuting their power level or even their desired QoS by assigning

hem to different TIM groups

Long sleeping periods IEEE 80211ah offers TIM Non-TIM and Un-

cheduled STAs the possibility to set very long doze times (up to

onths) The corresponding clock drift produced by such long doze

imes must be taken into consideration however as the higher the

ime an STA has been asleep the further in advance it should wake

p to avoid possible synchronisation problems with the network

Efficient small data transmission Three new enhancements have

een proposed to reduce the overhead when the data packet size

s small First while IEEE 80211 contains a 28-byte MAC header

EEE 80211ah proposes a short 18-byte version by using AIDs in-

tead of MAC addresses Second IEEE 80211 has defined several null

ata packet (NDP) frames which consist only of a PHY header These

rames can be used to create short ACKs short Block ACKs short CTSs

nd short PS-Polls Finally a Fast Frame Exchange mechanism has

een developed so that if an STA has data to transmit it can notify a

uccessful reception by transmitting its data frame instead of an ACK

Sectorisation Since the PHY layer is based on the IEEE 80211ac

mendment single and multi-user beam-forming are also supported

y IEEE 80211ah This allows the transmission of data to multiple

TAs simultaneously in the downlink increasing the system capacity

he use of the beam-forming capability of IEEE 80211ah APs is also

onsidered as a means to group STAs into different independent an-

enna sectors with the main goal of reducing interference issues This

ould be particularly useful in the case of overlapping with other

EEE 80211ah WLANs or in the presence of hidden nodes

12 Open challenges

A first open challenge is to understand the coverage and achiev-

ble transmission rates in IEEE 80211ah WLANs in both indoor

nd outdoor scenarios Propagation models for WLANs working at

B Bellalta et al Computer Communications 75 (2016) 1ndash25 13

f

c

8

p

R

p

i

m

l

a

m

c

p

p

f

m

t

a

e

p

t

m

a

l

T

m

a

i

n

t

t

t

u

p

c

w

f

a

t

8

a

m

a

s

l

a

m

b

m

d

S

i

a

o

a

R

t

n

G

fi

s

t

p

t

c

u

c

w

i

o

p

g

a

c

i

f

c

(

d

m

l

i

R

t

l

t

L

m

a

m

o

e

f

a

I

i

t

w

i

t

a

c

w

p

f

t

d

5

i

s

e

s

w

5

a

c

a

n

requencies lower than 1 GHz are evaluated in [115] The authors

ompare two path loss propagation models proposed by the IEEE

0211ah Task Group (one for macro and one for picohotzone de-

loyments) [116] with Lee and HatandashOkumura propagation models

esults show that the IEEE 80211ah channel models underestimate

ath loss with respect to Lee and Hata models Moreover in a compar-

son with empirical data it is observed that the IEEE 80211ah channel

odels also underestimate the initial loss and the slope of the path-

oss function A new model parameterisation is thus proposed by the

uthors In [117] the feasibility of an IEEE 80211ah WLAN deploy-

ent is also evaluated in terms of the achievable range and bit rate

omputed on the basis of the link budget using the same path loss

ropagation models as in [115] Results show that the transmission

ower limitations in the uplink can limit the overall network per-

ormance Finally in [118] the authors evaluate the achievable trans-

ission range for the different transmission rates and the achievable

hroughput for different combination of transmission power values

nd transmission rates Further studies are required due to the het-

rogeneity of scenarios in which IEEE 80211ah WLANs can be de-

loyed to characterise the effective coverage ranges with special at-

ention to mixed indoors and outdoors scenarios and in presence of

obile nodes

Due to the wide coverage IEEE 80211ah WLANs may be severely

ffected by the presence of hidden terminals To mitigate such a prob-

em smart solutions to guarantee that the STAs assigned to the same

IM Group are inside each other transmission range or the employ-

ent of RTSCTS-based solutions may contribute to mitigate such

problem An overview of long-range scenarios for IEEE 80211ah

s presented in [119] including a detailed description of the hidden

ode problem in them

Single-hop uplink transmissions from distant STAs require high

ransmission power to reach the AP Since not all STAs may be able

o transmit at the required power a solution could be the introduc-

ion of nodes able to relay transmissions from them However the

se of relays has to be efficiently harmonised with the operation in

ower saving mode allowing temporal periods in which the relays

an gather the data from their associated STAs and then periods in

hich they forward the data to the AP which is still an open challenge

or IEEE 80211ah WLANs In case relays are used data prediction and

ggregation techniques could be implemented to make more efficient

ransmissions [120]

Another challenge is the assessment of the efficiency of IEEE

0211ah PS mechanisms including their optimisation In [111] the

uthors evaluate the capacity of the TIM and page segmentation

echanisms in terms of the maximum number of nodes supported

nd the energy consumed given a certain network traffic profile Re-

ults confirm that a large number of STAs can be supported with

ow energy consumption The impact of the number of nodes oper-

ting in PS mode on the energy consumption and the delay perfor-

ance is analysed in [121] A detailed analysis of the TIM Group-

ased channel access adopted by the IEEE 80211ah task group is

ade in [122] where the STAs assigned to each TIM are uniformly

istributed between the slots of the RAW Since only one group of

TAs is allowed to transmit in each RAW slot the channel contention

s minimised The number of downlink and uplink slots in a RAW

s well as their duration is optimised on the basis of the number

f active STAs and the network traffic profile in [123] In [124] the

uthors propose an algorithm to determine the optimal size of the

AW interval for uplink transmissions based on the estimation of

he number of STAs transmitting and the duration of the RAW Fi-

ally in [125] the authors address the contention problems in Smart

rid communication networks with many nodes and periodic traf-

c when using an IEEE 80211ah WLAN It is still a challenge to con-

ider smart systems able to adapt the IEEE 80211ah parameters to

he instantaneous system state in order to save as much energy as

ossible

Since the PS mechanisms affect only the operation of TIM STAs

he performance of non-TIM and Unscheduled STAs has not yet been

onsidered in the literature To identify the scenarios in which the

se of non-TIM and Unscheduled STAs are of interest is still an open

hallenge as further investigating how non-TIM STAs negotiate over

hich PRAWs they can use to transmit and receive data Co-existence

ssues between the three types of STAs in a single network is also an

pen challenge that requires further work to guarantee a minimum

erformance level for all of them

The design of strategies to distribute the STAs between all the TIM

roups based on their specific traffic profile position battery level

nd application priority is another important challenge that is still

ompletely open EDCA is the default channel access scheme included

n the IEEE 80211ah amendment and provides some basic traffic dif-

erentiation capabilities at the packet level Besides the different ac-

ess categories (ACs) needing to be renamed and their parameters

Arbitration Inter-Frame Spacing or AIFS CWmin TXOP duration) up-

ated to fit the specific traffic profiles of M2M communications other

echanisms can be implemented A mechanism to assign the down-

ink RAW slots when packets from multiple priority levels are wait-

ng for transmission at the AP is required for example Also uplink

AW slots are currently selected randomly by the STAs that want to

ransmit a packet A mechanism to reserve some slots at each up-

ink RAW for high priority STAs may therefore be an option although

his may severely degrade the performance of the low priority STAs

astly different repetition patterns for different priority TIM groups

ay also allow priority STAs assigned to those priority TIM groups to

ccess the channel more often We expect this challenge will receive

uch attention in the upcoming years because of the heterogeneity

f sensors that will be connected though a single IEEE 80211ah AP

specially in urban scenarios

Similar to what has been discussed for 4G cellular networks used

or M2M communications [126] the development of mechanisms to

void and resolve congestion situations is another open challenge for

EEE 80211ah WLANs with many nodes For instance a mechanism

n which STAs have to wait a random number of DTIM periods before

hey can start a transmission may be implemented This approach

ould effectively distribute the traffic over a longer period of time

n overload conditions but otherwise would unnecessarily increase

he access delay Adaptive and load-aware solutions may play again

n important role to keep the optimal WLAN operation in all possible

ircumstances

Finally since IEEE 80211ah will compete with 4G5G cellular net-

orks and WSNs to provide M2M connectivity in many different ap-

lication domains such as smart cities or e-health comparative per-

ormance studies to determine the strong and weak points of each

echnology are required including also aspects such as the cost of the

evices and the system reliability

Cognitive radio technology for TV White Spaces

This section gives an overview of the IEEE 80211af amendment

ntroducing its most relevant features and open challenges Fig 9

hows the basic components and features considered for WLANs op-

rating in the TV White Spaces (TVWS) which include local spectrum

ensing by the AP and STAs and the use of geolocation data bases

ith information on channel availability

1 The IEEE 80211af amendment

Thanks to the transition from analog TV to digital TV several VHF

nd UHF spectrum channels used for decades for analog TV broad-

asting are now unused or are under-utilised in many geographical

reas [127ndash129] This ldquodigital dividendrdquo of spectrum has suggested

ational regulators such as the FCC in the US and Ofcom in the UK

14 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 9 WLAN operating in TVWS basic elements and functionalities

Fig 10 Hidden terminal problem in TVWS when the secondary network uses only in-

stantaneous channel sensing to decide whether or not a TV channel is occupied When

the STA is not able to detect the TV signal due to the presence of obstacles it may

initiate a transmission and create interference to the primary users

s

c

5

a

i

p

t

o

o

t

e

c

u

l

m

b

a

m

t

e

t

t

t

m

D

c

I

n

b

u

8

c

a

t

i

t

w

f

to discuss how to reuse these channels for unlicensed devicesrsquo com-

munications [130131] These potentially vacant channels in the VHF

and UHF bands are referred to as TVWS [132] and include spectrum

portions like the 470ndash790 MHz in Europe and non-contiguous 54ndash72

76ndash88 MHz 174ndash216 MHz 470ndash698 MHz and 698ndash806 MHz in USA

A snapshot of the TV spectrum occupancy in the city of Barcelona in

2012 is shown in [133]

The attractive characteristics of TVWS (not only for WLANs) in-

clude the ability to penetrate through walls and other obstacles much

more effectively than other widely used spectrum bands such as the

24 and 57 GHz ISM bands [134ndash137] This fact along with a progres-

sive spectrum scarcity in ISM bands has suggested the birth of the

IEEE 80211af amendment published in February 2014 [138] which

provides the IEEE 80211 operational characteristics for TVWS access

of unlicensed White Space Devices (WSD) A good summary of IEEE

80211af can be found in [137] The main advantage of operating IEEE

80211 WLANs in the TVWS comes from an increased coverage range

which can reach up to 1 km in rural areas and open fields [139]

and less energy needed to transmit However this comes at the cost

of an increased interference risk to other WSDs which creates co-

existence problems and demands new PHY and MAC layers to ef-

ficiently support channel access and operations preserving licensed

usersrsquo devices

WLANs operated in the TVWS could cover a number of interest-

ing and emerging use cases and scenarios eg Internet access in ru-

ral or sparsely populated areas Smart Grid sensor aggregation me-

tering and control Internet of Things advanced WLAN operations

and TVWS traffic offloading in indoor environments [140141] Nev-

ertheless the concept could be realised to create or extend com-

mercial or municipality WiFi services offered to citizens with cover-

age at the whole city level realised with reasonably limited network

infrastructures [142143]

IEEE 80211af will use the PHY and MAC layers derived from IEEE

80211ac It will adopt concepts such as the OFDM multi-user beam-

forming contiguous and non-contiguous channel bonding and packet

aggregation Among the mandatory and innovative behavioural and

operational parameters the most notable one is the channel acqui-

sition support realised through remote geolocation-based spectrum

allocation databases which maintain the channelsrsquo availability infor-

mation in any given area and time of day providing upon request the

list of free channels available for use

In the following section we highlight and summarise three main

novelties that IEEE 80211af introduces In Section 511 we explain

how the access infrastructure to the remote spectrum database is de-

signed and what the requirements are we discuss the coexistence is-

sues and methodologies adopted in IEEE 80211af and we also discuss

the novel concept of non-contiguous channel bonding The following

ections also contain an illustration of related works Open research

hallenges will be summarised in Section 512

11 Novel features

This section provides an illustration of novel features introduced

nd discussed in the path to IEEE 80211af and the directions provided

n related works and standardisation initiatives to resolve classical

roblems properly characterised in the new framework of TVWS

echnologies

Channel acquisition spectrum database and channel sensing Much

f the attention in operating in the TVWS is given to the protection

f the primary licensed users in the TVWS spectrum band In general

he primary users were considered as the Digital TV (DTV) broadcast-

rs and receivers In fact receivers would suffer the wasteful effect of

ollisions during the TV broadcast reception in the case of secondary

sersrsquo (SU) transmission interference (Fig 10)

It is a well-known problem that the definition of a coherent wire-

ess channel status (idlebusy) from the viewpoint of distributed pri-

ary users (eg DTV receivers) is made difficult by the difference

etween transmitters and receivers interpretation of signals in a vari-

ble time-space collision domain resulting in hiddenexposed ter-

inals concepts Solutions proposed to reduce the hiddenexposed

erminals included the RTSCTS mechanisms and beaconing How-

ver it is difficult to implement solutions that would effectively de-

ermine the channel status of TVWS channels based on localised dis-

ributed sensing activities performed by multiple secondary users in

he same collision domain On the other hand the use of RTSCTS

echanism would not be effective in this case because most of the

TV primary users (and specifically the receivers) were not originally

onceived to transmit signals or implement the RTSCTS handshake

n conclusion the actions to identify free (and busy) TVWS chan-

els should not be based on primary usersrsquo involvement but should

e almost totally implemented by specifically designed secondary

sersrsquo methodologies

There has been much debate on the methodology that IEEE

0211af secondary devices should adopt to get knowledge of free

hannels at a given time on a given target communication area in

n effort to guarantee verified optimised and reliable TVWS spec-

rum use The three main methodologies discussed include sens-

ng solutions geolocation databases (DB) and beaconing Moreover

wo approaches are possible in general distributed and centralised

ith their well known tradeoffs in terms of effectiveness required in-

rastructure deployment and coordination overhead The distributed

B Bellalta et al Computer Communications 75 (2016) 1ndash25 15

m

b

s

c

t

u

a

t

o

l

t

w

t

t

a

o

c

c

a

(

t

o

t

t

D

t

t

o

t

f

a

t

c

i

a

c

d

t

s

t

b

I

l

(

e

a

i

R

p

e

t

c

c

i

b

i

i

t

b

t

t

t

W

a

p

c

i

t

h

p

t

n

a

t

e

t

t

t

p

t

a

r

8

w

s

w

[

i

S

o

a

t

f

i

8

i

r

a

a

i

t

c

(

I

s

a

e

t

t

c

[

i

c

i

g

s

n

s

t

i

o

o

s

c

D

o

ethodology that is receiving more attention in the literature is

ased on distributed channel sensing solutions [127144145] These

olutions include the observation of spectrum use and possible

ooperative aggregation of spectrum sensing information to increase

he accuracy of detection and reduce the vulnerability of primary

sersrsquo transmissions The spectrum sensing capability is required in

ll the secondary devices with a minus114 dBm sensitivity Dynamic

ransmission power control must be provided and the upper limit

n emissions is 100 mW EIRP (20 dBm) for portable devices (further

imited in the case of adjacent channel use to reduce out-of-band in-

erference) However database (DB) coordination of TVWS combined

ith spectrum sensing is considered the most promising and effec-

ive technique compared with spectrum sensing alone The DB spec-

rum information has more chances to be effective (ie identifying

ll free channels opportunities) and reliable (ie not prone to attacks

r erroneous interpretation of channels status) In fact regulators de-

ided to push for the conservative solution of a remotely accessible

entralised spectrum DB that maintains the information on the avail-

bility of all TVWS channels at any given point in time and location

with a target accuracy around 50 m) under their responsibility For

his reason the spectrum databases have become a mandatory part

f many spectrum sharing systems and the dominant technical solu-

ion to support TV white spaces

The IEEE 80211af standard adheres to this vision and also has

he role of a common regulatory framework for different spectrum

B implementations defining a generalised coordination architec-

ure protocols and interfaces for spectrum queries and local spec-

rum information management The protocol explicitly considers out

f scope the communication protocol and technology adopted in

he background leaving the freedom to those players responsible

or the deployment to select any suitable Internet-based and local

ccess network technology Secondary devices that need to access

he TVWS spectrum should get information on which channels they

an effectively use and which parameters to adopt without impact-

ng primary usersrsquo transmissions (and receptions) by querying the

vailable spectrum DB The query must contain the transmission

haracteristics and accurate geographical position of the secondary

evice which can be determined by means of a geo-positioning sys-

em (GPS) regularly updated in the case of movement (or manually

et for static devices) The IEEE 80211af reference system architec-

ure for DB spectrum access is composed of multiple Geolocation-

ased Spectrum information Databases (GDBs) entities connected via

nternet to Registered Location Secure Servers (RLSS) RLSSs work as

ocal proxies of the GDB for a localised group of Basic Service Sets

BSS) RLSS are connected via a secure protocol architecture with the

quivalent of multiple Access Points (AP) in different BSS realising

trusted DB infrastructure APs locally coordinate the exchange of

nformation and channel access management between the GDB (via

LSS) and the secondary usersrsquo end stations (STA)

Mode I devices are those under the control of a device that em-

loys geo-location database access while Mode II devices are those

mploying geo-location database access by themselves The informa-

ion exchange provided between the GDB and a secondary user STA

an be provided in both an open-loop (eg adopted by FCC) and a

losed-loop (eg adopted by ETSI) implementation In an open loop

mplementation daily spectrum availability information is provided

y the GDB and no feedback on the spectrum information received

s provided by STAs thus the system approach for spectrum access

s much more conservative leading to low channelsrsquo utilisation po-

ential In a closed loop implementation the STA can provide feed-

ack to the GDB and there are more communication overheads due

o the high granularity of updates however the system is more effec-

ive and reliable in the exploitation of TVWS on behalf of STAs The

ypical information provided by the GDB includes (1) the updated

hite Space Maps (WSM) of frequencies allowed for secondary use

t the timespace of the querying STA and (2) the device-dependent

ower limitations for transmission (in general conservative and ac-

urate enough to avoid relevant interference effects on primary users

dentified in the area)

The basic IEEE 80211af mechanisms regulating the communica-

ion between STA and GDB can be found in [137138]

In parallel with IEEE 80211af standards such as the IEEE 19006

ave been created that consider interfaces and data structures sup-

orting spectrum sensing information exchange applicable to spec-

rum sensing (and particularly distributed spectrum sensing) sce-

arios In particular in October 2014 the IEEE 19006 WG initiated

project for a new standard called IEEE 19006b [146] concerning

he use of spectrum sensing information to support and optimise the

ffectiveness reliability and robustness of spectrum database solu-

ions The aim is to enhance the performance and capabilities of spec-

rum databases through the use of spectrum sensing information

Co-existence Support for co-existence mechanisms so that mul-

iple technologies can effectively utilise the TVWS spectrum is im-

ortant Self co-existence between network devices of a common

echnology (eg deployed by different operators in the same area)

nd co-existence among different technologies are relevant topic of

esearch for Cognitive Radio (CR) systems and specifically for IEEE

0211af Many solutions appeared on the research scene but no one

as so far finalised as the target solution for IEEE 80211af Specific

tandardisation has been started to regulate coexistence between

ireless standards of unlicensed devices including the IEEE 802191

147] The purpose of the IEEE 802191 standard is to enable the fam-

ly of IEEE 802 Wireless Standards to most effectively use TV White

pace by providing standard coexistence methods among dissimilar

r independently operated TVWS devices Early examples of gener-

lised coexistence mechanisms included Dynamic Frequency Selec-

ion (DFS) Transmission Power Control (TPC) listen before talk (eg

or contention based IEEE 80211 80215) time division multiplex-

ng (also among different techniques such as the IEEE 80216 80220

0222) and Message-based Spectrum Contention (that is beacon-

ng messages that carry coexistence information) Opportune met-

ics must be defined to assess the measurable coexistence achieved

mong different technologies as an example the hidden node prob-

bility for a target scenario or the estimate of percentage variation

n normalised network throughput and latency (before and during

he SU transmissions) On the other hand a centralised coexistence

ontrol mechanism could be effectively realised by a central manager

or coexistence-DB like the GDB) in critical scenarios To this end

EEE 19004 [148] (a standard for heterogeneous networks in dynamic

pectrum context part of IEEE Standards Coordinating Committee 41)

ims to standardise the overall system architecture and information

xchange between the network and mobile devices which will allow

hese elements to optimally choose from available radio resources

There are three possible classification of co-existence archi-

ectures and inter-network coordination channels for CR systems

entralised coordinated and autonomous co-existence mechanisms

149] In centralised co-existence schemes the co-existence is admin-

stered by a central entity (eg like in IEEE 802191) This solution

ould be applied in both homogeneous IEEE 80211af systems (eg

ncarnated by the centralised spectrum DB entity) and in hetero-

eneous inter-networks without requiring modifications to existing

tandards under the assumption that all the involved co-existing

etwork entities would adhere to the same centralised coordination

cheme and inter-network coordination protocol On the other hand

his assumption could be hard to satisfy in some practical scenar-

os The centralised schemes are effective in providing minimisation

f inter-network interference based on the availability and quality

f centralised co-existence information In coordinated co-existence

chemes the centralised entity could be present but not taking de-

isions in general The central entity could provide a co-existence

B with required information which can be used via in-band and

ut-band signalling for co-existence decisions taken by the cluster

16 B Bellalta et al Computer Communications 75 (2016) 1ndash25

c

e

n

c

o

s

i

d

a

S

8

n

a

b

o

b

i

t

a

8

b

[

t

i

n

b

5

t

a

c

b

s

a

c

n

s

p

i

c

l

r

c

f

a

v

t

a

e

c

t

Q

a

b

s

i

c

u

n

c

i

heads of many co-existing inter-networks for implementing a proper

common coordination protocol (eg out-band busy tone signalling)

These are often hybrid solutions realising a compromise between ef-

fectiveness and suitability In autonomous co-existence schemes all the

decisions and coordination are implemented in a distributed way

by the involved network entities In these schemes many policies

can be adopted to realise a sufficient level of co-existence under a

best effort approach Solutions that could be implemented include

the use of busy tones beaconing and other signalling protocols in

dedicated control channels dynamic distributed frequency selection

schemes listen before transmit token-based and dynamic reserva-

tion schemes All the above mentioned solutions have goods and

bads and a general illustration can be found in [149] Other proac-

tive co-existence techniques try to early detect and recovermitigate

co-existence issues with relaxed inter-network coordination eg re-

alised via spectrum sensing and interference avoidancesuppression

techniques

It must be clear that different aims exist regarding the co-

existence for primary TVWS users protection against secondary

users and for secondary devices mutual interference avoidance

Many co-existence problems have been analysed in the literature

for TVWS technologies in particular between the primary (licensed)

and secondary (unlicensed) devices The coexistence management in

TVWS is complex because primary users of DVB transmissions are in-

tended to be the pure receivers rather than the broadcasters In other

words potential protected users could be everywhere and the hid-

den terminal problem could arise for these systems The co-existence

problem is also complicated by many factors that cause asymme-

try and dynamicity in the TVWS such as the mobility variable den-

sity power asymmetry and heterogeneous MACPHY layers TVWS

co-existence problems originate between the IEEE 80211af technolo-

gies and the IEEE 80222 technologies for wireless regional area net-

works (WRAN) Results of analysis of TVWS usage in Europe show

that white spaces are typically present and fragmented They are typ-

ically more abundant in rural areas where larger contiguous blocks

of unused channels are available due to broadcast network planning

giving priorities linked to population density However the exploita-

tion of TVWS in urban areas is possible eg some recent research

was realised under the assumption of exploiting shadowing effects

created in the communication environment (eg by buildings obsta-

cles) in favour of frequency reuse [150]

Since different standards for opportunistic communication in the

TV White Spaces have now been published such as IEEE 80222 [151]

IEEE 802154m [152] Weightless [153] and of course IEEE 80211af

improved methods to guarantee the coexistence of different devices

operating on several protocols in the same bands must be deployed

We highlight here some work eg [150154ndash157] that focuses on

the coexistence between different technologies operating in the same

bands In particular [155] studies the performance degradation of

IEEE 80222 when an IEEE 80211af network is operated in the same

area The problem becomes even worse when the IEEE 80211af net-

work is located near 80222 user equipment causing both networks

to perceive a strong interference from each other One proposed solu-

tion is the Coexistence Beacon Protocol [158] studied for 80222 net-

works which foresees the exchange of a periodic beacon to identify

the neighbouring and possibly interfering networks

For what concerns the co-existence problems research has ad-

dressed other important related aspects of new WLANs on TV White

Spaces For instance the higher coverage range makes them attrac-

tive for applications in the smart grid and also for M2M on TVWS

[134150159] Here the increased range compared to standard tech-

nologies in the ISM bands can have beneficial effects for indoor

mobile devices [150] and M2M [159] Specifically indoor devices

communicating on TV White Spaces offer better propagation char-

acteristics and penetration through obstacles making it easier to

realise scalable home connectivity [135] although interference and

o-existence problems could be exacerbated and must be resolved

g see [139] However studies in the literature show how the sig-

al coming from an indoor transmitter on TV White Spaces remains

onstrained inside the house making it difficult to deploy indoor-to-

utdoor networks On the other hand this shadowing limit can be

een as an opportunity lowering the interference outside the build-

ng in which the TVWS network is locally operated [134150] Un-

er this approach HDTV streaming has recently been considered as

possible use case of WLAN networks operating in the TV White

pace [160]

Non-contiguous channel bonding A relevant novel feature of IEEE

0211af is the potential for contiguous and non-contiguous chan-

el bonding which permits aggregating basic channels (also non-

djacent ones) and leveraging the possible large frequency spread

etween multiple available channels Due to the rather static nature

f primary DTV transmissions where a busy channel is unlikely to

ecome free in the near future and state changes are coarse-grained

n general it is crucial to exploit the time-locality effect and exploit

he maximum physical channel availability that could be aggregated

t any given location With the methodology inherited from IEEE

0211ac IEEE 80211af is capable of bonding together two up to four

asic channels grouped in up to two different non-contiguous chunks

137] The spectrum bandwidth of a DVB-T basic channel can be ei-

her 6 7 or 8 MHz depending on the country in which the service

s operated As an example this creates a 144 to 168 OFDM chan-

elsrsquo bandwidth potential when up to four 6ndash7ndash8 MHz channels are

onded [137]

12 Open challenges

The deployment of WLANs in the TVWS at their maximum poten-

ial still requires the resolution of open problems Generally speaking

s mentioned in previous sections the challenges for such networks

an be divided into three main categories (1) spectrum sharing

etween opportunistic (secondary) and heterogeneous devices (2)

pectrum sensingmanagement and maximum exploitation of avail-

ble spectrum (potentially enabling the spectrum-on-demand con-

ept) and (3) co-existence and interference mitigation to the primary

etwork (primary user protection) and secondary networks [154]

The overall aim of spectrum sharing techniques is to maximise

eparation to avoid overlapping or contiguous channels operations

rovided that sufficient channels are available in the TVWS spectrum

n a given spacetime scenario [154] In general important research

ontributions have to be realised for the design of proper spectrum al-

ocation methodologies and techniques satisfying multi-factorial QoS

equirements at the system-level and the user-level Novel ideas in-

lude the spectrum sharing and spectrum sensing techniques being

urther divided into cooperative and non-cooperative

Non-cooperative solutions attempt to realise spectrum sharing

nd sensing on a local basis without direct cooperation between de-

ices Examples include the ldquolisten-before-transmitrdquo approach and

ransmission power control to limit the interference and spectrum

llocation policies that are based on local node feedback only Coop-

rative solutions rely on the existence of a common communication

hannel through which devices can tentatively agree on the spec-

rum allocation that provides the desired spectrum separation and

oS Major issues include the definition of a methodology to identify

common channel (or multiple channels mutually shared in space

y pairs of heterogeneous and distributed devices) and to efficiently

hare common information For the access to the spectrum DBs the

dentification of proper common access channels is still under dis-

ussion out-of-TVWS-band cellular communications can be widely

sed given their high coverage also indoor

Trying to avoid the use of complex and resource-hungry coordi-

ation schemes and dedicated control channels rendezvous proto-

ols play a significant role both for co-existence and for establish-

ng dynamic communication channels based on spectrum sharing

B Bellalta et al Computer Communications 75 (2016) 1ndash25 17

R

n

c

t

l

o

m

e

n

i

n

w

n

a

u

s

p

b

s

m

s

a

d

t

b

a

a

m

a

t

u

8

n

t

d

d

m

f

t

n

d

n

e

r

a

h

b

f

a

c

t

u

i

f

t

d

t

s

i

f

I

l

i

a

n

c

d

I

n

o

t

i

t

e

w

t

c

m

t

a

i

t

q

b

c

a

i

l

i

a

v

t

a

a

c

t

i

t

b

t

t

a

D

p

c

s

o

q

r

p

a

g

t

e

u

d

t

m

n

t

s

f

i

1

8

f

T

endezvous is considered a fundamental problem in generalised CR

etworks at the basis of many communication processes eg in-

luding neighbour discovery routing and broadcast Rendezvous pro-

ocols attempt to establish a new link for communication on a se-

ected and agreed frequency band (channel) identified within a set

f available resources (channels) thus creating the basis for com-

unication between two or more SUs This problem is often exac-

rbated by the high (and unknown) numbers of contending SUs high

umber of available channels and their ambiguous boundaries and

dentifiers and tough co-existence requirements satisfaction in cog-

itive radio systems [161] Preliminary approaches for rendezvous

ere based on the existence of a dedicated Common Control Chan-

el (CCC) agreed among all the SUs or the existence of a centralised

gent (decision maker) assigning channels to any pairs of requesting

sers This approach could be considered in the case of IEEE 80211af

olutions based on the spectrum DB implementation however the

roblem is complicated by the fact that the common channel must

e identified and made available for all the requesting SUs at the

ame time even when these are spread over multiple collision do-

ains in space On the other hand the scalability and reliability is-

ues which are caused by the congestion of CCCs and the vulner-

bility risks of centralised DBs motivate the research of alternative

istributed solutions called blind rendezvous protocols [162] Dis-

ributed solutions have been mainly based on beaconing andor slot-

ased channel hopping algorithms exploiting randomisation sensing

nd discoverysynchronisation messages spread in slotted time over

set of candidate available channels [161] These protocols aim to

inimise the convergence time to maximise the use of resources

nd to avoid collisions among SUs by identifying and agreeing on

he use of a given common channel among N non-overlapping and

nambiguously labelled channels The main complication in IEEE

0211af systems is given by the wide availability of many different

on-contiguous spectrum bands which cannot be uniformly quan-

ised in single channels and unambiguously labelled in space in or-

er to realise a common reference domain for all the IEEE 80211af

evices In fact the knowledge of the number of channels their com-

on labels the unique IDs of SUs and the number of SUs contending

or the channel assignment are the key factors determining the effec-

iveness and good properties of blind rendezvous protocols Recently

ew classes of rendezvous protocols have been proposed which are

istributed blind and oblivious Oblivious means that available chan-

els may be identified with different labels on behalf of the differ-

nt involved SUs A number of these distributed blind and oblivious

endezvous protocols have been proposed and have been referenced

nd analysed in [162] To conclude solutions for dynamic spectrum

andoff issues (eg in case PUs suddenly appear in a region) must

e resolved in advance to minimise the latency of re-establishing a

unctional communication channel eg by implementing a proactive

lternative rendezvous channel selection in background to ongoing

ommunication processes [163]

Another challenge is to provide cooperative techniques in a dis-

ributed vs centralised implementation by analysing their convergence

nder variable conditions overheads and tradeoffs In some cases as

n vehicular networks the exploitation of scenario characteristics and

actors such as the constrained mobility could help to realise effec-

ive dissemination of sensing information over an extended and pre-

ictable horizon under a cooperative approach [164ndash166] Coopera-

ion could be helpful in the realisation of the cooperative spectrum

ensing in combination with the Geolocation DB approach Another

nteresting direction of research is the investigation of mutual ef-

ects of coexisting cooperation-based vs non-cooperative techniques

n cooperation-based devices the exchange of information could al-

ow a quick convergence to a solution and avoidance of further

nterference [145] Certainly cooperative techniques have desirable

dvantages over non-cooperative ones However tight time synchro-

isation and high coordination overheads are required The common

hannel for cooperation purposes could become a bottleneck and re-

uce the potential advantages of dynamic spectrum allocation The

EEE 80219 foresees the presence of a shared common control chan-

el (CCC) to which devices need to tune in order to gather the state

f the network Solutions such as [141] propose the use of TVWS for

he CCC

A promising direction for research is the adoption of cluster-

ng schemes for secondary devices associated with a hybrid dis-

ributedcooperative vs centralisednon-cooperative approach in an

ffort to reduce overheads while maximising advantages In this

ay secondary nodes could implement cooperation for sensing func-

ions using properly identified common channels and delegate the

entralised decisions and spectrum-DB management to well instru-

ented cluster-leader nodes This concept is incarnated to some ex-

ent by RLSS in IEEE 80211af More recently there have been propos-

ls to build distributed databases that refresh their contents period-

cally by querying the remote spectrum database [167] Here Mas-

er devices (according to Ofcom terminology) periodically cache the

uery replies from the spectrum database in order to reply to and

roadcast the spectrum availability to neighbour Slave devices

Another challenging direction for research is the definition of ac-

urate models and efficient simulation tools enabling dynamic spectrum

nalysis for both rural and urban areas Recent developments of dig-

tal maps simulation tools and propagation models theoretically al-

ow improving the capability of predicting radio propagation effects

n complex scenarios with an acceptable computation time This en-

bling technology could be used in parallel to spectrum DBs to pro-

ide more accurate forecasting of dynamic frequency allocation in

imespace scenarios Another issue to be taken into account is the

ccuracy of the remote spectrum database regarding the channel

vailability estimation Several works have already shown the inac-

uracy of propagation models due to the complexity of parameters

o be considered and differences between modelled and real scenar-

os [132134168169] Better propagation models will make it possible

o more efficiently estimate the interference between devices and

uild Radio Environment Maps (REM) or White Space Maps (WSM)

hat will possibly lead to a more accurate sharing of the radio spec-

rum [144] To this end current research could focus on how to lever-

ge simulation models in order to build more accurate frequency

Bs

In general the adoption of remote spectrum DBs methodology

ushed by national regulators has made the spectrum sensing pro-

ess optional and conservatively realised (eg power control for

ensing-only devices is limited to below 17 dBm EIRP) Thus much

f the research has focused on techniques and solutions to successfully

uery the remote database also in challenging environments such as

ural areas and indoor However since the queries must provide the

osition of the mobile device the accuracy of positioning for mobile

nd handheld devices still needs more appropriate solutions Trian-

ulation can be used to estimate the position of mobile devices with

he accuracy needed by the current regulations outdoor (50 m) How-

ver a low GPS accuracy is possible indoors [170] Certified man-

al positioning can be provided by technicians for static and indoor

evices although limiting mobility Alternatives to costly infrastruc-

ures for positioning (eg based on short range beaconing devices)

ust be deployed in indoor scenarios

Finally regarding spectrum sensing and interference mitigation

ew challenges come from the heterogeneity and different charac-

eristics that networks operating in the TV White Spaces can have

uch as different transmitting power (4 W for static devices 100 mW

or mobile and portable devices and 40 mW for communication

n channels adjacent to occupied ones) different bandwidths (5

0 15 and 20 MHz are currently foreseen for IEEE 80211af IEEE

02154m can have narrower bandwidths or wider ones) and dif-

erent medium access schemes (protocols that use either CSMA or

DMA are required to co-exist in the TV White Spaces) With all this

18 B Bellalta et al Computer Communications 75 (2016) 1ndash25

d

a

T

i

a

c

t

l

b

o

t

a

d

p

a

r

p

n

a

m

t

p

p

d

d

o

t

i

D

w

p

i

a

a

d

b

fi

d

o

t

l

b

t

b

a

i

e

o

o

c

t

t

w

A

c

M

s

c

l

O

w

g

v

t

in place the problem of determining whether another secondary de-

vice is currently transmitting in the same channel is still an open

issue

Spectrum sensing is a very hot topic in the literature with many

proposals that leverage the cooperation between devices to raise the

accuracy of the detection process eg [144171] The current litera-

ture also offers room for proposals that suggest exploiting sensing to

build more accurate spectrum DBs [164172173] Lastly regarding the

interference mitigation to the primary network regulators provide a

conservative approach based on a remote spectrum database to avoid

high risk for the primary users However in particular for the indoor

scenario the concept of secondary usersrsquo transmission in TV Grey

Spaces has been proposed TV Grey Space identifies busy channels

that are formally used by primary users in the area of interest (eg

at the rooftop level of a building for DVB-T) but which could be con-

sidered usable indoors (eg at the basement level) without causing

effective interference to the primary users [134150174] This concept

is based on the generalised spectrum underlay paradigm [175] The

remaining research challenge is to provide TV Grey space access guar-

anteeing a sufficient amount of protection to the primary user while

providing a satisfactory QoS to opportunistic devices eg through

the use of customised and validated propagation models for indoor

TVWS networks and feedback mechanisms to constantly monitor the

interference generated to the primary network

6 Emerging new trends and technologies

In this last section we review three emerging trends that in our

opinion will have a large influence in the conception of future WLANs

as they change the way WLAN protocols and functionalities are de-

veloped implemented tested and integrated with other wireless

networks

61 Programmable Wireless LANs

Especially in the enterprise environment WLAN deployments

need to support a wide range of functionalities and services This is

intrinsically difficult because of the large number of APs that must be

managed which calls for scalable solutions Typical services include

channel assignment load balancing among APs authentication au-

thorisation and accounting (AAA) policy management support for

client mobility and interference coordination Another problem is the

fact that WLAN clients autonomously take several decisions such as

which APs to associate with when to hand-over etc Therefore sup-

porting roaming clients requires the management of a large number

of association states across several APs which is a challenge if sup-

port for real-time hand-over is desired Typically such management

schemes are centralised and most of them are proprietary such as

WLAN controller solutions from Aruba [176] or Cisco [177] although

the IEEE 80211u amendment has been released to allow mobile users

to seamless roam between WiFi networks with automatic authenti-

cation and handoff [178] For example Dyson [179] enables STAs to

send information such as radio channel conditions to a centralised

controller based on a custom API (eg based on Python) As the con-

troller has a centralised view of the network it can enforce a rich

set of policies to control the network also using historical informa-

tion A demo system has been implemented along with applications

such as airtime reservations for specific clients or optimised hand-

offs However Dyson requires STAs to be modified in order to use

those new services offered by the centralised controller TRANTOR

[180] is another example of a centralised management system that

requires changes of clients in order for the infrastructure to gather

information from them in terms of eg interference measurements

In addition control commands enable the infrastructure to exercise

control such as modifying the transmit power or influencing the as-

sociation procedure Clients still use standard CSMA MAC layer for

ata transmission In contrast to DenseAP [181] or DIRAC [182] which

re also based on a centralised controller but do not modify clients

RANTOR exercises much more control on clients by the use of a ded-

cated API which enables a significantly higher gain in coordination

nd thus capacity CENTAUR [183] is another example of a centralised

ontroller that centrally schedules hidden and exposed terminals in

he downlink while using standard CSMA MAC for uplink traffic and

egacy downlink traffic To achieve good performance it uses a fixed

ack-off packet staggering techniques and a hybrid data path that

nly schedules downlink transmissions towards hidden and exposed

erminals centrally All other traffic is sent using standard DCF with

standard back-off procedure In contrast to previous approaches it

oes not require any modifications to the clients but does require data

lane centralisation and it remains questionable how scalable such

solution is for large WLAN deployments with high PHY layer data

ates beyond 100 Mbps However new approaches for fast data plane

rocessing that are currently explored in the virtualisation commu-

ity which move all the packet processing into user space may be

n option to significantly increase speed Designing such manage-

ent systems raises several interesting research questions (a) Cen-

ralised infrastructures may incur high latency but are more simple to

rogram and maintain while designing a completely distributed ap-

roach that operates close to WLAN devices may result in significant

istributed coordination and consistency problems (b) what level of

istributed control is required in order to support the flexibility needs

f future WiFi based networks supporting a high level of mobility

Traditionally WLAN APs are built on proprietary operating sys-

ems that are tightly coupled with the hardware This design makes

t hard to create new applications on top of such networking devices

espite the fact that protocols and mechanisms are available that

ould greatly improve the utility of existing networks those new

rotocols are not deployed because the closed system design makes

t very difficult to extend their functionality An important aspect for

ll such new approaches is therefore how to provide open interfaces

nd open source to speed up innovation As an example Linux-based

evices are completely open source but in order to increase flexi-

ility much more work is needed in the area of open drivers and

rmware In fact the Atheros based ath9k driver has been the main

river of innovation in the open source community because it is the

nly driver that interworks with open firmware In general solutions

hat modify the OS driver focus mainly on programmable network

evel solutions for WLANs such as channel switching or handovers

etween APs While the ath9k driver is an excellent example in how

o speed up innovation it is still very cumbersome to support flexi-

le MAC engine reprogrammability with ath9k In contrast MAClets

nd Wireless MAC processors (WMP) [184] allow a much more flex-

ble reprogramming of MAC functionalities A MAC processor is an

ntity able to execute general MAC commands that specify the MAC

perations through a software-defined state machine The behaviour

f the MAC protocol can therefore be updated at run-time by simply

hanging the sequence in which those commands are executed (ie

he MAClets) As a proof-of-concept the authors of [184] implement

he proposed MAC processor solution in a commodity WLAN hard-

are card extending the basic DCF in three directions piggy-backing

CKs a pseudo TDMA and the use of multiple channels In [185] a

ontrol framework for this WMP system is also proposed to support

AClet code mobility ie for moving loading and activating MAC

oftware programs embedded into ordinary data packets (akin to the

apsule model of traditional active networks)

The difficulty in re-programming networking hardware has also

ed to the concept of Software Defined Networks (SDNs) based on the

penFlow protocol [186] The main idea of SDNs is to extend net-

orking devices with standardised APIs that allow third-party pro-

rammers to flexibly control the data path In addition SDNs pro-

ide higher level abstractions to network designers and programmers

hrough the use of a centralised control plane offered by a network

B Bellalta et al Computer Communications 75 (2016) 1ndash25 19

Fig 11 The CLOUDMAC architecture VAP stands for Virtual Access Point and WT

stands for Wireless Termination

c

a

c

t

m

W

n

t

t

c

p

t

u

h

a

i

e

s

o

c

a

a

f

T

r

s

a

s

p

o

M

C

i

T

t

fl

A

a

a

i

c

t

s

t

e

I

t

w

H

e

b

I

l

d

o

6

r

W

s

l

a

e

e

a

s

b

c

t

(

S

p

c

F

B

t

p

T

e

t

s

u

m

i

e

o

a

A

p

d

t

t

n

p

i

t

t

f

i

ontroller such as NOX [187] which allows reuse of components such

s topology discovery or network access control for different appli-

ations The SDN concept has recently been applied to WLAN archi-

ecture in order to enable fine grained control over mobility manage-

ent and data forwarding focusing on a programmable enterprise

LAN architecture An important part of the SDN architecture is the

etwork controller which provides a centralised view of (parts of)

he SDN enabled network and uses the OpenFlow protocol to install

he forwarding rules on SDN-enabled devices (routers switches ac-

ess points cellular base stations etc)

ODIN [188] is designed to support programmability in enter-

rise WLAN architecture by separating the association state from

he physical AP They implement Light Virtual Access Points (LVAP)

sing a Split-MAC approach where the infrastructure controls the

andover procedure among different WLAN APs By managing

ssociations through SDN controllers ODIN enables proactive mobil-

ty management and load balancing within the SDN enabled WLAN

nterprise network without the need for changes in the client WLAN

tack or IEEE 80211 MAC layer While ODIN requires agents to reside

n the APs that communicate with the Odin Master within the SDN

ontroller CLOUDMAC [189] is based purely on the concept of SDN

nd virtual APs Similar to ODIN CLOUDMAC implements a Split-MAC

pproach but in addition enables the processing of WLAN MAC layer

rames within a co-located Cloud using so-called Virtual APs (VAPs)

he physical APs in CLOUDMAC are lightweight WLAN APs that are

esponsible only for sending out IEEE 80211 based MAC layer ACKs to

tandard WLAN clients and tunnel WLAN MAC layer frames through

n SDN-enabled enterprise WLAN towards the VAPs As association

tates are kept in the VAPs fast mobility is supported using sim-

le OpenFlow forwarding rules Because of the additional processing

f IEEE 80211 WLAN MAC frames in the co-located Cloud CLOUD-

AC has higher latency than a standard WLAN deployment However

LOUDMAC offers a Webservice based API to third party applications

n order to program the enterprise WLAN and enable new services

he architecture of CLOUDMAC is depicted in Fig 11 It has been ex-

ended in [190] to support QoS management and in [191] to support

exible MAC management frame prioritisation based on IEEE 80211

system based on OpenFlow has been recently proposed in [192] to

llow the station to be associated with multiple APs simultaneously

nd to switch between APs with low overhead

An important aspect to consider in order to enable programmabil-

ty is backwards compatibility Several approaches (eg [180]) require

lients to be modified to utilise the features provided This is difficult

o do in practice because it requires changes in the operating system

oftware of all clients If not all clients can utilise those APIs it is ques-

ionable how usable the new architecture will be and how much ben-

fit in terms of aggregate performance such architecture will allow

n contrast the SDN based approaches are interesting in the sense

hat they do not require changes in the client WLAN stack and work

ith the standard IEEE 80211 MAC layer deployed within the clients

owever it remains questionable how scalable such solutions are For

xample an interesting open research topic is to evaluate the scala-

ility of approaches such as CLOUDMAC [189] In addition processing

EEE 80211 MAC frames within co-located private Cloud requires low

atency support from local Cloud solutions (such as OpenStack) in or-

er to reduce the IEEE 80211 MAC processing time which is an area

f active research

2 Prototyping and testing IEEE 80211 enhancements

Most of the new proposals for next-generation WLANs are cur-

ently only evaluated using mathematical analysis and simulation

hile both analysis and simulation are necessary to characterise and

tudy those enhancements in the initial design phase or to consider

arge-scale scenarios it is difficult to consider all practical aspects of

real-world scenario This can sometimes cause significant differ-

nces between what simulations and real experiments show How-

ver real experiments are challenging because of the high complexity

nd costs of building the new hardware and software for each specific

olution to test

To mitigate implementation complexity there are several flexi-

le hardware platforms where low-level MAC mechanisms can be

ompletely implemented in software Examples based on FPGAs are

he USRP (Universal Software Radio Peripheral) [193] and the WARP

Wireless Open-Access Research Platform) [194] As an alternative

ORA (Microsoft Research Software Radio) [195] works on general

urpose computers by taking some advantage from modern multiple

ore systems

A different approach is provided by OpenFWWF [196] Open-

WWF provides an open CSMACA firmware for specific models of

roadcom chipsets so the resulting firmware can be uploaded and

ested in real commercial hardware OpenFWWF implements a sim-

le State Machine (SM) for controlling the hardware in real time

he SM evolution is driven by a main loop that reacts to events by

xecuting specific handlers When a packet originally prepared by

he Linux kernel is ready in the NIC memory handler Packet_Ready

ets up the radio hardware according to the packet metadata sched-

les the transmission and jumps back to the main loop The Trans-

ission Engine (TXE) then takes care of accessing the channel ie

t decrements the back-off counter according to the DCF rules and

ventually starts the actual transmission This triggers the execution

f the TX_frame_now event that prepares the ACK time-out clock

nd finalises the MAC header If the ACK-frame is received or if the

CK time-out expires and the maximum number of attempts for this

acket is reached handler Update_Params resets the contention win-

ow to its minimum value or otherwise doubles it Finally it loads

he back-off counter with a fresh value

In the following we will review a list of selected papers that use

he OpenFWWF firmware and the WARP or USRP platforms to test

ew proposals for WLANs in real scenarios as examples A first im-

lementation of the groupcast mechanisms defined in IEEE 80211aa

s presented in [197] The authors modify the OpenFWWF firmware

o include those functionalities and evaluate them using a 30 STAs

estbed A second example of the use of OpenFWWF is the collision-

ree MAC protocol presented in [198] In [198] the authors design and

mplement in OpenFWWF a MAC protocol that is able to achieve a

20 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 12 LTE to WIFI offloading

a

n

c

t

[

d

a

W

v

t

t

o

b

f

S

g

l

i

I

s

I

b

n

s

m

a

p

d

c

N

o

v

t

t

n

l

i

a

t

a

collision-free operation by waiting a deterministic time after success-

ful transmissions before to start a new transmission This work shows

the experimental performance of a collision-free MAC (CF-MAC) pro-

tocol for WLANs using commercial hardware The results show that

the proposed CF-MAC protocol leads to a better distribution of the

available bandwidth among users a higher throughput and lower

losses than DCF

In [199] the authors use the WARP platform to test a variant of the

DCF ndash called IEEE 80211ec ndash that substitutes control packets such as

the RTS CTS and ACK with short detectable sequences Since those se-

quences are shorter than the control packets and can be detected cor-

rectly even at lower SNRs values a significant gain in performance is

achieved MIDAS (Multiple-Input Distributed Antenna Systems) [200]

implementation using the WARP platform shows the benefits of dis-

tributing the antennas over the area to cover instead of co-locating

them at the AP in terms of capacity when MU-MIMO is employed

The authors also propose a new MAC protocol to benefit from the

spatial reuse that the DAS allows taking as a basis the IEEE 80211ac

Experimental results show that their proposal is able to achieve up

to 200 gains versus the traditional approach where all antennas are

co-located at the AP

The USRP platform has recently been proposed to develop a first

implementation of IEEE 80211ah in order to obtain a preliminary ex-

perimental performance assessment of such a technology because

no commercial off-the-shelf hardware is yet available as the amend-

ment is still in progress [201] The USRP platform is also used in

[202] to evaluate TIMO (Technology Independent Multi-Output) a so-

lution to deal with high-power non-IEEE 80211 interferers in ISM

bands

To conclude this section it is worth mentioning here also to

MAClets and WMP as given such flexibility to implement and dis-

tribute new MAC protocols and other IEEE 80211 functionalities we

believe that if a use-friendly implementation of such a MAC proces-

sors framework is provided it would significantly contribute to the

development and testing of new MAC enhancements by the research

community

63 CellularWLAN interworking

Public hotspots that offer Internet access over a WLAN using IEEE

80211 technology are now nearly ubiquitous It is forecasted that the

cumulative installed base of WiFi hotspots worldwide will amount

to 551 million by 2018 excluding private hotspots (eg WiFi access

points deployed at home) [203] The sharp increase in the availabil-

ity of public WiFi was initially perceived by mobile cellular operators

as a threat due to the additional competition from wireline Internet

service providers or emerging crowdsourced WiFi networks such as

FON5 However as cellular operators are fighting to cope with the ex-

plosion of mobile data traffic created by the rising use of multimedia

content traffic over mobile devices [15] they are also starting to use

WLANs based on the IEEE 80211 technology to offload data from their

core and access networks In general mobile data offloading refers to

the use of complementary network technologies (in licensed or un-

licensed spectrum) for delivering data originally targeted to cellular

networks Intuitively the simplest type of offloading consists of ex-

ploiting connectivity to existing co-located WiFi networks and trans-

ferring data without any delay (Fig 12) Thus this offloading tech-

nique is known as on-the-spot offloading As a consequence of this

new trend the seamless integration of cellular (eg 3GLTE) and WiFi

technologies has attracted significant research interest in recent years

(see [204] for a survey) a few solutions have already been standard-

ised [205206] and roaming between cellular and WiFi is becoming

increasingly transparent to end users CellularWLAN interworking is

5 httpscorpfoncomen

fl

t

f

lso fostered by the support in the evolving 4G standards of heteroge-

eous network deployments (HetNets) in which the existing macro

ells are complemented with a number of small low-power base sta-

ions with the goal of increasing capacity in highly congested areas

207208] It is envisaged that small cells will be based on 4G stan-

ards (eg pico and femto cells) as well as IEEE 80211 technologies

nd multimode base stations that work simultaneously with LTE and

iFi are already entering the market

It is important to point out that LTE standards already support a

ariety of mechanisms that enable data offloading However most of

he existing solutions such as Local IP Access (LIPA) and Selected In-

ernet IP Traffic Offload (SIPTO) focus on data offloading at the core

f the network [205] For instance LIPA allows an IP-enabled mo-

ile device to transfer data to another device in the same pico or

emto cell without passing through the cellular access network while

IPTO enables the routing of selected IP data flows through different

ateways The only LTE offloading mechanism that supports seam-

ess interworking with IEEE 80211-based WLANs is IP Flow Mobil-

ty and Seamless Offload (IFOM) Specifically IFOM relies on Mobile

Pv6 technologies to allow a user terminal to simultaneously route

elected IP flows over different radio access technologies [209210]

n this way a user terminal can offload selected flows to a WLAN

ased on some operator-defined policy while keeping the LTE con-

ection running However to enable such an approach it is neces-

ary to support an entity in the cellular core network that can com-

unicate with the user terminal to exchange information about the

vailability and quality of neighbouring access networks as well as to

rovide the user terminal with predefined rules to manage the han-

over process This entity in current LTE standards is called the Ac-

ess Network Discovery and Selection Function (ANDSF) server [211]

ote that the problem of selecting the best communication technol-

gy in an heterogeneous wireless network has been extensively in-

estigated in the literature both with centralised as well as decen-

ralised approaches (see [212] for a survey) However the use of mul-

iple interfaces in parallel as well as per-flow offloading are relatively

ew concepts Thus the design of scalable and efficient network se-

ection strategies for the ANDSF framework is still an open issue For

nstance three offloading methods suitable for the ANDSF framework

re proposed in [213] which are based on coverage SNR and sys-

em load A reinforcement learning approach is designed in [214] that

llows multimode base stations to autonomously steer their traffic

ows across different access technologies depending on the traffic

ype the usersrsquo QoS requirements the network load and the inter-

erence levels A number of studies have also tried to quantify the

B Bellalta et al Computer Communications 75 (2016) 1ndash25 9

Fig 5 GCR service with different retransmission schemes

A

r

o

l

o

e

v

e

v

d

f

d

s

v

c

s

a

a

f

d

t

F

t

d

v

a

f

c

t

t

o

t

i

c

r

v

i

r

s

o

t

a

(

p

i

f

a

o

i

t

t

c

s

fi

o

c

I

n

i

a

t

a

t

f

i

B

T

f

t

a

g

a

a

i

s

a

d

s

p

m

m

i

a

t

b

t

i

(

r

a

S

i

c

s

c

T

Q

p

a

3

g

t

4 TSPEC is a traffic specification sent from a QoS capable wireless client that requests

a certain amount of network traffic from the AP for the traffic stream it represents

CK frame to all groupcast recipients or only to a subset of them to

educe overheads and delays The advantages of the GCR methods

ver broadcast and DMS have been extensively demonstrated in the

iterature [1668]

Intra-access category prioritisation The IEEE 80211e amendment

nly allows traffic differentiation between four different access cat-

gories (ACs) that are broadly mapped to four application classes

oice (VO) video (VD) best-effort (BE) and background (BK) How-

ver there is a variety of streaming services ranging from simple

ideoconferencing to HD streaming over IPTV systems which have

ifferent QoS requirements (see Table 3) To provide the ability to dif-

erentiate among individual streams IEEE 80211aa includes an ad-

itional scheduling layer with respect to IEEE 80211e IEEE 80211aa

plits each one of the transmission queues associated with voice and

ideo ACs into a primary and an alternate queue In this way spe-

ialised scheduling rules can be applied to decide which queue to

erve when the EDCA function for inter-AC collision resolution grants

n access opportunity to voice or video ACs To facilitate the man-

gement of service level agreements IEEE 80211aa follows the de-

ault mappings between user priority values and traffic types that are

efined in the IEEE 8021D standard [69] It is then straightforward

o further map traffic types onto transmission queues and ACs (see

ig 6) Finally it is important to point out that the intra-AC differen-

iation functionality can be used to provide more sophisticated traffic

ifferentiation than simple stream prioritisation For instance most

ideo applications use Scalable Video Coding (SVC) schemes that en-

ble the partitioning of a video sequence into multiple layers with dif-

erent qualities and rates [70] Typically an SVC-based video stream

ontains a base layer which provides a basic level of quality and mul-

iple enhancement layers which can only be decoded together with

he base layer to improve the video quality Thus the different layers

f the same encoded video steam can be easily mapped to different

ransmission queues to receive differentiated QoS [71]

Stream classification service The stream classification service (SCS)

s an optional service that can be provided by an AP to the asso-

iated stations to classify multimedia streams based on arbitrary

ules that are established directly by the stations instead of the con-

entional 8021D user priorities To this end the station request-

ng the use of SCS must initiate an SCS session by sending an SCS

equest frame to the AP which contains an identifier for the SCS

tream and the descriptor of the classification rule The AP may accept

r reject the requirements specified by the station Once accepted

he AP must assign all frames that match the classification rule to

specific AC When intra-access category prioritisation is enabled

see Section 331) calternate transmit queues so that finer grained

rioritisation can be applied Finally there is also a Drop Eligibil-

ty Indicator (DEI) bit in the SCS descriptor that indicates whether

rames from this traffic stream can be dropped in the case that there

re insufficient resources Specifically frames with the DEI bit set to

ne have a higher probability of being discarded because their max-

mum number of allowed retries is smaller than the default Note

hat how to combine intra-AC queues and frame dropping settings

o achieve graceful degradation of the audiovideo stream quality in

ase of bandwidth shortage is beyond the scope of the IEEE 80211aa

pecification

Overlapping Basic Service Set (OBSS) management Network densi-

cation ie a denser deployment of wireless infrastructure nodes is

ne of the key strategies that is used nowadays to easily increase the

apacity of wireless systems even for indoor WLANs [72] However

EEE 80211 networks have a limited number of orthogonal chan-

els available and even if optimised frequency planning is applied

t might happen that neighbouring APs are mutually interfering and

station may affect multiple overlapping BSSs In this case conges-

ion not only increases but it is also likely to observe an unfair us-

ge of wireless capacity with the channel retained by one AP for long

ime intervals This is mainly due to the neighbourhood capture ef-

ect ie hidden terminal phenomena among APs [73] To address this

ssue IEEE 80211aa specifies a new functionality called Overlapping

SS (OBSS) management which is based on two new mechanisms

he first defines a set of parameters to quantify the load and inter-

erence among neighbouring BSSs such as medium occupancy frac-

ion number of admitted audiovideo streams data traffic volumes

nd the number of BSSs that are using the same channel as the tar-

et one Note that the traffic load consists of two components the

llocated traffic which is derived on the basis of the TSPEC values of

dmitted streams4 and predicted traffic which is evaluated by track-

ng the maximum value of the allocated EDCA and HCCA traffic over

even-day periods Once load measurement reports are exchanged

mong the APs a second OBSS component is responsible for coor-

inated admission control procedures on the basis of two suggested

haring schemes proportional sharing and on-demand sharing The

urpose of both schemes is to keep the total allocated traffic below a

aximum value in order to provide some QoS protection to admitted

ultimedia streams Finally IEEE 80211aa recommends implement-

ng additional OBSS management procedures for channel selection

nd cooperatively creating HCCA schedules that do not collide

Interworking with IEEE 8021AVB Audio Video Bridging (AVB) is a

erm commonly used to denote a set of technical standards developed

y IEEE to support real-time streaming services with bounded latency

hrough IEEE 802 networks [74] This objective is achieved by spec-

fying mechanisms to allow the synchronisation of multiple streams

IEEE 8021AS [75]) and traffic shaping (IEEE 8021Qav [76]) and to

eserve network resources for specific audiovideo streams traversing

bridged local area network by using a signalling protocol called the

tream Reservation Protocol (SRP) (IEEE 8021Qat [77]) IEEE 80211aa

ntegrates the SRP operations with the EDCA admission control pro-

edures Specifically the SRP RequestResponse messages are encap-

ulated in the management frames that are used to carry the traffic

haracteristics and the QoS requirements during admission control

his enables the end-to-end management of resource reservation for

oS guaranteed streams even when one or more IEEE 80211 links are

art of a path from the stream producers (called IEEE 8021Q talkers)

nd the stream consumers (called IEEE 8021Q listeners)

32 Open challenges

In recent years several MAC enhancements have been investi-

ated to improve QoS guarantees for real-time multimedia applica-

ions in IEEE 80211 networks [20] and the IEEE 80211aa standard

10 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 6 Stream classification and inter-AC traffic prioritisation

a

l

t

r

n

a

t

n

i

t

i

t

r

d

p

t

t

e

m

b

t

a

i

s

o

s

a

e

t

i

f

e

m

8

t

I

i

fi

which was finalised in 2012 included several of these proposed im-

provements Significant research efforts have focused on improving

the transmission reliability of multicasting by integrating ARQ mech-

anisms in IEEE 80211-based multicast transmissions Modifications

to the legacy MAC protocol were proposed in [78] to enable the

RTSCTS option in multicast mode and to select one or more multi-

cast receivers (called leaders) for acknowledging multicast data pack-

ets However these enhancements require changes to the standard

specifications The main problems of leader-based ARQ schemes are

leader election and the trade-off between scalability and reliability

The authors in [79] propose selecting the multicast recipient operat-

ing in the worst channel conditions as the unique leader but this ap-

proach may perform poorly in lossy environments In the Batch mode

multicast MAC (BMMM) [80] all multicast recipients are polled by

the multicast originator to send individual ACKs but this scheme is

not suitable for large multicast groups The Enhanced Leader Based

Protocol (ELBP) is proposed in [81] on the basis of multiple ACK-

leaders and block acknowledgement techniques Analytical models

are then developed to help select optimal ACK-leaders to meet ap-

plication QoS requirements However the models apply only to sat-

urated traffic while multimedia streams are typically bursty Another

class of reliable multicast protocols relies on busy tones to reduce

packet losses due to collisions [82] but the additional radio inter-

face needed for the busy tone limits the practicality of such solu-

tions An alternative approach to avoid collisions of multicast pack-

ets is the multicast collision prevention (MCP) scheme [83] which

is based on the use of a shorter waiting time for transmitting mul-

ticast packets An interesting approach is also proposed in [84] to

retransmit lost packets using an online linear XOR coding algorithm

However a modification to the standard MAC protocol is required to

enable simultaneous ACK transmissions In summary several differ-

ent methods have been proposed to improve multicast transmission

reliability by integrating ARQ schemes into the protocol architecture

but there are not conclusive results on which is the best solution The

choice of the most efficient mechanism depends on a variety of in-

terdependent factors such as loss ratios channel congestion multi-

cast group size and QoS requirements of multimedia streams A com-

prehensive analytical framework is needed to optimise the setting of

the parameters for each scheme and to dynamically select the best

one

As discussed above one main difference between unicast services

nd multicast services in the legacy IEEE 80211 standard was the

ack of acknowledgements Another critical difference is that mul-

icast frames must be transmitted using a fixed rate in the basic

ate set while the transmission rate of unicast frames can be dy-

amically adapted to the channel and traffic conditions [85] Thus

group of research papers has investigated the use of rate adap-

ation to improve the throughput of multicast services in IEEE 80211

etworks [7186ndash89] For instance the authors in [86] propose us-

ng RTS frames to allow group members to estimate channel condi-

ions Each member will then send a dummy CTS frame with a length

nversely proportional to channel quality In this way the multicast

ransmitter can use the collision duration to predict the lowest data

ate that can be used for group transmissions The overhead intro-

uced by this mechanism is quite high however The solution pro-

osed in [88] called ARSM also relies on feedback messages sent by

he multicast receivers called multicast response frames to identify

he group member exhibiting the poorest channel conditions How-

ver in this case a different back off timer is associated with each

ulticast receiver depending on the SNR of previously received feed-

ack messages in order to prevent collision An approach similar to

he one employed in the Auto Rate Fallback (ARF) protocol a rate

daptation scheme originally proposed in [90] is used in [87] Specif-

cally the number of successful consecutive transmissions and con-

ecutive transmission failures are used to decide when to increase

r decrease the transmission data rate respectively A modified ARF

cheme is also proposed in [71] which can be applied to videos that

re encoded into two layers namely the base and enhancement lay-

rs However how to integrate rate adaptation with the different re-

ransmission policies that are defined in IEEE 80211aa is still an open

ssue

One research area that is expected to be crucial in the success-

ul development of IEEE 80211aa-based products is the design of

fficient scheduling algorithms for supporting voicevideo traffic Al-

ost all research work in this field has been triggered by the IEEE

0211e amendment that enhanced the original IEEE 80211 MAC with

wo new QoS-aware access mechanisms ie EDCA and HCCA [91]

n principle with a well-designed admission control and schedul-

ng scheme HCCA is able to provide hard QoS guarantees to traf-

c flows [9293] However HCCA is rarely implemented in IEEE

B Bellalta et al Computer Communications 75 (2016) 1ndash25 11

8

c

c

a

s

b

w

o

j

t

f

a

Q

t

t

A

p

t

s

a

f

s

H

a

i

c

i

t

s

f

s

s

m

i

t

s

l

t

t

[

t

i

o

b

r

W

v

m

8

s

i

o

p

q

4

c

e

t

d

a

t

m

Fig 7 WLANs for M2M communications STAs represent sensor and actuator devices

4

t

w

s

a

c

g

e

(

p

o

t

p

9

r

o

a

p

c

o

i

p

b

f

t

s

o

i

4

8

f

t

i

s

m

d

0211e-based WLANs owing to its higher complexity and cost con-

erns Instead EDCA is widely adopted Most papers have thus fo-

used on improving EDCA performance Many papers have proposed

nalytical models for various subsets of EDCA functionalities For in-

tance a saturation-based performance analysis is conducted in [94]

y differentiating the minimum back-off window size the back-off

indow-increasing factor and the retransmission limit The authors

f [9596] also model AIFS differentiation while the model in [97]

ointly captures all the four EDCA parameters for traffic differen-

iation More recent papers have analysed the EDCA performance

or non-saturated conditions and for arbitrary buffer sizes [98] The

uthors in [99] have developed an analytical model to predict the

oS levels that can be achieved once a new voicevideo flow is in-

roduced in the WLAN A Kalman filter is proposed in [100] to ob-

ain estimates on the number of active transmission queues of each

ccess Category in EDCA These analytical models can then be ex-

loited to derive the optimal configuration of the EDCA parameters

o achieve given performance criteria or to design admission control

chemes that preserve QoS constraints For instance a scheme that

ssigns contention-window values to achieve pre-defined weighted-

airness goals is proposed in [101] A control-theoretic scheme is de-

igned in [102] with the goal of minimising the video traffic delay

owever most of these solutions rely on non-realistic assumptions

bout video traffic dynamics An alternative class of solutions dynam-

cally updates the EDCA parameters based on the observed network

onditions In [103] the EDCA parameters are optimised consider-

ng a WLAN with rigid and elastic traffic simultaneously analysing

he interactions between both types of traffic The authors in [104]

pecify several bandwidth-sharing mechanisms with guaranteed QoS

or voice and video traffic Measurement-based admission control

chemes are proposed in [105] A TXOP adaptation method is de-

cribed in [106] that takes into account video frame sizes and trans-

it queue lengths However the main drawback of these solutions

s that they are based on heuristics and hence do not ensure op-

imal and guaranteed performance Finally a third category of re-

earch papers tries to improve video performance by designing cross-

ayer scheduling approaches Specifically these works take advan-

age of multi-layer video encoding to classify the frames according

o their importance and assign them to different access categories

107] For instance the authors in [108] define classifiers and waiting

ime priority schedulers that dynamically change the packet prior-

ties according to end-to-end delay measurements A disadvantage

f this approach however is that an additional adaptation layer may

e needed to implement the complex interactions that are typically

equired between the video coding applications and the MAC layer

e conclude this section by pointing out that existing studies pro-

ide the basic design principles and techniques for improving multi-

edia streaming performance in IEEE 80211 networks Still the IEEE

0211aa standard poses new research challenges that have not been

ufficiently explored and that will require innovative solutions For

nstance scheduling between primary and alternate queues is still an

pen research area as the mapping of individual frames to multi-

le queues in order to achieve graceful degradation of voicevideo

uality [16]

Sensor Networks and machine-type communications

As discussed in Section 2 M2M communications refer to any

ommunication technology that enables sensoractuator devices to

xchange information and perform actions without the manual assis-

ance of humans This section reviews the main features currently un-

er consideration in the development of the upcoming IEEE 80211ah

mendment which targets the main challenges of those networks as

he IoT in general such as the energy consumption or the manage-

ent of many devices

1 The IEEE 80211ah amendment

The IEEE 80211ah amendment [109] aims to provide WLANs with

he ability to both manage a large number of heterogeneous STAs

ithin a single BSS and minimise the energy consumption of the

ensor-type battery-powered STAs

The initial design requirements of the IEEE 80211ah amendment

re detailed in [110] these entail the support of up to 8192 STAs asso-

iated with a single AP the adoption of efficient power saving strate-

ies a minimum data rate of 100 kbps the operation in the license-

xempt sub 1 GHz band and a coverage up to 1 km in outdoor areas

see Fig 7 for an illustrative example) A preliminary assessment of

erformance of the IEEE 80211ah technology in terms of the number

f STAs that can be effectively supported in a single WLAN as well as

heir energy consumption is presented in [111]

IEEE 80211ah operates over different sub-1 GHz ISM bands de-

ending on country regulations 863ndash868 MHz in Europe 902ndash

28 MHz in the US and 9165ndash9275 MHz in Japan China South Ko-

ea and Singapore also have specific channelisations Channel widths

f 1 MHz and 2 MHz have been adopted although 4 8 and 16 MHz

re also supported in some countries IEEE 80211ah furthermore pro-

oses new PHY and MAC layers The IEEE 80211ah PHY layer can be

onsidered to some extent a sub-1 GHz version of the IEEE 80211ac

ne At the physical layer OFDM is the chosen modulation method us-

ng 32 or 64 tonessub-carriers that are spaced by 3125 kHz The sup-

orted modulations include BPSK QPSK and from 16 to 256-QAM A

road range of antenna technologies ranging from single-user beam-

orming to MIMO and DL-MU-MIMO which was first introduced in

he IEEE 80211ac amendment are also included in the IEEE 80211ah

pecification Similarly the IEEE 80211ah MAC protocol include most

f IEEE 80211 main characteristics further extending its power sav-

ng (PS) mechanisms

11 Novel features

This section overviews the extensions introduced by the IEEE

0211ah amendment to the IEEE 80211 PS mechanisms to account

or the specific characteristics of resource-constrained sensor and ac-

uator devices A more detailed review can be found in [110] includ-

ng a performance assessment of IEEE 80211ah in several of the key

cenarios for M2M communications such as agriculture and animal

onitoring smart metering and industrial automation plants In ad-

ition a detailed survey of the IEEE 80211ah is reported in [112]

12 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 8 IEEE 80211ah PS mode for TIM-STAs non-TIM STAs and Unscheduled STAs

a

d

w

i

a

a

c

m

i

a

e

t

s

m

t

t

u

b

i

I

s

d

f

a

b

s

a

b

S

T

c

t

w

I

4

a

a

which addresses aspects not considered in [110] such as the use of

sectorisation to avoid overlapping between multiple IEEE 80211ah

WLANs and the use of relays to further extend the IEEE 80211ah cov-

erage among other aspects Finally in [113] both IEEE 80211ah PHY

and MAC characteristics are introduced with emphasis on the ben-

efits of using the sub-1 GHz ISM band in terms of link budget the

location of the OFDM pilots for outdoors operation the use of bidi-

rectional transmission opportunities and the new packet formats to

minimise overheads among others

Enhanced power saving mechanism PS mechanisms for WLANs

were already considered in the development of the first IEEE 80211

standard with the goal of improving the lifetime of battery equipped

devices [114] In PS mode STAs keep the transceiver in sleeping mode

as much time as possible They periodically wake up to listen to the

beacons transmitted by the AP Those beacons indicate whether an

STA has packets waiting for it at the AP In the positive case that STA

remains awake and requests the delivery of those packets Otherwise

given it has nothing to receive it returns to sleep mode until the next

beacon is expected

In the IEEE 80211ah amendment time is divided into pages DTIM

(Delivery Traffic Indication Map) periods TIM (Traffic Indication Map)

periods and slots DTIM and TIM periods begin with the correspond-

ing DTIM and TIM beacons sent by the AP The functions of DTIM and

TIM beacons are described below

1 DTIM beacons They inform as to which TIM Groups (ie the

group of STAs assigned to the same TIM period) have pending

packets at the AP

2 TIM beacons Each TIM message informs a TIM Group about

which specific STA has pending data in the AP Between two

consecutive DTIMs there are as many TIM beacons as TIM

Groups

Using this DTIMTIM-based approach any STA can enter into a

power saving state if it does not have packets pending for transmis-

sion and one of two conditions is met (1) it observes in the DTIM

beacon that there is no downlink traffic addressed to its TIM Group or

(2) it observes in the DTIM beacon that there is some downlink traffic

addressed to its TIM Group but that STA does not explicitly appear in

the TIM beacon Compared to the preliminary IEEE 80211 PS mecha-

nism this approach reduces the size of the Traffic Indication Map in

each TIM beacon thus reducing the overhead and the time STAs need

to listen and process them In addition TIM periods can be organised

in pages which further increases the number of TIM groups between

two DTIM beacons

The temporal organisation of pages DTIM and TIM periods is re-

ported in Fig 8 which shows the division of TIM periods in RAW (Re-

stricted Access Window) and PRAW (Periodic RAW) The RAW is a

time interval in each TIM period where TIM stations can transmit and

receive data (see below the different types of STAs defined in IEEE

80211ah) In addition it can be divided into several downlink and

uplink slots for further granularity In the downlink the slots are as-

signed to a single STA or a group of STAs while the uplink slots are

randomly selected by the STAs with packets ready for transmission

The PRAW is the period of time in each TIM where non-TIM stations

can transmit and receive data

Types of STAs IEEE 80211ah supports three types of STAs TIM

non-TIM and Unscheduled STAs A TIM station is assigned to a TIM

Group Their data transmissions must be performed within a RAW

(Restricted Access Window) period Non-TIM stations do not have to

listen to beacons to transmit data During the association process

non-TIM devices directly negotiate with the AP to obtain a trans-

mission time allocated in a PRAW The following channel access can

either be renegotiated or occurs periodically depending on the re-

quirements set by the STA Unscheduled STAs do not need to listen to

any beacons similar to non-TIM stations Even inside any restricted

access window unscheduled STAs can send a poll frame to the AP

sking for immediate access to the channel The response frame in-

icates an interval (outside both restricted access windows) during

hich unscheduled stations can access the channel This procedure

s meant for STAs that transmit data very sporadically

Hierarchical station organisation To support a large number of STAs

nd their organisation in pages DTIM and TIM periods IEEE 80211ah

ssigns to each associated STA a unique identifier of 13 bits which is

alled the Association Identifier (AID) Using this new AID the maxi-

um number of supported STAs is increased from the original 2007

n IEEE 80211 to 8191 (= 213 minus 1) in IEEE 80211ah However it also

llows categorising STAs according to the type of application they are

xecuting their power level or even their desired QoS by assigning

hem to different TIM groups

Long sleeping periods IEEE 80211ah offers TIM Non-TIM and Un-

cheduled STAs the possibility to set very long doze times (up to

onths) The corresponding clock drift produced by such long doze

imes must be taken into consideration however as the higher the

ime an STA has been asleep the further in advance it should wake

p to avoid possible synchronisation problems with the network

Efficient small data transmission Three new enhancements have

een proposed to reduce the overhead when the data packet size

s small First while IEEE 80211 contains a 28-byte MAC header

EEE 80211ah proposes a short 18-byte version by using AIDs in-

tead of MAC addresses Second IEEE 80211 has defined several null

ata packet (NDP) frames which consist only of a PHY header These

rames can be used to create short ACKs short Block ACKs short CTSs

nd short PS-Polls Finally a Fast Frame Exchange mechanism has

een developed so that if an STA has data to transmit it can notify a

uccessful reception by transmitting its data frame instead of an ACK

Sectorisation Since the PHY layer is based on the IEEE 80211ac

mendment single and multi-user beam-forming are also supported

y IEEE 80211ah This allows the transmission of data to multiple

TAs simultaneously in the downlink increasing the system capacity

he use of the beam-forming capability of IEEE 80211ah APs is also

onsidered as a means to group STAs into different independent an-

enna sectors with the main goal of reducing interference issues This

ould be particularly useful in the case of overlapping with other

EEE 80211ah WLANs or in the presence of hidden nodes

12 Open challenges

A first open challenge is to understand the coverage and achiev-

ble transmission rates in IEEE 80211ah WLANs in both indoor

nd outdoor scenarios Propagation models for WLANs working at

B Bellalta et al Computer Communications 75 (2016) 1ndash25 13

f

c

8

p

R

p

i

m

l

a

m

c

p

p

f

m

t

a

e

p

t

m

a

l

T

m

a

i

n

t

t

t

u

p

c

w

f

a

t

8

a

m

a

s

l

a

m

b

m

d

S

i

a

o

a

R

t

n

G

fi

s

t

p

t

c

u

c

w

i

o

p

g

a

c

i

f

c

(

d

m

l

i

R

t

l

t

L

m

a

m

o

e

f

a

I

i

t

w

i

t

a

c

w

p

f

t

d

5

i

s

e

s

w

5

a

c

a

n

requencies lower than 1 GHz are evaluated in [115] The authors

ompare two path loss propagation models proposed by the IEEE

0211ah Task Group (one for macro and one for picohotzone de-

loyments) [116] with Lee and HatandashOkumura propagation models

esults show that the IEEE 80211ah channel models underestimate

ath loss with respect to Lee and Hata models Moreover in a compar-

son with empirical data it is observed that the IEEE 80211ah channel

odels also underestimate the initial loss and the slope of the path-

oss function A new model parameterisation is thus proposed by the

uthors In [117] the feasibility of an IEEE 80211ah WLAN deploy-

ent is also evaluated in terms of the achievable range and bit rate

omputed on the basis of the link budget using the same path loss

ropagation models as in [115] Results show that the transmission

ower limitations in the uplink can limit the overall network per-

ormance Finally in [118] the authors evaluate the achievable trans-

ission range for the different transmission rates and the achievable

hroughput for different combination of transmission power values

nd transmission rates Further studies are required due to the het-

rogeneity of scenarios in which IEEE 80211ah WLANs can be de-

loyed to characterise the effective coverage ranges with special at-

ention to mixed indoors and outdoors scenarios and in presence of

obile nodes

Due to the wide coverage IEEE 80211ah WLANs may be severely

ffected by the presence of hidden terminals To mitigate such a prob-

em smart solutions to guarantee that the STAs assigned to the same

IM Group are inside each other transmission range or the employ-

ent of RTSCTS-based solutions may contribute to mitigate such

problem An overview of long-range scenarios for IEEE 80211ah

s presented in [119] including a detailed description of the hidden

ode problem in them

Single-hop uplink transmissions from distant STAs require high

ransmission power to reach the AP Since not all STAs may be able

o transmit at the required power a solution could be the introduc-

ion of nodes able to relay transmissions from them However the

se of relays has to be efficiently harmonised with the operation in

ower saving mode allowing temporal periods in which the relays

an gather the data from their associated STAs and then periods in

hich they forward the data to the AP which is still an open challenge

or IEEE 80211ah WLANs In case relays are used data prediction and

ggregation techniques could be implemented to make more efficient

ransmissions [120]

Another challenge is the assessment of the efficiency of IEEE

0211ah PS mechanisms including their optimisation In [111] the

uthors evaluate the capacity of the TIM and page segmentation

echanisms in terms of the maximum number of nodes supported

nd the energy consumed given a certain network traffic profile Re-

ults confirm that a large number of STAs can be supported with

ow energy consumption The impact of the number of nodes oper-

ting in PS mode on the energy consumption and the delay perfor-

ance is analysed in [121] A detailed analysis of the TIM Group-

ased channel access adopted by the IEEE 80211ah task group is

ade in [122] where the STAs assigned to each TIM are uniformly

istributed between the slots of the RAW Since only one group of

TAs is allowed to transmit in each RAW slot the channel contention

s minimised The number of downlink and uplink slots in a RAW

s well as their duration is optimised on the basis of the number

f active STAs and the network traffic profile in [123] In [124] the

uthors propose an algorithm to determine the optimal size of the

AW interval for uplink transmissions based on the estimation of

he number of STAs transmitting and the duration of the RAW Fi-

ally in [125] the authors address the contention problems in Smart

rid communication networks with many nodes and periodic traf-

c when using an IEEE 80211ah WLAN It is still a challenge to con-

ider smart systems able to adapt the IEEE 80211ah parameters to

he instantaneous system state in order to save as much energy as

ossible

Since the PS mechanisms affect only the operation of TIM STAs

he performance of non-TIM and Unscheduled STAs has not yet been

onsidered in the literature To identify the scenarios in which the

se of non-TIM and Unscheduled STAs are of interest is still an open

hallenge as further investigating how non-TIM STAs negotiate over

hich PRAWs they can use to transmit and receive data Co-existence

ssues between the three types of STAs in a single network is also an

pen challenge that requires further work to guarantee a minimum

erformance level for all of them

The design of strategies to distribute the STAs between all the TIM

roups based on their specific traffic profile position battery level

nd application priority is another important challenge that is still

ompletely open EDCA is the default channel access scheme included

n the IEEE 80211ah amendment and provides some basic traffic dif-

erentiation capabilities at the packet level Besides the different ac-

ess categories (ACs) needing to be renamed and their parameters

Arbitration Inter-Frame Spacing or AIFS CWmin TXOP duration) up-

ated to fit the specific traffic profiles of M2M communications other

echanisms can be implemented A mechanism to assign the down-

ink RAW slots when packets from multiple priority levels are wait-

ng for transmission at the AP is required for example Also uplink

AW slots are currently selected randomly by the STAs that want to

ransmit a packet A mechanism to reserve some slots at each up-

ink RAW for high priority STAs may therefore be an option although

his may severely degrade the performance of the low priority STAs

astly different repetition patterns for different priority TIM groups

ay also allow priority STAs assigned to those priority TIM groups to

ccess the channel more often We expect this challenge will receive

uch attention in the upcoming years because of the heterogeneity

f sensors that will be connected though a single IEEE 80211ah AP

specially in urban scenarios

Similar to what has been discussed for 4G cellular networks used

or M2M communications [126] the development of mechanisms to

void and resolve congestion situations is another open challenge for

EEE 80211ah WLANs with many nodes For instance a mechanism

n which STAs have to wait a random number of DTIM periods before

hey can start a transmission may be implemented This approach

ould effectively distribute the traffic over a longer period of time

n overload conditions but otherwise would unnecessarily increase

he access delay Adaptive and load-aware solutions may play again

n important role to keep the optimal WLAN operation in all possible

ircumstances

Finally since IEEE 80211ah will compete with 4G5G cellular net-

orks and WSNs to provide M2M connectivity in many different ap-

lication domains such as smart cities or e-health comparative per-

ormance studies to determine the strong and weak points of each

echnology are required including also aspects such as the cost of the

evices and the system reliability

Cognitive radio technology for TV White Spaces

This section gives an overview of the IEEE 80211af amendment

ntroducing its most relevant features and open challenges Fig 9

hows the basic components and features considered for WLANs op-

rating in the TV White Spaces (TVWS) which include local spectrum

ensing by the AP and STAs and the use of geolocation data bases

ith information on channel availability

1 The IEEE 80211af amendment

Thanks to the transition from analog TV to digital TV several VHF

nd UHF spectrum channels used for decades for analog TV broad-

asting are now unused or are under-utilised in many geographical

reas [127ndash129] This ldquodigital dividendrdquo of spectrum has suggested

ational regulators such as the FCC in the US and Ofcom in the UK

14 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 9 WLAN operating in TVWS basic elements and functionalities

Fig 10 Hidden terminal problem in TVWS when the secondary network uses only in-

stantaneous channel sensing to decide whether or not a TV channel is occupied When

the STA is not able to detect the TV signal due to the presence of obstacles it may

initiate a transmission and create interference to the primary users

s

c

5

a

i

p

t

o

o

t

e

c

u

l

m

b

a

m

t

e

t

t

t

m

D

c

I

n

b

u

8

c

a

t

i

t

w

f

to discuss how to reuse these channels for unlicensed devicesrsquo com-

munications [130131] These potentially vacant channels in the VHF

and UHF bands are referred to as TVWS [132] and include spectrum

portions like the 470ndash790 MHz in Europe and non-contiguous 54ndash72

76ndash88 MHz 174ndash216 MHz 470ndash698 MHz and 698ndash806 MHz in USA

A snapshot of the TV spectrum occupancy in the city of Barcelona in

2012 is shown in [133]

The attractive characteristics of TVWS (not only for WLANs) in-

clude the ability to penetrate through walls and other obstacles much

more effectively than other widely used spectrum bands such as the

24 and 57 GHz ISM bands [134ndash137] This fact along with a progres-

sive spectrum scarcity in ISM bands has suggested the birth of the

IEEE 80211af amendment published in February 2014 [138] which

provides the IEEE 80211 operational characteristics for TVWS access

of unlicensed White Space Devices (WSD) A good summary of IEEE

80211af can be found in [137] The main advantage of operating IEEE

80211 WLANs in the TVWS comes from an increased coverage range

which can reach up to 1 km in rural areas and open fields [139]

and less energy needed to transmit However this comes at the cost

of an increased interference risk to other WSDs which creates co-

existence problems and demands new PHY and MAC layers to ef-

ficiently support channel access and operations preserving licensed

usersrsquo devices

WLANs operated in the TVWS could cover a number of interest-

ing and emerging use cases and scenarios eg Internet access in ru-

ral or sparsely populated areas Smart Grid sensor aggregation me-

tering and control Internet of Things advanced WLAN operations

and TVWS traffic offloading in indoor environments [140141] Nev-

ertheless the concept could be realised to create or extend com-

mercial or municipality WiFi services offered to citizens with cover-

age at the whole city level realised with reasonably limited network

infrastructures [142143]

IEEE 80211af will use the PHY and MAC layers derived from IEEE

80211ac It will adopt concepts such as the OFDM multi-user beam-

forming contiguous and non-contiguous channel bonding and packet

aggregation Among the mandatory and innovative behavioural and

operational parameters the most notable one is the channel acqui-

sition support realised through remote geolocation-based spectrum

allocation databases which maintain the channelsrsquo availability infor-

mation in any given area and time of day providing upon request the

list of free channels available for use

In the following section we highlight and summarise three main

novelties that IEEE 80211af introduces In Section 511 we explain

how the access infrastructure to the remote spectrum database is de-

signed and what the requirements are we discuss the coexistence is-

sues and methodologies adopted in IEEE 80211af and we also discuss

the novel concept of non-contiguous channel bonding The following

ections also contain an illustration of related works Open research

hallenges will be summarised in Section 512

11 Novel features

This section provides an illustration of novel features introduced

nd discussed in the path to IEEE 80211af and the directions provided

n related works and standardisation initiatives to resolve classical

roblems properly characterised in the new framework of TVWS

echnologies

Channel acquisition spectrum database and channel sensing Much

f the attention in operating in the TVWS is given to the protection

f the primary licensed users in the TVWS spectrum band In general

he primary users were considered as the Digital TV (DTV) broadcast-

rs and receivers In fact receivers would suffer the wasteful effect of

ollisions during the TV broadcast reception in the case of secondary

sersrsquo (SU) transmission interference (Fig 10)

It is a well-known problem that the definition of a coherent wire-

ess channel status (idlebusy) from the viewpoint of distributed pri-

ary users (eg DTV receivers) is made difficult by the difference

etween transmitters and receivers interpretation of signals in a vari-

ble time-space collision domain resulting in hiddenexposed ter-

inals concepts Solutions proposed to reduce the hiddenexposed

erminals included the RTSCTS mechanisms and beaconing How-

ver it is difficult to implement solutions that would effectively de-

ermine the channel status of TVWS channels based on localised dis-

ributed sensing activities performed by multiple secondary users in

he same collision domain On the other hand the use of RTSCTS

echanism would not be effective in this case because most of the

TV primary users (and specifically the receivers) were not originally

onceived to transmit signals or implement the RTSCTS handshake

n conclusion the actions to identify free (and busy) TVWS chan-

els should not be based on primary usersrsquo involvement but should

e almost totally implemented by specifically designed secondary

sersrsquo methodologies

There has been much debate on the methodology that IEEE

0211af secondary devices should adopt to get knowledge of free

hannels at a given time on a given target communication area in

n effort to guarantee verified optimised and reliable TVWS spec-

rum use The three main methodologies discussed include sens-

ng solutions geolocation databases (DB) and beaconing Moreover

wo approaches are possible in general distributed and centralised

ith their well known tradeoffs in terms of effectiveness required in-

rastructure deployment and coordination overhead The distributed

B Bellalta et al Computer Communications 75 (2016) 1ndash25 15

m

b

s

c

t

u

a

t

o

l

t

w

t

t

a

o

c

c

a

(

t

o

t

t

D

t

t

o

t

f

a

t

c

i

a

c

d

t

s

t

b

I

l

(

e

a

i

R

p

e

t

c

c

i

b

i

i

t

b

t

t

t

W

a

p

c

i

t

h

p

t

n

a

t

e

t

t

t

p

t

a

r

8

w

s

w

[

i

S

o

a

t

f

i

8

i

r

a

a

i

t

c

(

I

s

a

e

t

t

c

[

i

c

i

g

s

n

s

t

i

o

o

s

c

D

o

ethodology that is receiving more attention in the literature is

ased on distributed channel sensing solutions [127144145] These

olutions include the observation of spectrum use and possible

ooperative aggregation of spectrum sensing information to increase

he accuracy of detection and reduce the vulnerability of primary

sersrsquo transmissions The spectrum sensing capability is required in

ll the secondary devices with a minus114 dBm sensitivity Dynamic

ransmission power control must be provided and the upper limit

n emissions is 100 mW EIRP (20 dBm) for portable devices (further

imited in the case of adjacent channel use to reduce out-of-band in-

erference) However database (DB) coordination of TVWS combined

ith spectrum sensing is considered the most promising and effec-

ive technique compared with spectrum sensing alone The DB spec-

rum information has more chances to be effective (ie identifying

ll free channels opportunities) and reliable (ie not prone to attacks

r erroneous interpretation of channels status) In fact regulators de-

ided to push for the conservative solution of a remotely accessible

entralised spectrum DB that maintains the information on the avail-

bility of all TVWS channels at any given point in time and location

with a target accuracy around 50 m) under their responsibility For

his reason the spectrum databases have become a mandatory part

f many spectrum sharing systems and the dominant technical solu-

ion to support TV white spaces

The IEEE 80211af standard adheres to this vision and also has

he role of a common regulatory framework for different spectrum

B implementations defining a generalised coordination architec-

ure protocols and interfaces for spectrum queries and local spec-

rum information management The protocol explicitly considers out

f scope the communication protocol and technology adopted in

he background leaving the freedom to those players responsible

or the deployment to select any suitable Internet-based and local

ccess network technology Secondary devices that need to access

he TVWS spectrum should get information on which channels they

an effectively use and which parameters to adopt without impact-

ng primary usersrsquo transmissions (and receptions) by querying the

vailable spectrum DB The query must contain the transmission

haracteristics and accurate geographical position of the secondary

evice which can be determined by means of a geo-positioning sys-

em (GPS) regularly updated in the case of movement (or manually

et for static devices) The IEEE 80211af reference system architec-

ure for DB spectrum access is composed of multiple Geolocation-

ased Spectrum information Databases (GDBs) entities connected via

nternet to Registered Location Secure Servers (RLSS) RLSSs work as

ocal proxies of the GDB for a localised group of Basic Service Sets

BSS) RLSS are connected via a secure protocol architecture with the

quivalent of multiple Access Points (AP) in different BSS realising

trusted DB infrastructure APs locally coordinate the exchange of

nformation and channel access management between the GDB (via

LSS) and the secondary usersrsquo end stations (STA)

Mode I devices are those under the control of a device that em-

loys geo-location database access while Mode II devices are those

mploying geo-location database access by themselves The informa-

ion exchange provided between the GDB and a secondary user STA

an be provided in both an open-loop (eg adopted by FCC) and a

losed-loop (eg adopted by ETSI) implementation In an open loop

mplementation daily spectrum availability information is provided

y the GDB and no feedback on the spectrum information received

s provided by STAs thus the system approach for spectrum access

s much more conservative leading to low channelsrsquo utilisation po-

ential In a closed loop implementation the STA can provide feed-

ack to the GDB and there are more communication overheads due

o the high granularity of updates however the system is more effec-

ive and reliable in the exploitation of TVWS on behalf of STAs The

ypical information provided by the GDB includes (1) the updated

hite Space Maps (WSM) of frequencies allowed for secondary use

t the timespace of the querying STA and (2) the device-dependent

ower limitations for transmission (in general conservative and ac-

urate enough to avoid relevant interference effects on primary users

dentified in the area)

The basic IEEE 80211af mechanisms regulating the communica-

ion between STA and GDB can be found in [137138]

In parallel with IEEE 80211af standards such as the IEEE 19006

ave been created that consider interfaces and data structures sup-

orting spectrum sensing information exchange applicable to spec-

rum sensing (and particularly distributed spectrum sensing) sce-

arios In particular in October 2014 the IEEE 19006 WG initiated

project for a new standard called IEEE 19006b [146] concerning

he use of spectrum sensing information to support and optimise the

ffectiveness reliability and robustness of spectrum database solu-

ions The aim is to enhance the performance and capabilities of spec-

rum databases through the use of spectrum sensing information

Co-existence Support for co-existence mechanisms so that mul-

iple technologies can effectively utilise the TVWS spectrum is im-

ortant Self co-existence between network devices of a common

echnology (eg deployed by different operators in the same area)

nd co-existence among different technologies are relevant topic of

esearch for Cognitive Radio (CR) systems and specifically for IEEE

0211af Many solutions appeared on the research scene but no one

as so far finalised as the target solution for IEEE 80211af Specific

tandardisation has been started to regulate coexistence between

ireless standards of unlicensed devices including the IEEE 802191

147] The purpose of the IEEE 802191 standard is to enable the fam-

ly of IEEE 802 Wireless Standards to most effectively use TV White

pace by providing standard coexistence methods among dissimilar

r independently operated TVWS devices Early examples of gener-

lised coexistence mechanisms included Dynamic Frequency Selec-

ion (DFS) Transmission Power Control (TPC) listen before talk (eg

or contention based IEEE 80211 80215) time division multiplex-

ng (also among different techniques such as the IEEE 80216 80220

0222) and Message-based Spectrum Contention (that is beacon-

ng messages that carry coexistence information) Opportune met-

ics must be defined to assess the measurable coexistence achieved

mong different technologies as an example the hidden node prob-

bility for a target scenario or the estimate of percentage variation

n normalised network throughput and latency (before and during

he SU transmissions) On the other hand a centralised coexistence

ontrol mechanism could be effectively realised by a central manager

or coexistence-DB like the GDB) in critical scenarios To this end

EEE 19004 [148] (a standard for heterogeneous networks in dynamic

pectrum context part of IEEE Standards Coordinating Committee 41)

ims to standardise the overall system architecture and information

xchange between the network and mobile devices which will allow

hese elements to optimally choose from available radio resources

There are three possible classification of co-existence archi-

ectures and inter-network coordination channels for CR systems

entralised coordinated and autonomous co-existence mechanisms

149] In centralised co-existence schemes the co-existence is admin-

stered by a central entity (eg like in IEEE 802191) This solution

ould be applied in both homogeneous IEEE 80211af systems (eg

ncarnated by the centralised spectrum DB entity) and in hetero-

eneous inter-networks without requiring modifications to existing

tandards under the assumption that all the involved co-existing

etwork entities would adhere to the same centralised coordination

cheme and inter-network coordination protocol On the other hand

his assumption could be hard to satisfy in some practical scenar-

os The centralised schemes are effective in providing minimisation

f inter-network interference based on the availability and quality

f centralised co-existence information In coordinated co-existence

chemes the centralised entity could be present but not taking de-

isions in general The central entity could provide a co-existence

B with required information which can be used via in-band and

ut-band signalling for co-existence decisions taken by the cluster

16 B Bellalta et al Computer Communications 75 (2016) 1ndash25

c

e

n

c

o

s

i

d

a

S

8

n

a

b

o

b

i

t

a

8

b

[

t

i

n

b

5

t

a

c

b

s

a

c

n

s

p

i

c

l

r

c

f

a

v

t

a

e

c

t

Q

a

b

s

i

c

u

n

c

i

heads of many co-existing inter-networks for implementing a proper

common coordination protocol (eg out-band busy tone signalling)

These are often hybrid solutions realising a compromise between ef-

fectiveness and suitability In autonomous co-existence schemes all the

decisions and coordination are implemented in a distributed way

by the involved network entities In these schemes many policies

can be adopted to realise a sufficient level of co-existence under a

best effort approach Solutions that could be implemented include

the use of busy tones beaconing and other signalling protocols in

dedicated control channels dynamic distributed frequency selection

schemes listen before transmit token-based and dynamic reserva-

tion schemes All the above mentioned solutions have goods and

bads and a general illustration can be found in [149] Other proac-

tive co-existence techniques try to early detect and recovermitigate

co-existence issues with relaxed inter-network coordination eg re-

alised via spectrum sensing and interference avoidancesuppression

techniques

It must be clear that different aims exist regarding the co-

existence for primary TVWS users protection against secondary

users and for secondary devices mutual interference avoidance

Many co-existence problems have been analysed in the literature

for TVWS technologies in particular between the primary (licensed)

and secondary (unlicensed) devices The coexistence management in

TVWS is complex because primary users of DVB transmissions are in-

tended to be the pure receivers rather than the broadcasters In other

words potential protected users could be everywhere and the hid-

den terminal problem could arise for these systems The co-existence

problem is also complicated by many factors that cause asymme-

try and dynamicity in the TVWS such as the mobility variable den-

sity power asymmetry and heterogeneous MACPHY layers TVWS

co-existence problems originate between the IEEE 80211af technolo-

gies and the IEEE 80222 technologies for wireless regional area net-

works (WRAN) Results of analysis of TVWS usage in Europe show

that white spaces are typically present and fragmented They are typ-

ically more abundant in rural areas where larger contiguous blocks

of unused channels are available due to broadcast network planning

giving priorities linked to population density However the exploita-

tion of TVWS in urban areas is possible eg some recent research

was realised under the assumption of exploiting shadowing effects

created in the communication environment (eg by buildings obsta-

cles) in favour of frequency reuse [150]

Since different standards for opportunistic communication in the

TV White Spaces have now been published such as IEEE 80222 [151]

IEEE 802154m [152] Weightless [153] and of course IEEE 80211af

improved methods to guarantee the coexistence of different devices

operating on several protocols in the same bands must be deployed

We highlight here some work eg [150154ndash157] that focuses on

the coexistence between different technologies operating in the same

bands In particular [155] studies the performance degradation of

IEEE 80222 when an IEEE 80211af network is operated in the same

area The problem becomes even worse when the IEEE 80211af net-

work is located near 80222 user equipment causing both networks

to perceive a strong interference from each other One proposed solu-

tion is the Coexistence Beacon Protocol [158] studied for 80222 net-

works which foresees the exchange of a periodic beacon to identify

the neighbouring and possibly interfering networks

For what concerns the co-existence problems research has ad-

dressed other important related aspects of new WLANs on TV White

Spaces For instance the higher coverage range makes them attrac-

tive for applications in the smart grid and also for M2M on TVWS

[134150159] Here the increased range compared to standard tech-

nologies in the ISM bands can have beneficial effects for indoor

mobile devices [150] and M2M [159] Specifically indoor devices

communicating on TV White Spaces offer better propagation char-

acteristics and penetration through obstacles making it easier to

realise scalable home connectivity [135] although interference and

o-existence problems could be exacerbated and must be resolved

g see [139] However studies in the literature show how the sig-

al coming from an indoor transmitter on TV White Spaces remains

onstrained inside the house making it difficult to deploy indoor-to-

utdoor networks On the other hand this shadowing limit can be

een as an opportunity lowering the interference outside the build-

ng in which the TVWS network is locally operated [134150] Un-

er this approach HDTV streaming has recently been considered as

possible use case of WLAN networks operating in the TV White

pace [160]

Non-contiguous channel bonding A relevant novel feature of IEEE

0211af is the potential for contiguous and non-contiguous chan-

el bonding which permits aggregating basic channels (also non-

djacent ones) and leveraging the possible large frequency spread

etween multiple available channels Due to the rather static nature

f primary DTV transmissions where a busy channel is unlikely to

ecome free in the near future and state changes are coarse-grained

n general it is crucial to exploit the time-locality effect and exploit

he maximum physical channel availability that could be aggregated

t any given location With the methodology inherited from IEEE

0211ac IEEE 80211af is capable of bonding together two up to four

asic channels grouped in up to two different non-contiguous chunks

137] The spectrum bandwidth of a DVB-T basic channel can be ei-

her 6 7 or 8 MHz depending on the country in which the service

s operated As an example this creates a 144 to 168 OFDM chan-

elsrsquo bandwidth potential when up to four 6ndash7ndash8 MHz channels are

onded [137]

12 Open challenges

The deployment of WLANs in the TVWS at their maximum poten-

ial still requires the resolution of open problems Generally speaking

s mentioned in previous sections the challenges for such networks

an be divided into three main categories (1) spectrum sharing

etween opportunistic (secondary) and heterogeneous devices (2)

pectrum sensingmanagement and maximum exploitation of avail-

ble spectrum (potentially enabling the spectrum-on-demand con-

ept) and (3) co-existence and interference mitigation to the primary

etwork (primary user protection) and secondary networks [154]

The overall aim of spectrum sharing techniques is to maximise

eparation to avoid overlapping or contiguous channels operations

rovided that sufficient channels are available in the TVWS spectrum

n a given spacetime scenario [154] In general important research

ontributions have to be realised for the design of proper spectrum al-

ocation methodologies and techniques satisfying multi-factorial QoS

equirements at the system-level and the user-level Novel ideas in-

lude the spectrum sharing and spectrum sensing techniques being

urther divided into cooperative and non-cooperative

Non-cooperative solutions attempt to realise spectrum sharing

nd sensing on a local basis without direct cooperation between de-

ices Examples include the ldquolisten-before-transmitrdquo approach and

ransmission power control to limit the interference and spectrum

llocation policies that are based on local node feedback only Coop-

rative solutions rely on the existence of a common communication

hannel through which devices can tentatively agree on the spec-

rum allocation that provides the desired spectrum separation and

oS Major issues include the definition of a methodology to identify

common channel (or multiple channels mutually shared in space

y pairs of heterogeneous and distributed devices) and to efficiently

hare common information For the access to the spectrum DBs the

dentification of proper common access channels is still under dis-

ussion out-of-TVWS-band cellular communications can be widely

sed given their high coverage also indoor

Trying to avoid the use of complex and resource-hungry coordi-

ation schemes and dedicated control channels rendezvous proto-

ols play a significant role both for co-existence and for establish-

ng dynamic communication channels based on spectrum sharing

B Bellalta et al Computer Communications 75 (2016) 1ndash25 17

R

n

c

t

l

o

m

e

n

i

n

w

n

a

u

s

p

b

s

m

s

a

d

t

b

a

a

m

a

t

u

8

n

t

d

d

m

f

t

n

d

n

e

r

a

h

b

f

a

c

t

u

i

f

t

d

t

s

i

f

I

l

i

a

n

c

d

I

n

o

t

i

t

e

w

t

c

m

t

a

i

t

q

b

c

a

i

l

i

a

v

t

a

a

c

t

i

t

b

t

t

a

D

p

c

s

o

q

r

p

a

g

t

e

u

d

t

m

n

t

s

f

i

1

8

f

T

endezvous is considered a fundamental problem in generalised CR

etworks at the basis of many communication processes eg in-

luding neighbour discovery routing and broadcast Rendezvous pro-

ocols attempt to establish a new link for communication on a se-

ected and agreed frequency band (channel) identified within a set

f available resources (channels) thus creating the basis for com-

unication between two or more SUs This problem is often exac-

rbated by the high (and unknown) numbers of contending SUs high

umber of available channels and their ambiguous boundaries and

dentifiers and tough co-existence requirements satisfaction in cog-

itive radio systems [161] Preliminary approaches for rendezvous

ere based on the existence of a dedicated Common Control Chan-

el (CCC) agreed among all the SUs or the existence of a centralised

gent (decision maker) assigning channels to any pairs of requesting

sers This approach could be considered in the case of IEEE 80211af

olutions based on the spectrum DB implementation however the

roblem is complicated by the fact that the common channel must

e identified and made available for all the requesting SUs at the

ame time even when these are spread over multiple collision do-

ains in space On the other hand the scalability and reliability is-

ues which are caused by the congestion of CCCs and the vulner-

bility risks of centralised DBs motivate the research of alternative

istributed solutions called blind rendezvous protocols [162] Dis-

ributed solutions have been mainly based on beaconing andor slot-

ased channel hopping algorithms exploiting randomisation sensing

nd discoverysynchronisation messages spread in slotted time over

set of candidate available channels [161] These protocols aim to

inimise the convergence time to maximise the use of resources

nd to avoid collisions among SUs by identifying and agreeing on

he use of a given common channel among N non-overlapping and

nambiguously labelled channels The main complication in IEEE

0211af systems is given by the wide availability of many different

on-contiguous spectrum bands which cannot be uniformly quan-

ised in single channels and unambiguously labelled in space in or-

er to realise a common reference domain for all the IEEE 80211af

evices In fact the knowledge of the number of channels their com-

on labels the unique IDs of SUs and the number of SUs contending

or the channel assignment are the key factors determining the effec-

iveness and good properties of blind rendezvous protocols Recently

ew classes of rendezvous protocols have been proposed which are

istributed blind and oblivious Oblivious means that available chan-

els may be identified with different labels on behalf of the differ-

nt involved SUs A number of these distributed blind and oblivious

endezvous protocols have been proposed and have been referenced

nd analysed in [162] To conclude solutions for dynamic spectrum

andoff issues (eg in case PUs suddenly appear in a region) must

e resolved in advance to minimise the latency of re-establishing a

unctional communication channel eg by implementing a proactive

lternative rendezvous channel selection in background to ongoing

ommunication processes [163]

Another challenge is to provide cooperative techniques in a dis-

ributed vs centralised implementation by analysing their convergence

nder variable conditions overheads and tradeoffs In some cases as

n vehicular networks the exploitation of scenario characteristics and

actors such as the constrained mobility could help to realise effec-

ive dissemination of sensing information over an extended and pre-

ictable horizon under a cooperative approach [164ndash166] Coopera-

ion could be helpful in the realisation of the cooperative spectrum

ensing in combination with the Geolocation DB approach Another

nteresting direction of research is the investigation of mutual ef-

ects of coexisting cooperation-based vs non-cooperative techniques

n cooperation-based devices the exchange of information could al-

ow a quick convergence to a solution and avoidance of further

nterference [145] Certainly cooperative techniques have desirable

dvantages over non-cooperative ones However tight time synchro-

isation and high coordination overheads are required The common

hannel for cooperation purposes could become a bottleneck and re-

uce the potential advantages of dynamic spectrum allocation The

EEE 80219 foresees the presence of a shared common control chan-

el (CCC) to which devices need to tune in order to gather the state

f the network Solutions such as [141] propose the use of TVWS for

he CCC

A promising direction for research is the adoption of cluster-

ng schemes for secondary devices associated with a hybrid dis-

ributedcooperative vs centralisednon-cooperative approach in an

ffort to reduce overheads while maximising advantages In this

ay secondary nodes could implement cooperation for sensing func-

ions using properly identified common channels and delegate the

entralised decisions and spectrum-DB management to well instru-

ented cluster-leader nodes This concept is incarnated to some ex-

ent by RLSS in IEEE 80211af More recently there have been propos-

ls to build distributed databases that refresh their contents period-

cally by querying the remote spectrum database [167] Here Mas-

er devices (according to Ofcom terminology) periodically cache the

uery replies from the spectrum database in order to reply to and

roadcast the spectrum availability to neighbour Slave devices

Another challenging direction for research is the definition of ac-

urate models and efficient simulation tools enabling dynamic spectrum

nalysis for both rural and urban areas Recent developments of dig-

tal maps simulation tools and propagation models theoretically al-

ow improving the capability of predicting radio propagation effects

n complex scenarios with an acceptable computation time This en-

bling technology could be used in parallel to spectrum DBs to pro-

ide more accurate forecasting of dynamic frequency allocation in

imespace scenarios Another issue to be taken into account is the

ccuracy of the remote spectrum database regarding the channel

vailability estimation Several works have already shown the inac-

uracy of propagation models due to the complexity of parameters

o be considered and differences between modelled and real scenar-

os [132134168169] Better propagation models will make it possible

o more efficiently estimate the interference between devices and

uild Radio Environment Maps (REM) or White Space Maps (WSM)

hat will possibly lead to a more accurate sharing of the radio spec-

rum [144] To this end current research could focus on how to lever-

ge simulation models in order to build more accurate frequency

Bs

In general the adoption of remote spectrum DBs methodology

ushed by national regulators has made the spectrum sensing pro-

ess optional and conservatively realised (eg power control for

ensing-only devices is limited to below 17 dBm EIRP) Thus much

f the research has focused on techniques and solutions to successfully

uery the remote database also in challenging environments such as

ural areas and indoor However since the queries must provide the

osition of the mobile device the accuracy of positioning for mobile

nd handheld devices still needs more appropriate solutions Trian-

ulation can be used to estimate the position of mobile devices with

he accuracy needed by the current regulations outdoor (50 m) How-

ver a low GPS accuracy is possible indoors [170] Certified man-

al positioning can be provided by technicians for static and indoor

evices although limiting mobility Alternatives to costly infrastruc-

ures for positioning (eg based on short range beaconing devices)

ust be deployed in indoor scenarios

Finally regarding spectrum sensing and interference mitigation

ew challenges come from the heterogeneity and different charac-

eristics that networks operating in the TV White Spaces can have

uch as different transmitting power (4 W for static devices 100 mW

or mobile and portable devices and 40 mW for communication

n channels adjacent to occupied ones) different bandwidths (5

0 15 and 20 MHz are currently foreseen for IEEE 80211af IEEE

02154m can have narrower bandwidths or wider ones) and dif-

erent medium access schemes (protocols that use either CSMA or

DMA are required to co-exist in the TV White Spaces) With all this

18 B Bellalta et al Computer Communications 75 (2016) 1ndash25

d

a

T

i

a

c

t

l

b

o

t

a

d

p

a

r

p

n

a

m

t

p

p

d

d

o

t

i

D

w

p

i

a

a

d

b

fi

d

o

t

l

b

t

b

a

i

e

o

o

c

t

t

w

A

c

M

s

c

l

O

w

g

v

t

in place the problem of determining whether another secondary de-

vice is currently transmitting in the same channel is still an open

issue

Spectrum sensing is a very hot topic in the literature with many

proposals that leverage the cooperation between devices to raise the

accuracy of the detection process eg [144171] The current litera-

ture also offers room for proposals that suggest exploiting sensing to

build more accurate spectrum DBs [164172173] Lastly regarding the

interference mitigation to the primary network regulators provide a

conservative approach based on a remote spectrum database to avoid

high risk for the primary users However in particular for the indoor

scenario the concept of secondary usersrsquo transmission in TV Grey

Spaces has been proposed TV Grey Space identifies busy channels

that are formally used by primary users in the area of interest (eg

at the rooftop level of a building for DVB-T) but which could be con-

sidered usable indoors (eg at the basement level) without causing

effective interference to the primary users [134150174] This concept

is based on the generalised spectrum underlay paradigm [175] The

remaining research challenge is to provide TV Grey space access guar-

anteeing a sufficient amount of protection to the primary user while

providing a satisfactory QoS to opportunistic devices eg through

the use of customised and validated propagation models for indoor

TVWS networks and feedback mechanisms to constantly monitor the

interference generated to the primary network

6 Emerging new trends and technologies

In this last section we review three emerging trends that in our

opinion will have a large influence in the conception of future WLANs

as they change the way WLAN protocols and functionalities are de-

veloped implemented tested and integrated with other wireless

networks

61 Programmable Wireless LANs

Especially in the enterprise environment WLAN deployments

need to support a wide range of functionalities and services This is

intrinsically difficult because of the large number of APs that must be

managed which calls for scalable solutions Typical services include

channel assignment load balancing among APs authentication au-

thorisation and accounting (AAA) policy management support for

client mobility and interference coordination Another problem is the

fact that WLAN clients autonomously take several decisions such as

which APs to associate with when to hand-over etc Therefore sup-

porting roaming clients requires the management of a large number

of association states across several APs which is a challenge if sup-

port for real-time hand-over is desired Typically such management

schemes are centralised and most of them are proprietary such as

WLAN controller solutions from Aruba [176] or Cisco [177] although

the IEEE 80211u amendment has been released to allow mobile users

to seamless roam between WiFi networks with automatic authenti-

cation and handoff [178] For example Dyson [179] enables STAs to

send information such as radio channel conditions to a centralised

controller based on a custom API (eg based on Python) As the con-

troller has a centralised view of the network it can enforce a rich

set of policies to control the network also using historical informa-

tion A demo system has been implemented along with applications

such as airtime reservations for specific clients or optimised hand-

offs However Dyson requires STAs to be modified in order to use

those new services offered by the centralised controller TRANTOR

[180] is another example of a centralised management system that

requires changes of clients in order for the infrastructure to gather

information from them in terms of eg interference measurements

In addition control commands enable the infrastructure to exercise

control such as modifying the transmit power or influencing the as-

sociation procedure Clients still use standard CSMA MAC layer for

ata transmission In contrast to DenseAP [181] or DIRAC [182] which

re also based on a centralised controller but do not modify clients

RANTOR exercises much more control on clients by the use of a ded-

cated API which enables a significantly higher gain in coordination

nd thus capacity CENTAUR [183] is another example of a centralised

ontroller that centrally schedules hidden and exposed terminals in

he downlink while using standard CSMA MAC for uplink traffic and

egacy downlink traffic To achieve good performance it uses a fixed

ack-off packet staggering techniques and a hybrid data path that

nly schedules downlink transmissions towards hidden and exposed

erminals centrally All other traffic is sent using standard DCF with

standard back-off procedure In contrast to previous approaches it

oes not require any modifications to the clients but does require data

lane centralisation and it remains questionable how scalable such

solution is for large WLAN deployments with high PHY layer data

ates beyond 100 Mbps However new approaches for fast data plane

rocessing that are currently explored in the virtualisation commu-

ity which move all the packet processing into user space may be

n option to significantly increase speed Designing such manage-

ent systems raises several interesting research questions (a) Cen-

ralised infrastructures may incur high latency but are more simple to

rogram and maintain while designing a completely distributed ap-

roach that operates close to WLAN devices may result in significant

istributed coordination and consistency problems (b) what level of

istributed control is required in order to support the flexibility needs

f future WiFi based networks supporting a high level of mobility

Traditionally WLAN APs are built on proprietary operating sys-

ems that are tightly coupled with the hardware This design makes

t hard to create new applications on top of such networking devices

espite the fact that protocols and mechanisms are available that

ould greatly improve the utility of existing networks those new

rotocols are not deployed because the closed system design makes

t very difficult to extend their functionality An important aspect for

ll such new approaches is therefore how to provide open interfaces

nd open source to speed up innovation As an example Linux-based

evices are completely open source but in order to increase flexi-

ility much more work is needed in the area of open drivers and

rmware In fact the Atheros based ath9k driver has been the main

river of innovation in the open source community because it is the

nly driver that interworks with open firmware In general solutions

hat modify the OS driver focus mainly on programmable network

evel solutions for WLANs such as channel switching or handovers

etween APs While the ath9k driver is an excellent example in how

o speed up innovation it is still very cumbersome to support flexi-

le MAC engine reprogrammability with ath9k In contrast MAClets

nd Wireless MAC processors (WMP) [184] allow a much more flex-

ble reprogramming of MAC functionalities A MAC processor is an

ntity able to execute general MAC commands that specify the MAC

perations through a software-defined state machine The behaviour

f the MAC protocol can therefore be updated at run-time by simply

hanging the sequence in which those commands are executed (ie

he MAClets) As a proof-of-concept the authors of [184] implement

he proposed MAC processor solution in a commodity WLAN hard-

are card extending the basic DCF in three directions piggy-backing

CKs a pseudo TDMA and the use of multiple channels In [185] a

ontrol framework for this WMP system is also proposed to support

AClet code mobility ie for moving loading and activating MAC

oftware programs embedded into ordinary data packets (akin to the

apsule model of traditional active networks)

The difficulty in re-programming networking hardware has also

ed to the concept of Software Defined Networks (SDNs) based on the

penFlow protocol [186] The main idea of SDNs is to extend net-

orking devices with standardised APIs that allow third-party pro-

rammers to flexibly control the data path In addition SDNs pro-

ide higher level abstractions to network designers and programmers

hrough the use of a centralised control plane offered by a network

B Bellalta et al Computer Communications 75 (2016) 1ndash25 19

Fig 11 The CLOUDMAC architecture VAP stands for Virtual Access Point and WT

stands for Wireless Termination

c

a

c

t

m

W

n

t

t

c

p

t

u

h

a

i

e

s

o

c

a

a

f

T

r

s

a

s

p

o

M

C

i

T

t

fl

A

a

a

i

c

t

s

t

e

I

t

w

H

e

b

I

l

d

o

6

r

W

s

l

a

e

e

a

s

b

c

t

(

S

p

c

F

B

t

p

T

e

t

s

u

m

i

e

o

a

A

p

d

t

t

n

p

i

t

t

f

i

ontroller such as NOX [187] which allows reuse of components such

s topology discovery or network access control for different appli-

ations The SDN concept has recently been applied to WLAN archi-

ecture in order to enable fine grained control over mobility manage-

ent and data forwarding focusing on a programmable enterprise

LAN architecture An important part of the SDN architecture is the

etwork controller which provides a centralised view of (parts of)

he SDN enabled network and uses the OpenFlow protocol to install

he forwarding rules on SDN-enabled devices (routers switches ac-

ess points cellular base stations etc)

ODIN [188] is designed to support programmability in enter-

rise WLAN architecture by separating the association state from

he physical AP They implement Light Virtual Access Points (LVAP)

sing a Split-MAC approach where the infrastructure controls the

andover procedure among different WLAN APs By managing

ssociations through SDN controllers ODIN enables proactive mobil-

ty management and load balancing within the SDN enabled WLAN

nterprise network without the need for changes in the client WLAN

tack or IEEE 80211 MAC layer While ODIN requires agents to reside

n the APs that communicate with the Odin Master within the SDN

ontroller CLOUDMAC [189] is based purely on the concept of SDN

nd virtual APs Similar to ODIN CLOUDMAC implements a Split-MAC

pproach but in addition enables the processing of WLAN MAC layer

rames within a co-located Cloud using so-called Virtual APs (VAPs)

he physical APs in CLOUDMAC are lightweight WLAN APs that are

esponsible only for sending out IEEE 80211 based MAC layer ACKs to

tandard WLAN clients and tunnel WLAN MAC layer frames through

n SDN-enabled enterprise WLAN towards the VAPs As association

tates are kept in the VAPs fast mobility is supported using sim-

le OpenFlow forwarding rules Because of the additional processing

f IEEE 80211 WLAN MAC frames in the co-located Cloud CLOUD-

AC has higher latency than a standard WLAN deployment However

LOUDMAC offers a Webservice based API to third party applications

n order to program the enterprise WLAN and enable new services

he architecture of CLOUDMAC is depicted in Fig 11 It has been ex-

ended in [190] to support QoS management and in [191] to support

exible MAC management frame prioritisation based on IEEE 80211

system based on OpenFlow has been recently proposed in [192] to

llow the station to be associated with multiple APs simultaneously

nd to switch between APs with low overhead

An important aspect to consider in order to enable programmabil-

ty is backwards compatibility Several approaches (eg [180]) require

lients to be modified to utilise the features provided This is difficult

o do in practice because it requires changes in the operating system

oftware of all clients If not all clients can utilise those APIs it is ques-

ionable how usable the new architecture will be and how much ben-

fit in terms of aggregate performance such architecture will allow

n contrast the SDN based approaches are interesting in the sense

hat they do not require changes in the client WLAN stack and work

ith the standard IEEE 80211 MAC layer deployed within the clients

owever it remains questionable how scalable such solutions are For

xample an interesting open research topic is to evaluate the scala-

ility of approaches such as CLOUDMAC [189] In addition processing

EEE 80211 MAC frames within co-located private Cloud requires low

atency support from local Cloud solutions (such as OpenStack) in or-

er to reduce the IEEE 80211 MAC processing time which is an area

f active research

2 Prototyping and testing IEEE 80211 enhancements

Most of the new proposals for next-generation WLANs are cur-

ently only evaluated using mathematical analysis and simulation

hile both analysis and simulation are necessary to characterise and

tudy those enhancements in the initial design phase or to consider

arge-scale scenarios it is difficult to consider all practical aspects of

real-world scenario This can sometimes cause significant differ-

nces between what simulations and real experiments show How-

ver real experiments are challenging because of the high complexity

nd costs of building the new hardware and software for each specific

olution to test

To mitigate implementation complexity there are several flexi-

le hardware platforms where low-level MAC mechanisms can be

ompletely implemented in software Examples based on FPGAs are

he USRP (Universal Software Radio Peripheral) [193] and the WARP

Wireless Open-Access Research Platform) [194] As an alternative

ORA (Microsoft Research Software Radio) [195] works on general

urpose computers by taking some advantage from modern multiple

ore systems

A different approach is provided by OpenFWWF [196] Open-

WWF provides an open CSMACA firmware for specific models of

roadcom chipsets so the resulting firmware can be uploaded and

ested in real commercial hardware OpenFWWF implements a sim-

le State Machine (SM) for controlling the hardware in real time

he SM evolution is driven by a main loop that reacts to events by

xecuting specific handlers When a packet originally prepared by

he Linux kernel is ready in the NIC memory handler Packet_Ready

ets up the radio hardware according to the packet metadata sched-

les the transmission and jumps back to the main loop The Trans-

ission Engine (TXE) then takes care of accessing the channel ie

t decrements the back-off counter according to the DCF rules and

ventually starts the actual transmission This triggers the execution

f the TX_frame_now event that prepares the ACK time-out clock

nd finalises the MAC header If the ACK-frame is received or if the

CK time-out expires and the maximum number of attempts for this

acket is reached handler Update_Params resets the contention win-

ow to its minimum value or otherwise doubles it Finally it loads

he back-off counter with a fresh value

In the following we will review a list of selected papers that use

he OpenFWWF firmware and the WARP or USRP platforms to test

ew proposals for WLANs in real scenarios as examples A first im-

lementation of the groupcast mechanisms defined in IEEE 80211aa

s presented in [197] The authors modify the OpenFWWF firmware

o include those functionalities and evaluate them using a 30 STAs

estbed A second example of the use of OpenFWWF is the collision-

ree MAC protocol presented in [198] In [198] the authors design and

mplement in OpenFWWF a MAC protocol that is able to achieve a

20 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 12 LTE to WIFI offloading

a

n

c

t

[

d

a

W

v

t

t

o

b

f

S

g

l

i

I

s

I

b

n

s

m

a

p

d

c

N

o

v

t

t

n

l

i

a

t

a

collision-free operation by waiting a deterministic time after success-

ful transmissions before to start a new transmission This work shows

the experimental performance of a collision-free MAC (CF-MAC) pro-

tocol for WLANs using commercial hardware The results show that

the proposed CF-MAC protocol leads to a better distribution of the

available bandwidth among users a higher throughput and lower

losses than DCF

In [199] the authors use the WARP platform to test a variant of the

DCF ndash called IEEE 80211ec ndash that substitutes control packets such as

the RTS CTS and ACK with short detectable sequences Since those se-

quences are shorter than the control packets and can be detected cor-

rectly even at lower SNRs values a significant gain in performance is

achieved MIDAS (Multiple-Input Distributed Antenna Systems) [200]

implementation using the WARP platform shows the benefits of dis-

tributing the antennas over the area to cover instead of co-locating

them at the AP in terms of capacity when MU-MIMO is employed

The authors also propose a new MAC protocol to benefit from the

spatial reuse that the DAS allows taking as a basis the IEEE 80211ac

Experimental results show that their proposal is able to achieve up

to 200 gains versus the traditional approach where all antennas are

co-located at the AP

The USRP platform has recently been proposed to develop a first

implementation of IEEE 80211ah in order to obtain a preliminary ex-

perimental performance assessment of such a technology because

no commercial off-the-shelf hardware is yet available as the amend-

ment is still in progress [201] The USRP platform is also used in

[202] to evaluate TIMO (Technology Independent Multi-Output) a so-

lution to deal with high-power non-IEEE 80211 interferers in ISM

bands

To conclude this section it is worth mentioning here also to

MAClets and WMP as given such flexibility to implement and dis-

tribute new MAC protocols and other IEEE 80211 functionalities we

believe that if a use-friendly implementation of such a MAC proces-

sors framework is provided it would significantly contribute to the

development and testing of new MAC enhancements by the research

community

63 CellularWLAN interworking

Public hotspots that offer Internet access over a WLAN using IEEE

80211 technology are now nearly ubiquitous It is forecasted that the

cumulative installed base of WiFi hotspots worldwide will amount

to 551 million by 2018 excluding private hotspots (eg WiFi access

points deployed at home) [203] The sharp increase in the availabil-

ity of public WiFi was initially perceived by mobile cellular operators

as a threat due to the additional competition from wireline Internet

service providers or emerging crowdsourced WiFi networks such as

FON5 However as cellular operators are fighting to cope with the ex-

plosion of mobile data traffic created by the rising use of multimedia

content traffic over mobile devices [15] they are also starting to use

WLANs based on the IEEE 80211 technology to offload data from their

core and access networks In general mobile data offloading refers to

the use of complementary network technologies (in licensed or un-

licensed spectrum) for delivering data originally targeted to cellular

networks Intuitively the simplest type of offloading consists of ex-

ploiting connectivity to existing co-located WiFi networks and trans-

ferring data without any delay (Fig 12) Thus this offloading tech-

nique is known as on-the-spot offloading As a consequence of this

new trend the seamless integration of cellular (eg 3GLTE) and WiFi

technologies has attracted significant research interest in recent years

(see [204] for a survey) a few solutions have already been standard-

ised [205206] and roaming between cellular and WiFi is becoming

increasingly transparent to end users CellularWLAN interworking is

5 httpscorpfoncomen

fl

t

f

lso fostered by the support in the evolving 4G standards of heteroge-

eous network deployments (HetNets) in which the existing macro

ells are complemented with a number of small low-power base sta-

ions with the goal of increasing capacity in highly congested areas

207208] It is envisaged that small cells will be based on 4G stan-

ards (eg pico and femto cells) as well as IEEE 80211 technologies

nd multimode base stations that work simultaneously with LTE and

iFi are already entering the market

It is important to point out that LTE standards already support a

ariety of mechanisms that enable data offloading However most of

he existing solutions such as Local IP Access (LIPA) and Selected In-

ernet IP Traffic Offload (SIPTO) focus on data offloading at the core

f the network [205] For instance LIPA allows an IP-enabled mo-

ile device to transfer data to another device in the same pico or

emto cell without passing through the cellular access network while

IPTO enables the routing of selected IP data flows through different

ateways The only LTE offloading mechanism that supports seam-

ess interworking with IEEE 80211-based WLANs is IP Flow Mobil-

ty and Seamless Offload (IFOM) Specifically IFOM relies on Mobile

Pv6 technologies to allow a user terminal to simultaneously route

elected IP flows over different radio access technologies [209210]

n this way a user terminal can offload selected flows to a WLAN

ased on some operator-defined policy while keeping the LTE con-

ection running However to enable such an approach it is neces-

ary to support an entity in the cellular core network that can com-

unicate with the user terminal to exchange information about the

vailability and quality of neighbouring access networks as well as to

rovide the user terminal with predefined rules to manage the han-

over process This entity in current LTE standards is called the Ac-

ess Network Discovery and Selection Function (ANDSF) server [211]

ote that the problem of selecting the best communication technol-

gy in an heterogeneous wireless network has been extensively in-

estigated in the literature both with centralised as well as decen-

ralised approaches (see [212] for a survey) However the use of mul-

iple interfaces in parallel as well as per-flow offloading are relatively

ew concepts Thus the design of scalable and efficient network se-

ection strategies for the ANDSF framework is still an open issue For

nstance three offloading methods suitable for the ANDSF framework

re proposed in [213] which are based on coverage SNR and sys-

em load A reinforcement learning approach is designed in [214] that

llows multimode base stations to autonomously steer their traffic

ows across different access technologies depending on the traffic

ype the usersrsquo QoS requirements the network load and the inter-

erence levels A number of studies have also tried to quantify the

10 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 6 Stream classification and inter-AC traffic prioritisation

a

l

t

r

n

a

t

n

i

t

i

t

r

d

p

t

t

e

m

b

t

a

i

s

o

s

a

e

t

i

f

e

m

8

t

I

i

fi

which was finalised in 2012 included several of these proposed im-

provements Significant research efforts have focused on improving

the transmission reliability of multicasting by integrating ARQ mech-

anisms in IEEE 80211-based multicast transmissions Modifications

to the legacy MAC protocol were proposed in [78] to enable the

RTSCTS option in multicast mode and to select one or more multi-

cast receivers (called leaders) for acknowledging multicast data pack-

ets However these enhancements require changes to the standard

specifications The main problems of leader-based ARQ schemes are

leader election and the trade-off between scalability and reliability

The authors in [79] propose selecting the multicast recipient operat-

ing in the worst channel conditions as the unique leader but this ap-

proach may perform poorly in lossy environments In the Batch mode

multicast MAC (BMMM) [80] all multicast recipients are polled by

the multicast originator to send individual ACKs but this scheme is

not suitable for large multicast groups The Enhanced Leader Based

Protocol (ELBP) is proposed in [81] on the basis of multiple ACK-

leaders and block acknowledgement techniques Analytical models

are then developed to help select optimal ACK-leaders to meet ap-

plication QoS requirements However the models apply only to sat-

urated traffic while multimedia streams are typically bursty Another

class of reliable multicast protocols relies on busy tones to reduce

packet losses due to collisions [82] but the additional radio inter-

face needed for the busy tone limits the practicality of such solu-

tions An alternative approach to avoid collisions of multicast pack-

ets is the multicast collision prevention (MCP) scheme [83] which

is based on the use of a shorter waiting time for transmitting mul-

ticast packets An interesting approach is also proposed in [84] to

retransmit lost packets using an online linear XOR coding algorithm

However a modification to the standard MAC protocol is required to

enable simultaneous ACK transmissions In summary several differ-

ent methods have been proposed to improve multicast transmission

reliability by integrating ARQ schemes into the protocol architecture

but there are not conclusive results on which is the best solution The

choice of the most efficient mechanism depends on a variety of in-

terdependent factors such as loss ratios channel congestion multi-

cast group size and QoS requirements of multimedia streams A com-

prehensive analytical framework is needed to optimise the setting of

the parameters for each scheme and to dynamically select the best

one

As discussed above one main difference between unicast services

nd multicast services in the legacy IEEE 80211 standard was the

ack of acknowledgements Another critical difference is that mul-

icast frames must be transmitted using a fixed rate in the basic

ate set while the transmission rate of unicast frames can be dy-

amically adapted to the channel and traffic conditions [85] Thus

group of research papers has investigated the use of rate adap-

ation to improve the throughput of multicast services in IEEE 80211

etworks [7186ndash89] For instance the authors in [86] propose us-

ng RTS frames to allow group members to estimate channel condi-

ions Each member will then send a dummy CTS frame with a length

nversely proportional to channel quality In this way the multicast

ransmitter can use the collision duration to predict the lowest data

ate that can be used for group transmissions The overhead intro-

uced by this mechanism is quite high however The solution pro-

osed in [88] called ARSM also relies on feedback messages sent by

he multicast receivers called multicast response frames to identify

he group member exhibiting the poorest channel conditions How-

ver in this case a different back off timer is associated with each

ulticast receiver depending on the SNR of previously received feed-

ack messages in order to prevent collision An approach similar to

he one employed in the Auto Rate Fallback (ARF) protocol a rate

daptation scheme originally proposed in [90] is used in [87] Specif-

cally the number of successful consecutive transmissions and con-

ecutive transmission failures are used to decide when to increase

r decrease the transmission data rate respectively A modified ARF

cheme is also proposed in [71] which can be applied to videos that

re encoded into two layers namely the base and enhancement lay-

rs However how to integrate rate adaptation with the different re-

ransmission policies that are defined in IEEE 80211aa is still an open

ssue

One research area that is expected to be crucial in the success-

ul development of IEEE 80211aa-based products is the design of

fficient scheduling algorithms for supporting voicevideo traffic Al-

ost all research work in this field has been triggered by the IEEE

0211e amendment that enhanced the original IEEE 80211 MAC with

wo new QoS-aware access mechanisms ie EDCA and HCCA [91]

n principle with a well-designed admission control and schedul-

ng scheme HCCA is able to provide hard QoS guarantees to traf-

c flows [9293] However HCCA is rarely implemented in IEEE

B Bellalta et al Computer Communications 75 (2016) 1ndash25 11

8

c

c

a

s

b

w

o

j

t

f

a

Q

t

t

A

p

t

s

a

f

s

H

a

i

c

i

t

s

f

s

s

m

i

t

s

l

t

t

[

t

i

o

b

r

W

v

m

8

s

i

o

p

q

4

c

e

t

d

a

t

m

Fig 7 WLANs for M2M communications STAs represent sensor and actuator devices

4

t

w

s

a

c

g

e

(

p

o

t

p

9

r

o

a

p

c

o

i

p

b

f

t

s

o

i

4

8

f

t

i

s

m

d

0211e-based WLANs owing to its higher complexity and cost con-

erns Instead EDCA is widely adopted Most papers have thus fo-

used on improving EDCA performance Many papers have proposed

nalytical models for various subsets of EDCA functionalities For in-

tance a saturation-based performance analysis is conducted in [94]

y differentiating the minimum back-off window size the back-off

indow-increasing factor and the retransmission limit The authors

f [9596] also model AIFS differentiation while the model in [97]

ointly captures all the four EDCA parameters for traffic differen-

iation More recent papers have analysed the EDCA performance

or non-saturated conditions and for arbitrary buffer sizes [98] The

uthors in [99] have developed an analytical model to predict the

oS levels that can be achieved once a new voicevideo flow is in-

roduced in the WLAN A Kalman filter is proposed in [100] to ob-

ain estimates on the number of active transmission queues of each

ccess Category in EDCA These analytical models can then be ex-

loited to derive the optimal configuration of the EDCA parameters

o achieve given performance criteria or to design admission control

chemes that preserve QoS constraints For instance a scheme that

ssigns contention-window values to achieve pre-defined weighted-

airness goals is proposed in [101] A control-theoretic scheme is de-

igned in [102] with the goal of minimising the video traffic delay

owever most of these solutions rely on non-realistic assumptions

bout video traffic dynamics An alternative class of solutions dynam-

cally updates the EDCA parameters based on the observed network

onditions In [103] the EDCA parameters are optimised consider-

ng a WLAN with rigid and elastic traffic simultaneously analysing

he interactions between both types of traffic The authors in [104]

pecify several bandwidth-sharing mechanisms with guaranteed QoS

or voice and video traffic Measurement-based admission control

chemes are proposed in [105] A TXOP adaptation method is de-

cribed in [106] that takes into account video frame sizes and trans-

it queue lengths However the main drawback of these solutions

s that they are based on heuristics and hence do not ensure op-

imal and guaranteed performance Finally a third category of re-

earch papers tries to improve video performance by designing cross-

ayer scheduling approaches Specifically these works take advan-

age of multi-layer video encoding to classify the frames according

o their importance and assign them to different access categories

107] For instance the authors in [108] define classifiers and waiting

ime priority schedulers that dynamically change the packet prior-

ties according to end-to-end delay measurements A disadvantage

f this approach however is that an additional adaptation layer may

e needed to implement the complex interactions that are typically

equired between the video coding applications and the MAC layer

e conclude this section by pointing out that existing studies pro-

ide the basic design principles and techniques for improving multi-

edia streaming performance in IEEE 80211 networks Still the IEEE

0211aa standard poses new research challenges that have not been

ufficiently explored and that will require innovative solutions For

nstance scheduling between primary and alternate queues is still an

pen research area as the mapping of individual frames to multi-

le queues in order to achieve graceful degradation of voicevideo

uality [16]

Sensor Networks and machine-type communications

As discussed in Section 2 M2M communications refer to any

ommunication technology that enables sensoractuator devices to

xchange information and perform actions without the manual assis-

ance of humans This section reviews the main features currently un-

er consideration in the development of the upcoming IEEE 80211ah

mendment which targets the main challenges of those networks as

he IoT in general such as the energy consumption or the manage-

ent of many devices

1 The IEEE 80211ah amendment

The IEEE 80211ah amendment [109] aims to provide WLANs with

he ability to both manage a large number of heterogeneous STAs

ithin a single BSS and minimise the energy consumption of the

ensor-type battery-powered STAs

The initial design requirements of the IEEE 80211ah amendment

re detailed in [110] these entail the support of up to 8192 STAs asso-

iated with a single AP the adoption of efficient power saving strate-

ies a minimum data rate of 100 kbps the operation in the license-

xempt sub 1 GHz band and a coverage up to 1 km in outdoor areas

see Fig 7 for an illustrative example) A preliminary assessment of

erformance of the IEEE 80211ah technology in terms of the number

f STAs that can be effectively supported in a single WLAN as well as

heir energy consumption is presented in [111]

IEEE 80211ah operates over different sub-1 GHz ISM bands de-

ending on country regulations 863ndash868 MHz in Europe 902ndash

28 MHz in the US and 9165ndash9275 MHz in Japan China South Ko-

ea and Singapore also have specific channelisations Channel widths

f 1 MHz and 2 MHz have been adopted although 4 8 and 16 MHz

re also supported in some countries IEEE 80211ah furthermore pro-

oses new PHY and MAC layers The IEEE 80211ah PHY layer can be

onsidered to some extent a sub-1 GHz version of the IEEE 80211ac

ne At the physical layer OFDM is the chosen modulation method us-

ng 32 or 64 tonessub-carriers that are spaced by 3125 kHz The sup-

orted modulations include BPSK QPSK and from 16 to 256-QAM A

road range of antenna technologies ranging from single-user beam-

orming to MIMO and DL-MU-MIMO which was first introduced in

he IEEE 80211ac amendment are also included in the IEEE 80211ah

pecification Similarly the IEEE 80211ah MAC protocol include most

f IEEE 80211 main characteristics further extending its power sav-

ng (PS) mechanisms

11 Novel features

This section overviews the extensions introduced by the IEEE

0211ah amendment to the IEEE 80211 PS mechanisms to account

or the specific characteristics of resource-constrained sensor and ac-

uator devices A more detailed review can be found in [110] includ-

ng a performance assessment of IEEE 80211ah in several of the key

cenarios for M2M communications such as agriculture and animal

onitoring smart metering and industrial automation plants In ad-

ition a detailed survey of the IEEE 80211ah is reported in [112]

12 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 8 IEEE 80211ah PS mode for TIM-STAs non-TIM STAs and Unscheduled STAs

a

d

w

i

a

a

c

m

i

a

e

t

s

m

t

t

u

b

i

I

s

d

f

a

b

s

a

b

S

T

c

t

w

I

4

a

a

which addresses aspects not considered in [110] such as the use of

sectorisation to avoid overlapping between multiple IEEE 80211ah

WLANs and the use of relays to further extend the IEEE 80211ah cov-

erage among other aspects Finally in [113] both IEEE 80211ah PHY

and MAC characteristics are introduced with emphasis on the ben-

efits of using the sub-1 GHz ISM band in terms of link budget the

location of the OFDM pilots for outdoors operation the use of bidi-

rectional transmission opportunities and the new packet formats to

minimise overheads among others

Enhanced power saving mechanism PS mechanisms for WLANs

were already considered in the development of the first IEEE 80211

standard with the goal of improving the lifetime of battery equipped

devices [114] In PS mode STAs keep the transceiver in sleeping mode

as much time as possible They periodically wake up to listen to the

beacons transmitted by the AP Those beacons indicate whether an

STA has packets waiting for it at the AP In the positive case that STA

remains awake and requests the delivery of those packets Otherwise

given it has nothing to receive it returns to sleep mode until the next

beacon is expected

In the IEEE 80211ah amendment time is divided into pages DTIM

(Delivery Traffic Indication Map) periods TIM (Traffic Indication Map)

periods and slots DTIM and TIM periods begin with the correspond-

ing DTIM and TIM beacons sent by the AP The functions of DTIM and

TIM beacons are described below

1 DTIM beacons They inform as to which TIM Groups (ie the

group of STAs assigned to the same TIM period) have pending

packets at the AP

2 TIM beacons Each TIM message informs a TIM Group about

which specific STA has pending data in the AP Between two

consecutive DTIMs there are as many TIM beacons as TIM

Groups

Using this DTIMTIM-based approach any STA can enter into a

power saving state if it does not have packets pending for transmis-

sion and one of two conditions is met (1) it observes in the DTIM

beacon that there is no downlink traffic addressed to its TIM Group or

(2) it observes in the DTIM beacon that there is some downlink traffic

addressed to its TIM Group but that STA does not explicitly appear in

the TIM beacon Compared to the preliminary IEEE 80211 PS mecha-

nism this approach reduces the size of the Traffic Indication Map in

each TIM beacon thus reducing the overhead and the time STAs need

to listen and process them In addition TIM periods can be organised

in pages which further increases the number of TIM groups between

two DTIM beacons

The temporal organisation of pages DTIM and TIM periods is re-

ported in Fig 8 which shows the division of TIM periods in RAW (Re-

stricted Access Window) and PRAW (Periodic RAW) The RAW is a

time interval in each TIM period where TIM stations can transmit and

receive data (see below the different types of STAs defined in IEEE

80211ah) In addition it can be divided into several downlink and

uplink slots for further granularity In the downlink the slots are as-

signed to a single STA or a group of STAs while the uplink slots are

randomly selected by the STAs with packets ready for transmission

The PRAW is the period of time in each TIM where non-TIM stations

can transmit and receive data

Types of STAs IEEE 80211ah supports three types of STAs TIM

non-TIM and Unscheduled STAs A TIM station is assigned to a TIM

Group Their data transmissions must be performed within a RAW

(Restricted Access Window) period Non-TIM stations do not have to

listen to beacons to transmit data During the association process

non-TIM devices directly negotiate with the AP to obtain a trans-

mission time allocated in a PRAW The following channel access can

either be renegotiated or occurs periodically depending on the re-

quirements set by the STA Unscheduled STAs do not need to listen to

any beacons similar to non-TIM stations Even inside any restricted

access window unscheduled STAs can send a poll frame to the AP

sking for immediate access to the channel The response frame in-

icates an interval (outside both restricted access windows) during

hich unscheduled stations can access the channel This procedure

s meant for STAs that transmit data very sporadically

Hierarchical station organisation To support a large number of STAs

nd their organisation in pages DTIM and TIM periods IEEE 80211ah

ssigns to each associated STA a unique identifier of 13 bits which is

alled the Association Identifier (AID) Using this new AID the maxi-

um number of supported STAs is increased from the original 2007

n IEEE 80211 to 8191 (= 213 minus 1) in IEEE 80211ah However it also

llows categorising STAs according to the type of application they are

xecuting their power level or even their desired QoS by assigning

hem to different TIM groups

Long sleeping periods IEEE 80211ah offers TIM Non-TIM and Un-

cheduled STAs the possibility to set very long doze times (up to

onths) The corresponding clock drift produced by such long doze

imes must be taken into consideration however as the higher the

ime an STA has been asleep the further in advance it should wake

p to avoid possible synchronisation problems with the network

Efficient small data transmission Three new enhancements have

een proposed to reduce the overhead when the data packet size

s small First while IEEE 80211 contains a 28-byte MAC header

EEE 80211ah proposes a short 18-byte version by using AIDs in-

tead of MAC addresses Second IEEE 80211 has defined several null

ata packet (NDP) frames which consist only of a PHY header These

rames can be used to create short ACKs short Block ACKs short CTSs

nd short PS-Polls Finally a Fast Frame Exchange mechanism has

een developed so that if an STA has data to transmit it can notify a

uccessful reception by transmitting its data frame instead of an ACK

Sectorisation Since the PHY layer is based on the IEEE 80211ac

mendment single and multi-user beam-forming are also supported

y IEEE 80211ah This allows the transmission of data to multiple

TAs simultaneously in the downlink increasing the system capacity

he use of the beam-forming capability of IEEE 80211ah APs is also

onsidered as a means to group STAs into different independent an-

enna sectors with the main goal of reducing interference issues This

ould be particularly useful in the case of overlapping with other

EEE 80211ah WLANs or in the presence of hidden nodes

12 Open challenges

A first open challenge is to understand the coverage and achiev-

ble transmission rates in IEEE 80211ah WLANs in both indoor

nd outdoor scenarios Propagation models for WLANs working at

B Bellalta et al Computer Communications 75 (2016) 1ndash25 13

f

c

8

p

R

p

i

m

l

a

m

c

p

p

f

m

t

a

e

p

t

m

a

l

T

m

a

i

n

t

t

t

u

p

c

w

f

a

t

8

a

m

a

s

l

a

m

b

m

d

S

i

a

o

a

R

t

n

G

fi

s

t

p

t

c

u

c

w

i

o

p

g

a

c

i

f

c

(

d

m

l

i

R

t

l

t

L

m

a

m

o

e

f

a

I

i

t

w

i

t

a

c

w

p

f

t

d

5

i

s

e

s

w

5

a

c

a

n

requencies lower than 1 GHz are evaluated in [115] The authors

ompare two path loss propagation models proposed by the IEEE

0211ah Task Group (one for macro and one for picohotzone de-

loyments) [116] with Lee and HatandashOkumura propagation models

esults show that the IEEE 80211ah channel models underestimate

ath loss with respect to Lee and Hata models Moreover in a compar-

son with empirical data it is observed that the IEEE 80211ah channel

odels also underestimate the initial loss and the slope of the path-

oss function A new model parameterisation is thus proposed by the

uthors In [117] the feasibility of an IEEE 80211ah WLAN deploy-

ent is also evaluated in terms of the achievable range and bit rate

omputed on the basis of the link budget using the same path loss

ropagation models as in [115] Results show that the transmission

ower limitations in the uplink can limit the overall network per-

ormance Finally in [118] the authors evaluate the achievable trans-

ission range for the different transmission rates and the achievable

hroughput for different combination of transmission power values

nd transmission rates Further studies are required due to the het-

rogeneity of scenarios in which IEEE 80211ah WLANs can be de-

loyed to characterise the effective coverage ranges with special at-

ention to mixed indoors and outdoors scenarios and in presence of

obile nodes

Due to the wide coverage IEEE 80211ah WLANs may be severely

ffected by the presence of hidden terminals To mitigate such a prob-

em smart solutions to guarantee that the STAs assigned to the same

IM Group are inside each other transmission range or the employ-

ent of RTSCTS-based solutions may contribute to mitigate such

problem An overview of long-range scenarios for IEEE 80211ah

s presented in [119] including a detailed description of the hidden

ode problem in them

Single-hop uplink transmissions from distant STAs require high

ransmission power to reach the AP Since not all STAs may be able

o transmit at the required power a solution could be the introduc-

ion of nodes able to relay transmissions from them However the

se of relays has to be efficiently harmonised with the operation in

ower saving mode allowing temporal periods in which the relays

an gather the data from their associated STAs and then periods in

hich they forward the data to the AP which is still an open challenge

or IEEE 80211ah WLANs In case relays are used data prediction and

ggregation techniques could be implemented to make more efficient

ransmissions [120]

Another challenge is the assessment of the efficiency of IEEE

0211ah PS mechanisms including their optimisation In [111] the

uthors evaluate the capacity of the TIM and page segmentation

echanisms in terms of the maximum number of nodes supported

nd the energy consumed given a certain network traffic profile Re-

ults confirm that a large number of STAs can be supported with

ow energy consumption The impact of the number of nodes oper-

ting in PS mode on the energy consumption and the delay perfor-

ance is analysed in [121] A detailed analysis of the TIM Group-

ased channel access adopted by the IEEE 80211ah task group is

ade in [122] where the STAs assigned to each TIM are uniformly

istributed between the slots of the RAW Since only one group of

TAs is allowed to transmit in each RAW slot the channel contention

s minimised The number of downlink and uplink slots in a RAW

s well as their duration is optimised on the basis of the number

f active STAs and the network traffic profile in [123] In [124] the

uthors propose an algorithm to determine the optimal size of the

AW interval for uplink transmissions based on the estimation of

he number of STAs transmitting and the duration of the RAW Fi-

ally in [125] the authors address the contention problems in Smart

rid communication networks with many nodes and periodic traf-

c when using an IEEE 80211ah WLAN It is still a challenge to con-

ider smart systems able to adapt the IEEE 80211ah parameters to

he instantaneous system state in order to save as much energy as

ossible

Since the PS mechanisms affect only the operation of TIM STAs

he performance of non-TIM and Unscheduled STAs has not yet been

onsidered in the literature To identify the scenarios in which the

se of non-TIM and Unscheduled STAs are of interest is still an open

hallenge as further investigating how non-TIM STAs negotiate over

hich PRAWs they can use to transmit and receive data Co-existence

ssues between the three types of STAs in a single network is also an

pen challenge that requires further work to guarantee a minimum

erformance level for all of them

The design of strategies to distribute the STAs between all the TIM

roups based on their specific traffic profile position battery level

nd application priority is another important challenge that is still

ompletely open EDCA is the default channel access scheme included

n the IEEE 80211ah amendment and provides some basic traffic dif-

erentiation capabilities at the packet level Besides the different ac-

ess categories (ACs) needing to be renamed and their parameters

Arbitration Inter-Frame Spacing or AIFS CWmin TXOP duration) up-

ated to fit the specific traffic profiles of M2M communications other

echanisms can be implemented A mechanism to assign the down-

ink RAW slots when packets from multiple priority levels are wait-

ng for transmission at the AP is required for example Also uplink

AW slots are currently selected randomly by the STAs that want to

ransmit a packet A mechanism to reserve some slots at each up-

ink RAW for high priority STAs may therefore be an option although

his may severely degrade the performance of the low priority STAs

astly different repetition patterns for different priority TIM groups

ay also allow priority STAs assigned to those priority TIM groups to

ccess the channel more often We expect this challenge will receive

uch attention in the upcoming years because of the heterogeneity

f sensors that will be connected though a single IEEE 80211ah AP

specially in urban scenarios

Similar to what has been discussed for 4G cellular networks used

or M2M communications [126] the development of mechanisms to

void and resolve congestion situations is another open challenge for

EEE 80211ah WLANs with many nodes For instance a mechanism

n which STAs have to wait a random number of DTIM periods before

hey can start a transmission may be implemented This approach

ould effectively distribute the traffic over a longer period of time

n overload conditions but otherwise would unnecessarily increase

he access delay Adaptive and load-aware solutions may play again

n important role to keep the optimal WLAN operation in all possible

ircumstances

Finally since IEEE 80211ah will compete with 4G5G cellular net-

orks and WSNs to provide M2M connectivity in many different ap-

lication domains such as smart cities or e-health comparative per-

ormance studies to determine the strong and weak points of each

echnology are required including also aspects such as the cost of the

evices and the system reliability

Cognitive radio technology for TV White Spaces

This section gives an overview of the IEEE 80211af amendment

ntroducing its most relevant features and open challenges Fig 9

hows the basic components and features considered for WLANs op-

rating in the TV White Spaces (TVWS) which include local spectrum

ensing by the AP and STAs and the use of geolocation data bases

ith information on channel availability

1 The IEEE 80211af amendment

Thanks to the transition from analog TV to digital TV several VHF

nd UHF spectrum channels used for decades for analog TV broad-

asting are now unused or are under-utilised in many geographical

reas [127ndash129] This ldquodigital dividendrdquo of spectrum has suggested

ational regulators such as the FCC in the US and Ofcom in the UK

14 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 9 WLAN operating in TVWS basic elements and functionalities

Fig 10 Hidden terminal problem in TVWS when the secondary network uses only in-

stantaneous channel sensing to decide whether or not a TV channel is occupied When

the STA is not able to detect the TV signal due to the presence of obstacles it may

initiate a transmission and create interference to the primary users

s

c

5

a

i

p

t

o

o

t

e

c

u

l

m

b

a

m

t

e

t

t

t

m

D

c

I

n

b

u

8

c

a

t

i

t

w

f

to discuss how to reuse these channels for unlicensed devicesrsquo com-

munications [130131] These potentially vacant channels in the VHF

and UHF bands are referred to as TVWS [132] and include spectrum

portions like the 470ndash790 MHz in Europe and non-contiguous 54ndash72

76ndash88 MHz 174ndash216 MHz 470ndash698 MHz and 698ndash806 MHz in USA

A snapshot of the TV spectrum occupancy in the city of Barcelona in

2012 is shown in [133]

The attractive characteristics of TVWS (not only for WLANs) in-

clude the ability to penetrate through walls and other obstacles much

more effectively than other widely used spectrum bands such as the

24 and 57 GHz ISM bands [134ndash137] This fact along with a progres-

sive spectrum scarcity in ISM bands has suggested the birth of the

IEEE 80211af amendment published in February 2014 [138] which

provides the IEEE 80211 operational characteristics for TVWS access

of unlicensed White Space Devices (WSD) A good summary of IEEE

80211af can be found in [137] The main advantage of operating IEEE

80211 WLANs in the TVWS comes from an increased coverage range

which can reach up to 1 km in rural areas and open fields [139]

and less energy needed to transmit However this comes at the cost

of an increased interference risk to other WSDs which creates co-

existence problems and demands new PHY and MAC layers to ef-

ficiently support channel access and operations preserving licensed

usersrsquo devices

WLANs operated in the TVWS could cover a number of interest-

ing and emerging use cases and scenarios eg Internet access in ru-

ral or sparsely populated areas Smart Grid sensor aggregation me-

tering and control Internet of Things advanced WLAN operations

and TVWS traffic offloading in indoor environments [140141] Nev-

ertheless the concept could be realised to create or extend com-

mercial or municipality WiFi services offered to citizens with cover-

age at the whole city level realised with reasonably limited network

infrastructures [142143]

IEEE 80211af will use the PHY and MAC layers derived from IEEE

80211ac It will adopt concepts such as the OFDM multi-user beam-

forming contiguous and non-contiguous channel bonding and packet

aggregation Among the mandatory and innovative behavioural and

operational parameters the most notable one is the channel acqui-

sition support realised through remote geolocation-based spectrum

allocation databases which maintain the channelsrsquo availability infor-

mation in any given area and time of day providing upon request the

list of free channels available for use

In the following section we highlight and summarise three main

novelties that IEEE 80211af introduces In Section 511 we explain

how the access infrastructure to the remote spectrum database is de-

signed and what the requirements are we discuss the coexistence is-

sues and methodologies adopted in IEEE 80211af and we also discuss

the novel concept of non-contiguous channel bonding The following

ections also contain an illustration of related works Open research

hallenges will be summarised in Section 512

11 Novel features

This section provides an illustration of novel features introduced

nd discussed in the path to IEEE 80211af and the directions provided

n related works and standardisation initiatives to resolve classical

roblems properly characterised in the new framework of TVWS

echnologies

Channel acquisition spectrum database and channel sensing Much

f the attention in operating in the TVWS is given to the protection

f the primary licensed users in the TVWS spectrum band In general

he primary users were considered as the Digital TV (DTV) broadcast-

rs and receivers In fact receivers would suffer the wasteful effect of

ollisions during the TV broadcast reception in the case of secondary

sersrsquo (SU) transmission interference (Fig 10)

It is a well-known problem that the definition of a coherent wire-

ess channel status (idlebusy) from the viewpoint of distributed pri-

ary users (eg DTV receivers) is made difficult by the difference

etween transmitters and receivers interpretation of signals in a vari-

ble time-space collision domain resulting in hiddenexposed ter-

inals concepts Solutions proposed to reduce the hiddenexposed

erminals included the RTSCTS mechanisms and beaconing How-

ver it is difficult to implement solutions that would effectively de-

ermine the channel status of TVWS channels based on localised dis-

ributed sensing activities performed by multiple secondary users in

he same collision domain On the other hand the use of RTSCTS

echanism would not be effective in this case because most of the

TV primary users (and specifically the receivers) were not originally

onceived to transmit signals or implement the RTSCTS handshake

n conclusion the actions to identify free (and busy) TVWS chan-

els should not be based on primary usersrsquo involvement but should

e almost totally implemented by specifically designed secondary

sersrsquo methodologies

There has been much debate on the methodology that IEEE

0211af secondary devices should adopt to get knowledge of free

hannels at a given time on a given target communication area in

n effort to guarantee verified optimised and reliable TVWS spec-

rum use The three main methodologies discussed include sens-

ng solutions geolocation databases (DB) and beaconing Moreover

wo approaches are possible in general distributed and centralised

ith their well known tradeoffs in terms of effectiveness required in-

rastructure deployment and coordination overhead The distributed

B Bellalta et al Computer Communications 75 (2016) 1ndash25 15

m

b

s

c

t

u

a

t

o

l

t

w

t

t

a

o

c

c

a

(

t

o

t

t

D

t

t

o

t

f

a

t

c

i

a

c

d

t

s

t

b

I

l

(

e

a

i

R

p

e

t

c

c

i

b

i

i

t

b

t

t

t

W

a

p

c

i

t

h

p

t

n

a

t

e

t

t

t

p

t

a

r

8

w

s

w

[

i

S

o

a

t

f

i

8

i

r

a

a

i

t

c

(

I

s

a

e

t

t

c

[

i

c

i

g

s

n

s

t

i

o

o

s

c

D

o

ethodology that is receiving more attention in the literature is

ased on distributed channel sensing solutions [127144145] These

olutions include the observation of spectrum use and possible

ooperative aggregation of spectrum sensing information to increase

he accuracy of detection and reduce the vulnerability of primary

sersrsquo transmissions The spectrum sensing capability is required in

ll the secondary devices with a minus114 dBm sensitivity Dynamic

ransmission power control must be provided and the upper limit

n emissions is 100 mW EIRP (20 dBm) for portable devices (further

imited in the case of adjacent channel use to reduce out-of-band in-

erference) However database (DB) coordination of TVWS combined

ith spectrum sensing is considered the most promising and effec-

ive technique compared with spectrum sensing alone The DB spec-

rum information has more chances to be effective (ie identifying

ll free channels opportunities) and reliable (ie not prone to attacks

r erroneous interpretation of channels status) In fact regulators de-

ided to push for the conservative solution of a remotely accessible

entralised spectrum DB that maintains the information on the avail-

bility of all TVWS channels at any given point in time and location

with a target accuracy around 50 m) under their responsibility For

his reason the spectrum databases have become a mandatory part

f many spectrum sharing systems and the dominant technical solu-

ion to support TV white spaces

The IEEE 80211af standard adheres to this vision and also has

he role of a common regulatory framework for different spectrum

B implementations defining a generalised coordination architec-

ure protocols and interfaces for spectrum queries and local spec-

rum information management The protocol explicitly considers out

f scope the communication protocol and technology adopted in

he background leaving the freedom to those players responsible

or the deployment to select any suitable Internet-based and local

ccess network technology Secondary devices that need to access

he TVWS spectrum should get information on which channels they

an effectively use and which parameters to adopt without impact-

ng primary usersrsquo transmissions (and receptions) by querying the

vailable spectrum DB The query must contain the transmission

haracteristics and accurate geographical position of the secondary

evice which can be determined by means of a geo-positioning sys-

em (GPS) regularly updated in the case of movement (or manually

et for static devices) The IEEE 80211af reference system architec-

ure for DB spectrum access is composed of multiple Geolocation-

ased Spectrum information Databases (GDBs) entities connected via

nternet to Registered Location Secure Servers (RLSS) RLSSs work as

ocal proxies of the GDB for a localised group of Basic Service Sets

BSS) RLSS are connected via a secure protocol architecture with the

quivalent of multiple Access Points (AP) in different BSS realising

trusted DB infrastructure APs locally coordinate the exchange of

nformation and channel access management between the GDB (via

LSS) and the secondary usersrsquo end stations (STA)

Mode I devices are those under the control of a device that em-

loys geo-location database access while Mode II devices are those

mploying geo-location database access by themselves The informa-

ion exchange provided between the GDB and a secondary user STA

an be provided in both an open-loop (eg adopted by FCC) and a

losed-loop (eg adopted by ETSI) implementation In an open loop

mplementation daily spectrum availability information is provided

y the GDB and no feedback on the spectrum information received

s provided by STAs thus the system approach for spectrum access

s much more conservative leading to low channelsrsquo utilisation po-

ential In a closed loop implementation the STA can provide feed-

ack to the GDB and there are more communication overheads due

o the high granularity of updates however the system is more effec-

ive and reliable in the exploitation of TVWS on behalf of STAs The

ypical information provided by the GDB includes (1) the updated

hite Space Maps (WSM) of frequencies allowed for secondary use

t the timespace of the querying STA and (2) the device-dependent

ower limitations for transmission (in general conservative and ac-

urate enough to avoid relevant interference effects on primary users

dentified in the area)

The basic IEEE 80211af mechanisms regulating the communica-

ion between STA and GDB can be found in [137138]

In parallel with IEEE 80211af standards such as the IEEE 19006

ave been created that consider interfaces and data structures sup-

orting spectrum sensing information exchange applicable to spec-

rum sensing (and particularly distributed spectrum sensing) sce-

arios In particular in October 2014 the IEEE 19006 WG initiated

project for a new standard called IEEE 19006b [146] concerning

he use of spectrum sensing information to support and optimise the

ffectiveness reliability and robustness of spectrum database solu-

ions The aim is to enhance the performance and capabilities of spec-

rum databases through the use of spectrum sensing information

Co-existence Support for co-existence mechanisms so that mul-

iple technologies can effectively utilise the TVWS spectrum is im-

ortant Self co-existence between network devices of a common

echnology (eg deployed by different operators in the same area)

nd co-existence among different technologies are relevant topic of

esearch for Cognitive Radio (CR) systems and specifically for IEEE

0211af Many solutions appeared on the research scene but no one

as so far finalised as the target solution for IEEE 80211af Specific

tandardisation has been started to regulate coexistence between

ireless standards of unlicensed devices including the IEEE 802191

147] The purpose of the IEEE 802191 standard is to enable the fam-

ly of IEEE 802 Wireless Standards to most effectively use TV White

pace by providing standard coexistence methods among dissimilar

r independently operated TVWS devices Early examples of gener-

lised coexistence mechanisms included Dynamic Frequency Selec-

ion (DFS) Transmission Power Control (TPC) listen before talk (eg

or contention based IEEE 80211 80215) time division multiplex-

ng (also among different techniques such as the IEEE 80216 80220

0222) and Message-based Spectrum Contention (that is beacon-

ng messages that carry coexistence information) Opportune met-

ics must be defined to assess the measurable coexistence achieved

mong different technologies as an example the hidden node prob-

bility for a target scenario or the estimate of percentage variation

n normalised network throughput and latency (before and during

he SU transmissions) On the other hand a centralised coexistence

ontrol mechanism could be effectively realised by a central manager

or coexistence-DB like the GDB) in critical scenarios To this end

EEE 19004 [148] (a standard for heterogeneous networks in dynamic

pectrum context part of IEEE Standards Coordinating Committee 41)

ims to standardise the overall system architecture and information

xchange between the network and mobile devices which will allow

hese elements to optimally choose from available radio resources

There are three possible classification of co-existence archi-

ectures and inter-network coordination channels for CR systems

entralised coordinated and autonomous co-existence mechanisms

149] In centralised co-existence schemes the co-existence is admin-

stered by a central entity (eg like in IEEE 802191) This solution

ould be applied in both homogeneous IEEE 80211af systems (eg

ncarnated by the centralised spectrum DB entity) and in hetero-

eneous inter-networks without requiring modifications to existing

tandards under the assumption that all the involved co-existing

etwork entities would adhere to the same centralised coordination

cheme and inter-network coordination protocol On the other hand

his assumption could be hard to satisfy in some practical scenar-

os The centralised schemes are effective in providing minimisation

f inter-network interference based on the availability and quality

f centralised co-existence information In coordinated co-existence

chemes the centralised entity could be present but not taking de-

isions in general The central entity could provide a co-existence

B with required information which can be used via in-band and

ut-band signalling for co-existence decisions taken by the cluster

16 B Bellalta et al Computer Communications 75 (2016) 1ndash25

c

e

n

c

o

s

i

d

a

S

8

n

a

b

o

b

i

t

a

8

b

[

t

i

n

b

5

t

a

c

b

s

a

c

n

s

p

i

c

l

r

c

f

a

v

t

a

e

c

t

Q

a

b

s

i

c

u

n

c

i

heads of many co-existing inter-networks for implementing a proper

common coordination protocol (eg out-band busy tone signalling)

These are often hybrid solutions realising a compromise between ef-

fectiveness and suitability In autonomous co-existence schemes all the

decisions and coordination are implemented in a distributed way

by the involved network entities In these schemes many policies

can be adopted to realise a sufficient level of co-existence under a

best effort approach Solutions that could be implemented include

the use of busy tones beaconing and other signalling protocols in

dedicated control channels dynamic distributed frequency selection

schemes listen before transmit token-based and dynamic reserva-

tion schemes All the above mentioned solutions have goods and

bads and a general illustration can be found in [149] Other proac-

tive co-existence techniques try to early detect and recovermitigate

co-existence issues with relaxed inter-network coordination eg re-

alised via spectrum sensing and interference avoidancesuppression

techniques

It must be clear that different aims exist regarding the co-

existence for primary TVWS users protection against secondary

users and for secondary devices mutual interference avoidance

Many co-existence problems have been analysed in the literature

for TVWS technologies in particular between the primary (licensed)

and secondary (unlicensed) devices The coexistence management in

TVWS is complex because primary users of DVB transmissions are in-

tended to be the pure receivers rather than the broadcasters In other

words potential protected users could be everywhere and the hid-

den terminal problem could arise for these systems The co-existence

problem is also complicated by many factors that cause asymme-

try and dynamicity in the TVWS such as the mobility variable den-

sity power asymmetry and heterogeneous MACPHY layers TVWS

co-existence problems originate between the IEEE 80211af technolo-

gies and the IEEE 80222 technologies for wireless regional area net-

works (WRAN) Results of analysis of TVWS usage in Europe show

that white spaces are typically present and fragmented They are typ-

ically more abundant in rural areas where larger contiguous blocks

of unused channels are available due to broadcast network planning

giving priorities linked to population density However the exploita-

tion of TVWS in urban areas is possible eg some recent research

was realised under the assumption of exploiting shadowing effects

created in the communication environment (eg by buildings obsta-

cles) in favour of frequency reuse [150]

Since different standards for opportunistic communication in the

TV White Spaces have now been published such as IEEE 80222 [151]

IEEE 802154m [152] Weightless [153] and of course IEEE 80211af

improved methods to guarantee the coexistence of different devices

operating on several protocols in the same bands must be deployed

We highlight here some work eg [150154ndash157] that focuses on

the coexistence between different technologies operating in the same

bands In particular [155] studies the performance degradation of

IEEE 80222 when an IEEE 80211af network is operated in the same

area The problem becomes even worse when the IEEE 80211af net-

work is located near 80222 user equipment causing both networks

to perceive a strong interference from each other One proposed solu-

tion is the Coexistence Beacon Protocol [158] studied for 80222 net-

works which foresees the exchange of a periodic beacon to identify

the neighbouring and possibly interfering networks

For what concerns the co-existence problems research has ad-

dressed other important related aspects of new WLANs on TV White

Spaces For instance the higher coverage range makes them attrac-

tive for applications in the smart grid and also for M2M on TVWS

[134150159] Here the increased range compared to standard tech-

nologies in the ISM bands can have beneficial effects for indoor

mobile devices [150] and M2M [159] Specifically indoor devices

communicating on TV White Spaces offer better propagation char-

acteristics and penetration through obstacles making it easier to

realise scalable home connectivity [135] although interference and

o-existence problems could be exacerbated and must be resolved

g see [139] However studies in the literature show how the sig-

al coming from an indoor transmitter on TV White Spaces remains

onstrained inside the house making it difficult to deploy indoor-to-

utdoor networks On the other hand this shadowing limit can be

een as an opportunity lowering the interference outside the build-

ng in which the TVWS network is locally operated [134150] Un-

er this approach HDTV streaming has recently been considered as

possible use case of WLAN networks operating in the TV White

pace [160]

Non-contiguous channel bonding A relevant novel feature of IEEE

0211af is the potential for contiguous and non-contiguous chan-

el bonding which permits aggregating basic channels (also non-

djacent ones) and leveraging the possible large frequency spread

etween multiple available channels Due to the rather static nature

f primary DTV transmissions where a busy channel is unlikely to

ecome free in the near future and state changes are coarse-grained

n general it is crucial to exploit the time-locality effect and exploit

he maximum physical channel availability that could be aggregated

t any given location With the methodology inherited from IEEE

0211ac IEEE 80211af is capable of bonding together two up to four

asic channels grouped in up to two different non-contiguous chunks

137] The spectrum bandwidth of a DVB-T basic channel can be ei-

her 6 7 or 8 MHz depending on the country in which the service

s operated As an example this creates a 144 to 168 OFDM chan-

elsrsquo bandwidth potential when up to four 6ndash7ndash8 MHz channels are

onded [137]

12 Open challenges

The deployment of WLANs in the TVWS at their maximum poten-

ial still requires the resolution of open problems Generally speaking

s mentioned in previous sections the challenges for such networks

an be divided into three main categories (1) spectrum sharing

etween opportunistic (secondary) and heterogeneous devices (2)

pectrum sensingmanagement and maximum exploitation of avail-

ble spectrum (potentially enabling the spectrum-on-demand con-

ept) and (3) co-existence and interference mitigation to the primary

etwork (primary user protection) and secondary networks [154]

The overall aim of spectrum sharing techniques is to maximise

eparation to avoid overlapping or contiguous channels operations

rovided that sufficient channels are available in the TVWS spectrum

n a given spacetime scenario [154] In general important research

ontributions have to be realised for the design of proper spectrum al-

ocation methodologies and techniques satisfying multi-factorial QoS

equirements at the system-level and the user-level Novel ideas in-

lude the spectrum sharing and spectrum sensing techniques being

urther divided into cooperative and non-cooperative

Non-cooperative solutions attempt to realise spectrum sharing

nd sensing on a local basis without direct cooperation between de-

ices Examples include the ldquolisten-before-transmitrdquo approach and

ransmission power control to limit the interference and spectrum

llocation policies that are based on local node feedback only Coop-

rative solutions rely on the existence of a common communication

hannel through which devices can tentatively agree on the spec-

rum allocation that provides the desired spectrum separation and

oS Major issues include the definition of a methodology to identify

common channel (or multiple channels mutually shared in space

y pairs of heterogeneous and distributed devices) and to efficiently

hare common information For the access to the spectrum DBs the

dentification of proper common access channels is still under dis-

ussion out-of-TVWS-band cellular communications can be widely

sed given their high coverage also indoor

Trying to avoid the use of complex and resource-hungry coordi-

ation schemes and dedicated control channels rendezvous proto-

ols play a significant role both for co-existence and for establish-

ng dynamic communication channels based on spectrum sharing

B Bellalta et al Computer Communications 75 (2016) 1ndash25 17

R

n

c

t

l

o

m

e

n

i

n

w

n

a

u

s

p

b

s

m

s

a

d

t

b

a

a

m

a

t

u

8

n

t

d

d

m

f

t

n

d

n

e

r

a

h

b

f

a

c

t

u

i

f

t

d

t

s

i

f

I

l

i

a

n

c

d

I

n

o

t

i

t

e

w

t

c

m

t

a

i

t

q

b

c

a

i

l

i

a

v

t

a

a

c

t

i

t

b

t

t

a

D

p

c

s

o

q

r

p

a

g

t

e

u

d

t

m

n

t

s

f

i

1

8

f

T

endezvous is considered a fundamental problem in generalised CR

etworks at the basis of many communication processes eg in-

luding neighbour discovery routing and broadcast Rendezvous pro-

ocols attempt to establish a new link for communication on a se-

ected and agreed frequency band (channel) identified within a set

f available resources (channels) thus creating the basis for com-

unication between two or more SUs This problem is often exac-

rbated by the high (and unknown) numbers of contending SUs high

umber of available channels and their ambiguous boundaries and

dentifiers and tough co-existence requirements satisfaction in cog-

itive radio systems [161] Preliminary approaches for rendezvous

ere based on the existence of a dedicated Common Control Chan-

el (CCC) agreed among all the SUs or the existence of a centralised

gent (decision maker) assigning channels to any pairs of requesting

sers This approach could be considered in the case of IEEE 80211af

olutions based on the spectrum DB implementation however the

roblem is complicated by the fact that the common channel must

e identified and made available for all the requesting SUs at the

ame time even when these are spread over multiple collision do-

ains in space On the other hand the scalability and reliability is-

ues which are caused by the congestion of CCCs and the vulner-

bility risks of centralised DBs motivate the research of alternative

istributed solutions called blind rendezvous protocols [162] Dis-

ributed solutions have been mainly based on beaconing andor slot-

ased channel hopping algorithms exploiting randomisation sensing

nd discoverysynchronisation messages spread in slotted time over

set of candidate available channels [161] These protocols aim to

inimise the convergence time to maximise the use of resources

nd to avoid collisions among SUs by identifying and agreeing on

he use of a given common channel among N non-overlapping and

nambiguously labelled channels The main complication in IEEE

0211af systems is given by the wide availability of many different

on-contiguous spectrum bands which cannot be uniformly quan-

ised in single channels and unambiguously labelled in space in or-

er to realise a common reference domain for all the IEEE 80211af

evices In fact the knowledge of the number of channels their com-

on labels the unique IDs of SUs and the number of SUs contending

or the channel assignment are the key factors determining the effec-

iveness and good properties of blind rendezvous protocols Recently

ew classes of rendezvous protocols have been proposed which are

istributed blind and oblivious Oblivious means that available chan-

els may be identified with different labels on behalf of the differ-

nt involved SUs A number of these distributed blind and oblivious

endezvous protocols have been proposed and have been referenced

nd analysed in [162] To conclude solutions for dynamic spectrum

andoff issues (eg in case PUs suddenly appear in a region) must

e resolved in advance to minimise the latency of re-establishing a

unctional communication channel eg by implementing a proactive

lternative rendezvous channel selection in background to ongoing

ommunication processes [163]

Another challenge is to provide cooperative techniques in a dis-

ributed vs centralised implementation by analysing their convergence

nder variable conditions overheads and tradeoffs In some cases as

n vehicular networks the exploitation of scenario characteristics and

actors such as the constrained mobility could help to realise effec-

ive dissemination of sensing information over an extended and pre-

ictable horizon under a cooperative approach [164ndash166] Coopera-

ion could be helpful in the realisation of the cooperative spectrum

ensing in combination with the Geolocation DB approach Another

nteresting direction of research is the investigation of mutual ef-

ects of coexisting cooperation-based vs non-cooperative techniques

n cooperation-based devices the exchange of information could al-

ow a quick convergence to a solution and avoidance of further

nterference [145] Certainly cooperative techniques have desirable

dvantages over non-cooperative ones However tight time synchro-

isation and high coordination overheads are required The common

hannel for cooperation purposes could become a bottleneck and re-

uce the potential advantages of dynamic spectrum allocation The

EEE 80219 foresees the presence of a shared common control chan-

el (CCC) to which devices need to tune in order to gather the state

f the network Solutions such as [141] propose the use of TVWS for

he CCC

A promising direction for research is the adoption of cluster-

ng schemes for secondary devices associated with a hybrid dis-

ributedcooperative vs centralisednon-cooperative approach in an

ffort to reduce overheads while maximising advantages In this

ay secondary nodes could implement cooperation for sensing func-

ions using properly identified common channels and delegate the

entralised decisions and spectrum-DB management to well instru-

ented cluster-leader nodes This concept is incarnated to some ex-

ent by RLSS in IEEE 80211af More recently there have been propos-

ls to build distributed databases that refresh their contents period-

cally by querying the remote spectrum database [167] Here Mas-

er devices (according to Ofcom terminology) periodically cache the

uery replies from the spectrum database in order to reply to and

roadcast the spectrum availability to neighbour Slave devices

Another challenging direction for research is the definition of ac-

urate models and efficient simulation tools enabling dynamic spectrum

nalysis for both rural and urban areas Recent developments of dig-

tal maps simulation tools and propagation models theoretically al-

ow improving the capability of predicting radio propagation effects

n complex scenarios with an acceptable computation time This en-

bling technology could be used in parallel to spectrum DBs to pro-

ide more accurate forecasting of dynamic frequency allocation in

imespace scenarios Another issue to be taken into account is the

ccuracy of the remote spectrum database regarding the channel

vailability estimation Several works have already shown the inac-

uracy of propagation models due to the complexity of parameters

o be considered and differences between modelled and real scenar-

os [132134168169] Better propagation models will make it possible

o more efficiently estimate the interference between devices and

uild Radio Environment Maps (REM) or White Space Maps (WSM)

hat will possibly lead to a more accurate sharing of the radio spec-

rum [144] To this end current research could focus on how to lever-

ge simulation models in order to build more accurate frequency

Bs

In general the adoption of remote spectrum DBs methodology

ushed by national regulators has made the spectrum sensing pro-

ess optional and conservatively realised (eg power control for

ensing-only devices is limited to below 17 dBm EIRP) Thus much

f the research has focused on techniques and solutions to successfully

uery the remote database also in challenging environments such as

ural areas and indoor However since the queries must provide the

osition of the mobile device the accuracy of positioning for mobile

nd handheld devices still needs more appropriate solutions Trian-

ulation can be used to estimate the position of mobile devices with

he accuracy needed by the current regulations outdoor (50 m) How-

ver a low GPS accuracy is possible indoors [170] Certified man-

al positioning can be provided by technicians for static and indoor

evices although limiting mobility Alternatives to costly infrastruc-

ures for positioning (eg based on short range beaconing devices)

ust be deployed in indoor scenarios

Finally regarding spectrum sensing and interference mitigation

ew challenges come from the heterogeneity and different charac-

eristics that networks operating in the TV White Spaces can have

uch as different transmitting power (4 W for static devices 100 mW

or mobile and portable devices and 40 mW for communication

n channels adjacent to occupied ones) different bandwidths (5

0 15 and 20 MHz are currently foreseen for IEEE 80211af IEEE

02154m can have narrower bandwidths or wider ones) and dif-

erent medium access schemes (protocols that use either CSMA or

DMA are required to co-exist in the TV White Spaces) With all this

18 B Bellalta et al Computer Communications 75 (2016) 1ndash25

d

a

T

i

a

c

t

l

b

o

t

a

d

p

a

r

p

n

a

m

t

p

p

d

d

o

t

i

D

w

p

i

a

a

d

b

fi

d

o

t

l

b

t

b

a

i

e

o

o

c

t

t

w

A

c

M

s

c

l

O

w

g

v

t

in place the problem of determining whether another secondary de-

vice is currently transmitting in the same channel is still an open

issue

Spectrum sensing is a very hot topic in the literature with many

proposals that leverage the cooperation between devices to raise the

accuracy of the detection process eg [144171] The current litera-

ture also offers room for proposals that suggest exploiting sensing to

build more accurate spectrum DBs [164172173] Lastly regarding the

interference mitigation to the primary network regulators provide a

conservative approach based on a remote spectrum database to avoid

high risk for the primary users However in particular for the indoor

scenario the concept of secondary usersrsquo transmission in TV Grey

Spaces has been proposed TV Grey Space identifies busy channels

that are formally used by primary users in the area of interest (eg

at the rooftop level of a building for DVB-T) but which could be con-

sidered usable indoors (eg at the basement level) without causing

effective interference to the primary users [134150174] This concept

is based on the generalised spectrum underlay paradigm [175] The

remaining research challenge is to provide TV Grey space access guar-

anteeing a sufficient amount of protection to the primary user while

providing a satisfactory QoS to opportunistic devices eg through

the use of customised and validated propagation models for indoor

TVWS networks and feedback mechanisms to constantly monitor the

interference generated to the primary network

6 Emerging new trends and technologies

In this last section we review three emerging trends that in our

opinion will have a large influence in the conception of future WLANs

as they change the way WLAN protocols and functionalities are de-

veloped implemented tested and integrated with other wireless

networks

61 Programmable Wireless LANs

Especially in the enterprise environment WLAN deployments

need to support a wide range of functionalities and services This is

intrinsically difficult because of the large number of APs that must be

managed which calls for scalable solutions Typical services include

channel assignment load balancing among APs authentication au-

thorisation and accounting (AAA) policy management support for

client mobility and interference coordination Another problem is the

fact that WLAN clients autonomously take several decisions such as

which APs to associate with when to hand-over etc Therefore sup-

porting roaming clients requires the management of a large number

of association states across several APs which is a challenge if sup-

port for real-time hand-over is desired Typically such management

schemes are centralised and most of them are proprietary such as

WLAN controller solutions from Aruba [176] or Cisco [177] although

the IEEE 80211u amendment has been released to allow mobile users

to seamless roam between WiFi networks with automatic authenti-

cation and handoff [178] For example Dyson [179] enables STAs to

send information such as radio channel conditions to a centralised

controller based on a custom API (eg based on Python) As the con-

troller has a centralised view of the network it can enforce a rich

set of policies to control the network also using historical informa-

tion A demo system has been implemented along with applications

such as airtime reservations for specific clients or optimised hand-

offs However Dyson requires STAs to be modified in order to use

those new services offered by the centralised controller TRANTOR

[180] is another example of a centralised management system that

requires changes of clients in order for the infrastructure to gather

information from them in terms of eg interference measurements

In addition control commands enable the infrastructure to exercise

control such as modifying the transmit power or influencing the as-

sociation procedure Clients still use standard CSMA MAC layer for

ata transmission In contrast to DenseAP [181] or DIRAC [182] which

re also based on a centralised controller but do not modify clients

RANTOR exercises much more control on clients by the use of a ded-

cated API which enables a significantly higher gain in coordination

nd thus capacity CENTAUR [183] is another example of a centralised

ontroller that centrally schedules hidden and exposed terminals in

he downlink while using standard CSMA MAC for uplink traffic and

egacy downlink traffic To achieve good performance it uses a fixed

ack-off packet staggering techniques and a hybrid data path that

nly schedules downlink transmissions towards hidden and exposed

erminals centrally All other traffic is sent using standard DCF with

standard back-off procedure In contrast to previous approaches it

oes not require any modifications to the clients but does require data

lane centralisation and it remains questionable how scalable such

solution is for large WLAN deployments with high PHY layer data

ates beyond 100 Mbps However new approaches for fast data plane

rocessing that are currently explored in the virtualisation commu-

ity which move all the packet processing into user space may be

n option to significantly increase speed Designing such manage-

ent systems raises several interesting research questions (a) Cen-

ralised infrastructures may incur high latency but are more simple to

rogram and maintain while designing a completely distributed ap-

roach that operates close to WLAN devices may result in significant

istributed coordination and consistency problems (b) what level of

istributed control is required in order to support the flexibility needs

f future WiFi based networks supporting a high level of mobility

Traditionally WLAN APs are built on proprietary operating sys-

ems that are tightly coupled with the hardware This design makes

t hard to create new applications on top of such networking devices

espite the fact that protocols and mechanisms are available that

ould greatly improve the utility of existing networks those new

rotocols are not deployed because the closed system design makes

t very difficult to extend their functionality An important aspect for

ll such new approaches is therefore how to provide open interfaces

nd open source to speed up innovation As an example Linux-based

evices are completely open source but in order to increase flexi-

ility much more work is needed in the area of open drivers and

rmware In fact the Atheros based ath9k driver has been the main

river of innovation in the open source community because it is the

nly driver that interworks with open firmware In general solutions

hat modify the OS driver focus mainly on programmable network

evel solutions for WLANs such as channel switching or handovers

etween APs While the ath9k driver is an excellent example in how

o speed up innovation it is still very cumbersome to support flexi-

le MAC engine reprogrammability with ath9k In contrast MAClets

nd Wireless MAC processors (WMP) [184] allow a much more flex-

ble reprogramming of MAC functionalities A MAC processor is an

ntity able to execute general MAC commands that specify the MAC

perations through a software-defined state machine The behaviour

f the MAC protocol can therefore be updated at run-time by simply

hanging the sequence in which those commands are executed (ie

he MAClets) As a proof-of-concept the authors of [184] implement

he proposed MAC processor solution in a commodity WLAN hard-

are card extending the basic DCF in three directions piggy-backing

CKs a pseudo TDMA and the use of multiple channels In [185] a

ontrol framework for this WMP system is also proposed to support

AClet code mobility ie for moving loading and activating MAC

oftware programs embedded into ordinary data packets (akin to the

apsule model of traditional active networks)

The difficulty in re-programming networking hardware has also

ed to the concept of Software Defined Networks (SDNs) based on the

penFlow protocol [186] The main idea of SDNs is to extend net-

orking devices with standardised APIs that allow third-party pro-

rammers to flexibly control the data path In addition SDNs pro-

ide higher level abstractions to network designers and programmers

hrough the use of a centralised control plane offered by a network

B Bellalta et al Computer Communications 75 (2016) 1ndash25 19

Fig 11 The CLOUDMAC architecture VAP stands for Virtual Access Point and WT

stands for Wireless Termination

c

a

c

t

m

W

n

t

t

c

p

t

u

h

a

i

e

s

o

c

a

a

f

T

r

s

a

s

p

o

M

C

i

T

t

fl

A

a

a

i

c

t

s

t

e

I

t

w

H

e

b

I

l

d

o

6

r

W

s

l

a

e

e

a

s

b

c

t

(

S

p

c

F

B

t

p

T

e

t

s

u

m

i

e

o

a

A

p

d

t

t

n

p

i

t

t

f

i

ontroller such as NOX [187] which allows reuse of components such

s topology discovery or network access control for different appli-

ations The SDN concept has recently been applied to WLAN archi-

ecture in order to enable fine grained control over mobility manage-

ent and data forwarding focusing on a programmable enterprise

LAN architecture An important part of the SDN architecture is the

etwork controller which provides a centralised view of (parts of)

he SDN enabled network and uses the OpenFlow protocol to install

he forwarding rules on SDN-enabled devices (routers switches ac-

ess points cellular base stations etc)

ODIN [188] is designed to support programmability in enter-

rise WLAN architecture by separating the association state from

he physical AP They implement Light Virtual Access Points (LVAP)

sing a Split-MAC approach where the infrastructure controls the

andover procedure among different WLAN APs By managing

ssociations through SDN controllers ODIN enables proactive mobil-

ty management and load balancing within the SDN enabled WLAN

nterprise network without the need for changes in the client WLAN

tack or IEEE 80211 MAC layer While ODIN requires agents to reside

n the APs that communicate with the Odin Master within the SDN

ontroller CLOUDMAC [189] is based purely on the concept of SDN

nd virtual APs Similar to ODIN CLOUDMAC implements a Split-MAC

pproach but in addition enables the processing of WLAN MAC layer

rames within a co-located Cloud using so-called Virtual APs (VAPs)

he physical APs in CLOUDMAC are lightweight WLAN APs that are

esponsible only for sending out IEEE 80211 based MAC layer ACKs to

tandard WLAN clients and tunnel WLAN MAC layer frames through

n SDN-enabled enterprise WLAN towards the VAPs As association

tates are kept in the VAPs fast mobility is supported using sim-

le OpenFlow forwarding rules Because of the additional processing

f IEEE 80211 WLAN MAC frames in the co-located Cloud CLOUD-

AC has higher latency than a standard WLAN deployment However

LOUDMAC offers a Webservice based API to third party applications

n order to program the enterprise WLAN and enable new services

he architecture of CLOUDMAC is depicted in Fig 11 It has been ex-

ended in [190] to support QoS management and in [191] to support

exible MAC management frame prioritisation based on IEEE 80211

system based on OpenFlow has been recently proposed in [192] to

llow the station to be associated with multiple APs simultaneously

nd to switch between APs with low overhead

An important aspect to consider in order to enable programmabil-

ty is backwards compatibility Several approaches (eg [180]) require

lients to be modified to utilise the features provided This is difficult

o do in practice because it requires changes in the operating system

oftware of all clients If not all clients can utilise those APIs it is ques-

ionable how usable the new architecture will be and how much ben-

fit in terms of aggregate performance such architecture will allow

n contrast the SDN based approaches are interesting in the sense

hat they do not require changes in the client WLAN stack and work

ith the standard IEEE 80211 MAC layer deployed within the clients

owever it remains questionable how scalable such solutions are For

xample an interesting open research topic is to evaluate the scala-

ility of approaches such as CLOUDMAC [189] In addition processing

EEE 80211 MAC frames within co-located private Cloud requires low

atency support from local Cloud solutions (such as OpenStack) in or-

er to reduce the IEEE 80211 MAC processing time which is an area

f active research

2 Prototyping and testing IEEE 80211 enhancements

Most of the new proposals for next-generation WLANs are cur-

ently only evaluated using mathematical analysis and simulation

hile both analysis and simulation are necessary to characterise and

tudy those enhancements in the initial design phase or to consider

arge-scale scenarios it is difficult to consider all practical aspects of

real-world scenario This can sometimes cause significant differ-

nces between what simulations and real experiments show How-

ver real experiments are challenging because of the high complexity

nd costs of building the new hardware and software for each specific

olution to test

To mitigate implementation complexity there are several flexi-

le hardware platforms where low-level MAC mechanisms can be

ompletely implemented in software Examples based on FPGAs are

he USRP (Universal Software Radio Peripheral) [193] and the WARP

Wireless Open-Access Research Platform) [194] As an alternative

ORA (Microsoft Research Software Radio) [195] works on general

urpose computers by taking some advantage from modern multiple

ore systems

A different approach is provided by OpenFWWF [196] Open-

WWF provides an open CSMACA firmware for specific models of

roadcom chipsets so the resulting firmware can be uploaded and

ested in real commercial hardware OpenFWWF implements a sim-

le State Machine (SM) for controlling the hardware in real time

he SM evolution is driven by a main loop that reacts to events by

xecuting specific handlers When a packet originally prepared by

he Linux kernel is ready in the NIC memory handler Packet_Ready

ets up the radio hardware according to the packet metadata sched-

les the transmission and jumps back to the main loop The Trans-

ission Engine (TXE) then takes care of accessing the channel ie

t decrements the back-off counter according to the DCF rules and

ventually starts the actual transmission This triggers the execution

f the TX_frame_now event that prepares the ACK time-out clock

nd finalises the MAC header If the ACK-frame is received or if the

CK time-out expires and the maximum number of attempts for this

acket is reached handler Update_Params resets the contention win-

ow to its minimum value or otherwise doubles it Finally it loads

he back-off counter with a fresh value

In the following we will review a list of selected papers that use

he OpenFWWF firmware and the WARP or USRP platforms to test

ew proposals for WLANs in real scenarios as examples A first im-

lementation of the groupcast mechanisms defined in IEEE 80211aa

s presented in [197] The authors modify the OpenFWWF firmware

o include those functionalities and evaluate them using a 30 STAs

estbed A second example of the use of OpenFWWF is the collision-

ree MAC protocol presented in [198] In [198] the authors design and

mplement in OpenFWWF a MAC protocol that is able to achieve a

20 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 12 LTE to WIFI offloading

a

n

c

t

[

d

a

W

v

t

t

o

b

f

S

g

l

i

I

s

I

b

n

s

m

a

p

d

c

N

o

v

t

t

n

l

i

a

t

a

collision-free operation by waiting a deterministic time after success-

ful transmissions before to start a new transmission This work shows

the experimental performance of a collision-free MAC (CF-MAC) pro-

tocol for WLANs using commercial hardware The results show that

the proposed CF-MAC protocol leads to a better distribution of the

available bandwidth among users a higher throughput and lower

losses than DCF

In [199] the authors use the WARP platform to test a variant of the

DCF ndash called IEEE 80211ec ndash that substitutes control packets such as

the RTS CTS and ACK with short detectable sequences Since those se-

quences are shorter than the control packets and can be detected cor-

rectly even at lower SNRs values a significant gain in performance is

achieved MIDAS (Multiple-Input Distributed Antenna Systems) [200]

implementation using the WARP platform shows the benefits of dis-

tributing the antennas over the area to cover instead of co-locating

them at the AP in terms of capacity when MU-MIMO is employed

The authors also propose a new MAC protocol to benefit from the

spatial reuse that the DAS allows taking as a basis the IEEE 80211ac

Experimental results show that their proposal is able to achieve up

to 200 gains versus the traditional approach where all antennas are

co-located at the AP

The USRP platform has recently been proposed to develop a first

implementation of IEEE 80211ah in order to obtain a preliminary ex-

perimental performance assessment of such a technology because

no commercial off-the-shelf hardware is yet available as the amend-

ment is still in progress [201] The USRP platform is also used in

[202] to evaluate TIMO (Technology Independent Multi-Output) a so-

lution to deal with high-power non-IEEE 80211 interferers in ISM

bands

To conclude this section it is worth mentioning here also to

MAClets and WMP as given such flexibility to implement and dis-

tribute new MAC protocols and other IEEE 80211 functionalities we

believe that if a use-friendly implementation of such a MAC proces-

sors framework is provided it would significantly contribute to the

development and testing of new MAC enhancements by the research

community

63 CellularWLAN interworking

Public hotspots that offer Internet access over a WLAN using IEEE

80211 technology are now nearly ubiquitous It is forecasted that the

cumulative installed base of WiFi hotspots worldwide will amount

to 551 million by 2018 excluding private hotspots (eg WiFi access

points deployed at home) [203] The sharp increase in the availabil-

ity of public WiFi was initially perceived by mobile cellular operators

as a threat due to the additional competition from wireline Internet

service providers or emerging crowdsourced WiFi networks such as

FON5 However as cellular operators are fighting to cope with the ex-

plosion of mobile data traffic created by the rising use of multimedia

content traffic over mobile devices [15] they are also starting to use

WLANs based on the IEEE 80211 technology to offload data from their

core and access networks In general mobile data offloading refers to

the use of complementary network technologies (in licensed or un-

licensed spectrum) for delivering data originally targeted to cellular

networks Intuitively the simplest type of offloading consists of ex-

ploiting connectivity to existing co-located WiFi networks and trans-

ferring data without any delay (Fig 12) Thus this offloading tech-

nique is known as on-the-spot offloading As a consequence of this

new trend the seamless integration of cellular (eg 3GLTE) and WiFi

technologies has attracted significant research interest in recent years

(see [204] for a survey) a few solutions have already been standard-

ised [205206] and roaming between cellular and WiFi is becoming

increasingly transparent to end users CellularWLAN interworking is

5 httpscorpfoncomen

fl

t

f

lso fostered by the support in the evolving 4G standards of heteroge-

eous network deployments (HetNets) in which the existing macro

ells are complemented with a number of small low-power base sta-

ions with the goal of increasing capacity in highly congested areas

207208] It is envisaged that small cells will be based on 4G stan-

ards (eg pico and femto cells) as well as IEEE 80211 technologies

nd multimode base stations that work simultaneously with LTE and

iFi are already entering the market

It is important to point out that LTE standards already support a

ariety of mechanisms that enable data offloading However most of

he existing solutions such as Local IP Access (LIPA) and Selected In-

ernet IP Traffic Offload (SIPTO) focus on data offloading at the core

f the network [205] For instance LIPA allows an IP-enabled mo-

ile device to transfer data to another device in the same pico or

emto cell without passing through the cellular access network while

IPTO enables the routing of selected IP data flows through different

ateways The only LTE offloading mechanism that supports seam-

ess interworking with IEEE 80211-based WLANs is IP Flow Mobil-

ty and Seamless Offload (IFOM) Specifically IFOM relies on Mobile

Pv6 technologies to allow a user terminal to simultaneously route

elected IP flows over different radio access technologies [209210]

n this way a user terminal can offload selected flows to a WLAN

ased on some operator-defined policy while keeping the LTE con-

ection running However to enable such an approach it is neces-

ary to support an entity in the cellular core network that can com-

unicate with the user terminal to exchange information about the

vailability and quality of neighbouring access networks as well as to

rovide the user terminal with predefined rules to manage the han-

over process This entity in current LTE standards is called the Ac-

ess Network Discovery and Selection Function (ANDSF) server [211]

ote that the problem of selecting the best communication technol-

gy in an heterogeneous wireless network has been extensively in-

estigated in the literature both with centralised as well as decen-

ralised approaches (see [212] for a survey) However the use of mul-

iple interfaces in parallel as well as per-flow offloading are relatively

ew concepts Thus the design of scalable and efficient network se-

ection strategies for the ANDSF framework is still an open issue For

nstance three offloading methods suitable for the ANDSF framework

re proposed in [213] which are based on coverage SNR and sys-

em load A reinforcement learning approach is designed in [214] that

llows multimode base stations to autonomously steer their traffic

ows across different access technologies depending on the traffic

ype the usersrsquo QoS requirements the network load and the inter-

erence levels A number of studies have also tried to quantify the

B Bellalta et al Computer Communications 75 (2016) 1ndash25 11

8

c

c

a

s

b

w

o

j

t

f

a

Q

t

t

A

p

t

s

a

f

s

H

a

i

c

i

t

s

f

s

s

m

i

t

s

l

t

t

[

t

i

o

b

r

W

v

m

8

s

i

o

p

q

4

c

e

t

d

a

t

m

Fig 7 WLANs for M2M communications STAs represent sensor and actuator devices

4

t

w

s

a

c

g

e

(

p

o

t

p

9

r

o

a

p

c

o

i

p

b

f

t

s

o

i

4

8

f

t

i

s

m

d

0211e-based WLANs owing to its higher complexity and cost con-

erns Instead EDCA is widely adopted Most papers have thus fo-

used on improving EDCA performance Many papers have proposed

nalytical models for various subsets of EDCA functionalities For in-

tance a saturation-based performance analysis is conducted in [94]

y differentiating the minimum back-off window size the back-off

indow-increasing factor and the retransmission limit The authors

f [9596] also model AIFS differentiation while the model in [97]

ointly captures all the four EDCA parameters for traffic differen-

iation More recent papers have analysed the EDCA performance

or non-saturated conditions and for arbitrary buffer sizes [98] The

uthors in [99] have developed an analytical model to predict the

oS levels that can be achieved once a new voicevideo flow is in-

roduced in the WLAN A Kalman filter is proposed in [100] to ob-

ain estimates on the number of active transmission queues of each

ccess Category in EDCA These analytical models can then be ex-

loited to derive the optimal configuration of the EDCA parameters

o achieve given performance criteria or to design admission control

chemes that preserve QoS constraints For instance a scheme that

ssigns contention-window values to achieve pre-defined weighted-

airness goals is proposed in [101] A control-theoretic scheme is de-

igned in [102] with the goal of minimising the video traffic delay

owever most of these solutions rely on non-realistic assumptions

bout video traffic dynamics An alternative class of solutions dynam-

cally updates the EDCA parameters based on the observed network

onditions In [103] the EDCA parameters are optimised consider-

ng a WLAN with rigid and elastic traffic simultaneously analysing

he interactions between both types of traffic The authors in [104]

pecify several bandwidth-sharing mechanisms with guaranteed QoS

or voice and video traffic Measurement-based admission control

chemes are proposed in [105] A TXOP adaptation method is de-

cribed in [106] that takes into account video frame sizes and trans-

it queue lengths However the main drawback of these solutions

s that they are based on heuristics and hence do not ensure op-

imal and guaranteed performance Finally a third category of re-

earch papers tries to improve video performance by designing cross-

ayer scheduling approaches Specifically these works take advan-

age of multi-layer video encoding to classify the frames according

o their importance and assign them to different access categories

107] For instance the authors in [108] define classifiers and waiting

ime priority schedulers that dynamically change the packet prior-

ties according to end-to-end delay measurements A disadvantage

f this approach however is that an additional adaptation layer may

e needed to implement the complex interactions that are typically

equired between the video coding applications and the MAC layer

e conclude this section by pointing out that existing studies pro-

ide the basic design principles and techniques for improving multi-

edia streaming performance in IEEE 80211 networks Still the IEEE

0211aa standard poses new research challenges that have not been

ufficiently explored and that will require innovative solutions For

nstance scheduling between primary and alternate queues is still an

pen research area as the mapping of individual frames to multi-

le queues in order to achieve graceful degradation of voicevideo

uality [16]

Sensor Networks and machine-type communications

As discussed in Section 2 M2M communications refer to any

ommunication technology that enables sensoractuator devices to

xchange information and perform actions without the manual assis-

ance of humans This section reviews the main features currently un-

er consideration in the development of the upcoming IEEE 80211ah

mendment which targets the main challenges of those networks as

he IoT in general such as the energy consumption or the manage-

ent of many devices

1 The IEEE 80211ah amendment

The IEEE 80211ah amendment [109] aims to provide WLANs with

he ability to both manage a large number of heterogeneous STAs

ithin a single BSS and minimise the energy consumption of the

ensor-type battery-powered STAs

The initial design requirements of the IEEE 80211ah amendment

re detailed in [110] these entail the support of up to 8192 STAs asso-

iated with a single AP the adoption of efficient power saving strate-

ies a minimum data rate of 100 kbps the operation in the license-

xempt sub 1 GHz band and a coverage up to 1 km in outdoor areas

see Fig 7 for an illustrative example) A preliminary assessment of

erformance of the IEEE 80211ah technology in terms of the number

f STAs that can be effectively supported in a single WLAN as well as

heir energy consumption is presented in [111]

IEEE 80211ah operates over different sub-1 GHz ISM bands de-

ending on country regulations 863ndash868 MHz in Europe 902ndash

28 MHz in the US and 9165ndash9275 MHz in Japan China South Ko-

ea and Singapore also have specific channelisations Channel widths

f 1 MHz and 2 MHz have been adopted although 4 8 and 16 MHz

re also supported in some countries IEEE 80211ah furthermore pro-

oses new PHY and MAC layers The IEEE 80211ah PHY layer can be

onsidered to some extent a sub-1 GHz version of the IEEE 80211ac

ne At the physical layer OFDM is the chosen modulation method us-

ng 32 or 64 tonessub-carriers that are spaced by 3125 kHz The sup-

orted modulations include BPSK QPSK and from 16 to 256-QAM A

road range of antenna technologies ranging from single-user beam-

orming to MIMO and DL-MU-MIMO which was first introduced in

he IEEE 80211ac amendment are also included in the IEEE 80211ah

pecification Similarly the IEEE 80211ah MAC protocol include most

f IEEE 80211 main characteristics further extending its power sav-

ng (PS) mechanisms

11 Novel features

This section overviews the extensions introduced by the IEEE

0211ah amendment to the IEEE 80211 PS mechanisms to account

or the specific characteristics of resource-constrained sensor and ac-

uator devices A more detailed review can be found in [110] includ-

ng a performance assessment of IEEE 80211ah in several of the key

cenarios for M2M communications such as agriculture and animal

onitoring smart metering and industrial automation plants In ad-

ition a detailed survey of the IEEE 80211ah is reported in [112]

12 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 8 IEEE 80211ah PS mode for TIM-STAs non-TIM STAs and Unscheduled STAs

a

d

w

i

a

a

c

m

i

a

e

t

s

m

t

t

u

b

i

I

s

d

f

a

b

s

a

b

S

T

c

t

w

I

4

a

a

which addresses aspects not considered in [110] such as the use of

sectorisation to avoid overlapping between multiple IEEE 80211ah

WLANs and the use of relays to further extend the IEEE 80211ah cov-

erage among other aspects Finally in [113] both IEEE 80211ah PHY

and MAC characteristics are introduced with emphasis on the ben-

efits of using the sub-1 GHz ISM band in terms of link budget the

location of the OFDM pilots for outdoors operation the use of bidi-

rectional transmission opportunities and the new packet formats to

minimise overheads among others

Enhanced power saving mechanism PS mechanisms for WLANs

were already considered in the development of the first IEEE 80211

standard with the goal of improving the lifetime of battery equipped

devices [114] In PS mode STAs keep the transceiver in sleeping mode

as much time as possible They periodically wake up to listen to the

beacons transmitted by the AP Those beacons indicate whether an

STA has packets waiting for it at the AP In the positive case that STA

remains awake and requests the delivery of those packets Otherwise

given it has nothing to receive it returns to sleep mode until the next

beacon is expected

In the IEEE 80211ah amendment time is divided into pages DTIM

(Delivery Traffic Indication Map) periods TIM (Traffic Indication Map)

periods and slots DTIM and TIM periods begin with the correspond-

ing DTIM and TIM beacons sent by the AP The functions of DTIM and

TIM beacons are described below

1 DTIM beacons They inform as to which TIM Groups (ie the

group of STAs assigned to the same TIM period) have pending

packets at the AP

2 TIM beacons Each TIM message informs a TIM Group about

which specific STA has pending data in the AP Between two

consecutive DTIMs there are as many TIM beacons as TIM

Groups

Using this DTIMTIM-based approach any STA can enter into a

power saving state if it does not have packets pending for transmis-

sion and one of two conditions is met (1) it observes in the DTIM

beacon that there is no downlink traffic addressed to its TIM Group or

(2) it observes in the DTIM beacon that there is some downlink traffic

addressed to its TIM Group but that STA does not explicitly appear in

the TIM beacon Compared to the preliminary IEEE 80211 PS mecha-

nism this approach reduces the size of the Traffic Indication Map in

each TIM beacon thus reducing the overhead and the time STAs need

to listen and process them In addition TIM periods can be organised

in pages which further increases the number of TIM groups between

two DTIM beacons

The temporal organisation of pages DTIM and TIM periods is re-

ported in Fig 8 which shows the division of TIM periods in RAW (Re-

stricted Access Window) and PRAW (Periodic RAW) The RAW is a

time interval in each TIM period where TIM stations can transmit and

receive data (see below the different types of STAs defined in IEEE

80211ah) In addition it can be divided into several downlink and

uplink slots for further granularity In the downlink the slots are as-

signed to a single STA or a group of STAs while the uplink slots are

randomly selected by the STAs with packets ready for transmission

The PRAW is the period of time in each TIM where non-TIM stations

can transmit and receive data

Types of STAs IEEE 80211ah supports three types of STAs TIM

non-TIM and Unscheduled STAs A TIM station is assigned to a TIM

Group Their data transmissions must be performed within a RAW

(Restricted Access Window) period Non-TIM stations do not have to

listen to beacons to transmit data During the association process

non-TIM devices directly negotiate with the AP to obtain a trans-

mission time allocated in a PRAW The following channel access can

either be renegotiated or occurs periodically depending on the re-

quirements set by the STA Unscheduled STAs do not need to listen to

any beacons similar to non-TIM stations Even inside any restricted

access window unscheduled STAs can send a poll frame to the AP

sking for immediate access to the channel The response frame in-

icates an interval (outside both restricted access windows) during

hich unscheduled stations can access the channel This procedure

s meant for STAs that transmit data very sporadically

Hierarchical station organisation To support a large number of STAs

nd their organisation in pages DTIM and TIM periods IEEE 80211ah

ssigns to each associated STA a unique identifier of 13 bits which is

alled the Association Identifier (AID) Using this new AID the maxi-

um number of supported STAs is increased from the original 2007

n IEEE 80211 to 8191 (= 213 minus 1) in IEEE 80211ah However it also

llows categorising STAs according to the type of application they are

xecuting their power level or even their desired QoS by assigning

hem to different TIM groups

Long sleeping periods IEEE 80211ah offers TIM Non-TIM and Un-

cheduled STAs the possibility to set very long doze times (up to

onths) The corresponding clock drift produced by such long doze

imes must be taken into consideration however as the higher the

ime an STA has been asleep the further in advance it should wake

p to avoid possible synchronisation problems with the network

Efficient small data transmission Three new enhancements have

een proposed to reduce the overhead when the data packet size

s small First while IEEE 80211 contains a 28-byte MAC header

EEE 80211ah proposes a short 18-byte version by using AIDs in-

tead of MAC addresses Second IEEE 80211 has defined several null

ata packet (NDP) frames which consist only of a PHY header These

rames can be used to create short ACKs short Block ACKs short CTSs

nd short PS-Polls Finally a Fast Frame Exchange mechanism has

een developed so that if an STA has data to transmit it can notify a

uccessful reception by transmitting its data frame instead of an ACK

Sectorisation Since the PHY layer is based on the IEEE 80211ac

mendment single and multi-user beam-forming are also supported

y IEEE 80211ah This allows the transmission of data to multiple

TAs simultaneously in the downlink increasing the system capacity

he use of the beam-forming capability of IEEE 80211ah APs is also

onsidered as a means to group STAs into different independent an-

enna sectors with the main goal of reducing interference issues This

ould be particularly useful in the case of overlapping with other

EEE 80211ah WLANs or in the presence of hidden nodes

12 Open challenges

A first open challenge is to understand the coverage and achiev-

ble transmission rates in IEEE 80211ah WLANs in both indoor

nd outdoor scenarios Propagation models for WLANs working at

B Bellalta et al Computer Communications 75 (2016) 1ndash25 13

f

c

8

p

R

p

i

m

l

a

m

c

p

p

f

m

t

a

e

p

t

m

a

l

T

m

a

i

n

t

t

t

u

p

c

w

f

a

t

8

a

m

a

s

l

a

m

b

m

d

S

i

a

o

a

R

t

n

G

fi

s

t

p

t

c

u

c

w

i

o

p

g

a

c

i

f

c

(

d

m

l

i

R

t

l

t

L

m

a

m

o

e

f

a

I

i

t

w

i

t

a

c

w

p

f

t

d

5

i

s

e

s

w

5

a

c

a

n

requencies lower than 1 GHz are evaluated in [115] The authors

ompare two path loss propagation models proposed by the IEEE

0211ah Task Group (one for macro and one for picohotzone de-

loyments) [116] with Lee and HatandashOkumura propagation models

esults show that the IEEE 80211ah channel models underestimate

ath loss with respect to Lee and Hata models Moreover in a compar-

son with empirical data it is observed that the IEEE 80211ah channel

odels also underestimate the initial loss and the slope of the path-

oss function A new model parameterisation is thus proposed by the

uthors In [117] the feasibility of an IEEE 80211ah WLAN deploy-

ent is also evaluated in terms of the achievable range and bit rate

omputed on the basis of the link budget using the same path loss

ropagation models as in [115] Results show that the transmission

ower limitations in the uplink can limit the overall network per-

ormance Finally in [118] the authors evaluate the achievable trans-

ission range for the different transmission rates and the achievable

hroughput for different combination of transmission power values

nd transmission rates Further studies are required due to the het-

rogeneity of scenarios in which IEEE 80211ah WLANs can be de-

loyed to characterise the effective coverage ranges with special at-

ention to mixed indoors and outdoors scenarios and in presence of

obile nodes

Due to the wide coverage IEEE 80211ah WLANs may be severely

ffected by the presence of hidden terminals To mitigate such a prob-

em smart solutions to guarantee that the STAs assigned to the same

IM Group are inside each other transmission range or the employ-

ent of RTSCTS-based solutions may contribute to mitigate such

problem An overview of long-range scenarios for IEEE 80211ah

s presented in [119] including a detailed description of the hidden

ode problem in them

Single-hop uplink transmissions from distant STAs require high

ransmission power to reach the AP Since not all STAs may be able

o transmit at the required power a solution could be the introduc-

ion of nodes able to relay transmissions from them However the

se of relays has to be efficiently harmonised with the operation in

ower saving mode allowing temporal periods in which the relays

an gather the data from their associated STAs and then periods in

hich they forward the data to the AP which is still an open challenge

or IEEE 80211ah WLANs In case relays are used data prediction and

ggregation techniques could be implemented to make more efficient

ransmissions [120]

Another challenge is the assessment of the efficiency of IEEE

0211ah PS mechanisms including their optimisation In [111] the

uthors evaluate the capacity of the TIM and page segmentation

echanisms in terms of the maximum number of nodes supported

nd the energy consumed given a certain network traffic profile Re-

ults confirm that a large number of STAs can be supported with

ow energy consumption The impact of the number of nodes oper-

ting in PS mode on the energy consumption and the delay perfor-

ance is analysed in [121] A detailed analysis of the TIM Group-

ased channel access adopted by the IEEE 80211ah task group is

ade in [122] where the STAs assigned to each TIM are uniformly

istributed between the slots of the RAW Since only one group of

TAs is allowed to transmit in each RAW slot the channel contention

s minimised The number of downlink and uplink slots in a RAW

s well as their duration is optimised on the basis of the number

f active STAs and the network traffic profile in [123] In [124] the

uthors propose an algorithm to determine the optimal size of the

AW interval for uplink transmissions based on the estimation of

he number of STAs transmitting and the duration of the RAW Fi-

ally in [125] the authors address the contention problems in Smart

rid communication networks with many nodes and periodic traf-

c when using an IEEE 80211ah WLAN It is still a challenge to con-

ider smart systems able to adapt the IEEE 80211ah parameters to

he instantaneous system state in order to save as much energy as

ossible

Since the PS mechanisms affect only the operation of TIM STAs

he performance of non-TIM and Unscheduled STAs has not yet been

onsidered in the literature To identify the scenarios in which the

se of non-TIM and Unscheduled STAs are of interest is still an open

hallenge as further investigating how non-TIM STAs negotiate over

hich PRAWs they can use to transmit and receive data Co-existence

ssues between the three types of STAs in a single network is also an

pen challenge that requires further work to guarantee a minimum

erformance level for all of them

The design of strategies to distribute the STAs between all the TIM

roups based on their specific traffic profile position battery level

nd application priority is another important challenge that is still

ompletely open EDCA is the default channel access scheme included

n the IEEE 80211ah amendment and provides some basic traffic dif-

erentiation capabilities at the packet level Besides the different ac-

ess categories (ACs) needing to be renamed and their parameters

Arbitration Inter-Frame Spacing or AIFS CWmin TXOP duration) up-

ated to fit the specific traffic profiles of M2M communications other

echanisms can be implemented A mechanism to assign the down-

ink RAW slots when packets from multiple priority levels are wait-

ng for transmission at the AP is required for example Also uplink

AW slots are currently selected randomly by the STAs that want to

ransmit a packet A mechanism to reserve some slots at each up-

ink RAW for high priority STAs may therefore be an option although

his may severely degrade the performance of the low priority STAs

astly different repetition patterns for different priority TIM groups

ay also allow priority STAs assigned to those priority TIM groups to

ccess the channel more often We expect this challenge will receive

uch attention in the upcoming years because of the heterogeneity

f sensors that will be connected though a single IEEE 80211ah AP

specially in urban scenarios

Similar to what has been discussed for 4G cellular networks used

or M2M communications [126] the development of mechanisms to

void and resolve congestion situations is another open challenge for

EEE 80211ah WLANs with many nodes For instance a mechanism

n which STAs have to wait a random number of DTIM periods before

hey can start a transmission may be implemented This approach

ould effectively distribute the traffic over a longer period of time

n overload conditions but otherwise would unnecessarily increase

he access delay Adaptive and load-aware solutions may play again

n important role to keep the optimal WLAN operation in all possible

ircumstances

Finally since IEEE 80211ah will compete with 4G5G cellular net-

orks and WSNs to provide M2M connectivity in many different ap-

lication domains such as smart cities or e-health comparative per-

ormance studies to determine the strong and weak points of each

echnology are required including also aspects such as the cost of the

evices and the system reliability

Cognitive radio technology for TV White Spaces

This section gives an overview of the IEEE 80211af amendment

ntroducing its most relevant features and open challenges Fig 9

hows the basic components and features considered for WLANs op-

rating in the TV White Spaces (TVWS) which include local spectrum

ensing by the AP and STAs and the use of geolocation data bases

ith information on channel availability

1 The IEEE 80211af amendment

Thanks to the transition from analog TV to digital TV several VHF

nd UHF spectrum channels used for decades for analog TV broad-

asting are now unused or are under-utilised in many geographical

reas [127ndash129] This ldquodigital dividendrdquo of spectrum has suggested

ational regulators such as the FCC in the US and Ofcom in the UK

14 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 9 WLAN operating in TVWS basic elements and functionalities

Fig 10 Hidden terminal problem in TVWS when the secondary network uses only in-

stantaneous channel sensing to decide whether or not a TV channel is occupied When

the STA is not able to detect the TV signal due to the presence of obstacles it may

initiate a transmission and create interference to the primary users

s

c

5

a

i

p

t

o

o

t

e

c

u

l

m

b

a

m

t

e

t

t

t

m

D

c

I

n

b

u

8

c

a

t

i

t

w

f

to discuss how to reuse these channels for unlicensed devicesrsquo com-

munications [130131] These potentially vacant channels in the VHF

and UHF bands are referred to as TVWS [132] and include spectrum

portions like the 470ndash790 MHz in Europe and non-contiguous 54ndash72

76ndash88 MHz 174ndash216 MHz 470ndash698 MHz and 698ndash806 MHz in USA

A snapshot of the TV spectrum occupancy in the city of Barcelona in

2012 is shown in [133]

The attractive characteristics of TVWS (not only for WLANs) in-

clude the ability to penetrate through walls and other obstacles much

more effectively than other widely used spectrum bands such as the

24 and 57 GHz ISM bands [134ndash137] This fact along with a progres-

sive spectrum scarcity in ISM bands has suggested the birth of the

IEEE 80211af amendment published in February 2014 [138] which

provides the IEEE 80211 operational characteristics for TVWS access

of unlicensed White Space Devices (WSD) A good summary of IEEE

80211af can be found in [137] The main advantage of operating IEEE

80211 WLANs in the TVWS comes from an increased coverage range

which can reach up to 1 km in rural areas and open fields [139]

and less energy needed to transmit However this comes at the cost

of an increased interference risk to other WSDs which creates co-

existence problems and demands new PHY and MAC layers to ef-

ficiently support channel access and operations preserving licensed

usersrsquo devices

WLANs operated in the TVWS could cover a number of interest-

ing and emerging use cases and scenarios eg Internet access in ru-

ral or sparsely populated areas Smart Grid sensor aggregation me-

tering and control Internet of Things advanced WLAN operations

and TVWS traffic offloading in indoor environments [140141] Nev-

ertheless the concept could be realised to create or extend com-

mercial or municipality WiFi services offered to citizens with cover-

age at the whole city level realised with reasonably limited network

infrastructures [142143]

IEEE 80211af will use the PHY and MAC layers derived from IEEE

80211ac It will adopt concepts such as the OFDM multi-user beam-

forming contiguous and non-contiguous channel bonding and packet

aggregation Among the mandatory and innovative behavioural and

operational parameters the most notable one is the channel acqui-

sition support realised through remote geolocation-based spectrum

allocation databases which maintain the channelsrsquo availability infor-

mation in any given area and time of day providing upon request the

list of free channels available for use

In the following section we highlight and summarise three main

novelties that IEEE 80211af introduces In Section 511 we explain

how the access infrastructure to the remote spectrum database is de-

signed and what the requirements are we discuss the coexistence is-

sues and methodologies adopted in IEEE 80211af and we also discuss

the novel concept of non-contiguous channel bonding The following

ections also contain an illustration of related works Open research

hallenges will be summarised in Section 512

11 Novel features

This section provides an illustration of novel features introduced

nd discussed in the path to IEEE 80211af and the directions provided

n related works and standardisation initiatives to resolve classical

roblems properly characterised in the new framework of TVWS

echnologies

Channel acquisition spectrum database and channel sensing Much

f the attention in operating in the TVWS is given to the protection

f the primary licensed users in the TVWS spectrum band In general

he primary users were considered as the Digital TV (DTV) broadcast-

rs and receivers In fact receivers would suffer the wasteful effect of

ollisions during the TV broadcast reception in the case of secondary

sersrsquo (SU) transmission interference (Fig 10)

It is a well-known problem that the definition of a coherent wire-

ess channel status (idlebusy) from the viewpoint of distributed pri-

ary users (eg DTV receivers) is made difficult by the difference

etween transmitters and receivers interpretation of signals in a vari-

ble time-space collision domain resulting in hiddenexposed ter-

inals concepts Solutions proposed to reduce the hiddenexposed

erminals included the RTSCTS mechanisms and beaconing How-

ver it is difficult to implement solutions that would effectively de-

ermine the channel status of TVWS channels based on localised dis-

ributed sensing activities performed by multiple secondary users in

he same collision domain On the other hand the use of RTSCTS

echanism would not be effective in this case because most of the

TV primary users (and specifically the receivers) were not originally

onceived to transmit signals or implement the RTSCTS handshake

n conclusion the actions to identify free (and busy) TVWS chan-

els should not be based on primary usersrsquo involvement but should

e almost totally implemented by specifically designed secondary

sersrsquo methodologies

There has been much debate on the methodology that IEEE

0211af secondary devices should adopt to get knowledge of free

hannels at a given time on a given target communication area in

n effort to guarantee verified optimised and reliable TVWS spec-

rum use The three main methodologies discussed include sens-

ng solutions geolocation databases (DB) and beaconing Moreover

wo approaches are possible in general distributed and centralised

ith their well known tradeoffs in terms of effectiveness required in-

rastructure deployment and coordination overhead The distributed

B Bellalta et al Computer Communications 75 (2016) 1ndash25 15

m

b

s

c

t

u

a

t

o

l

t

w

t

t

a

o

c

c

a

(

t

o

t

t

D

t

t

o

t

f

a

t

c

i

a

c

d

t

s

t

b

I

l

(

e

a

i

R

p

e

t

c

c

i

b

i

i

t

b

t

t

t

W

a

p

c

i

t

h

p

t

n

a

t

e

t

t

t

p

t

a

r

8

w

s

w

[

i

S

o

a

t

f

i

8

i

r

a

a

i

t

c

(

I

s

a

e

t

t

c

[

i

c

i

g

s

n

s

t

i

o

o

s

c

D

o

ethodology that is receiving more attention in the literature is

ased on distributed channel sensing solutions [127144145] These

olutions include the observation of spectrum use and possible

ooperative aggregation of spectrum sensing information to increase

he accuracy of detection and reduce the vulnerability of primary

sersrsquo transmissions The spectrum sensing capability is required in

ll the secondary devices with a minus114 dBm sensitivity Dynamic

ransmission power control must be provided and the upper limit

n emissions is 100 mW EIRP (20 dBm) for portable devices (further

imited in the case of adjacent channel use to reduce out-of-band in-

erference) However database (DB) coordination of TVWS combined

ith spectrum sensing is considered the most promising and effec-

ive technique compared with spectrum sensing alone The DB spec-

rum information has more chances to be effective (ie identifying

ll free channels opportunities) and reliable (ie not prone to attacks

r erroneous interpretation of channels status) In fact regulators de-

ided to push for the conservative solution of a remotely accessible

entralised spectrum DB that maintains the information on the avail-

bility of all TVWS channels at any given point in time and location

with a target accuracy around 50 m) under their responsibility For

his reason the spectrum databases have become a mandatory part

f many spectrum sharing systems and the dominant technical solu-

ion to support TV white spaces

The IEEE 80211af standard adheres to this vision and also has

he role of a common regulatory framework for different spectrum

B implementations defining a generalised coordination architec-

ure protocols and interfaces for spectrum queries and local spec-

rum information management The protocol explicitly considers out

f scope the communication protocol and technology adopted in

he background leaving the freedom to those players responsible

or the deployment to select any suitable Internet-based and local

ccess network technology Secondary devices that need to access

he TVWS spectrum should get information on which channels they

an effectively use and which parameters to adopt without impact-

ng primary usersrsquo transmissions (and receptions) by querying the

vailable spectrum DB The query must contain the transmission

haracteristics and accurate geographical position of the secondary

evice which can be determined by means of a geo-positioning sys-

em (GPS) regularly updated in the case of movement (or manually

et for static devices) The IEEE 80211af reference system architec-

ure for DB spectrum access is composed of multiple Geolocation-

ased Spectrum information Databases (GDBs) entities connected via

nternet to Registered Location Secure Servers (RLSS) RLSSs work as

ocal proxies of the GDB for a localised group of Basic Service Sets

BSS) RLSS are connected via a secure protocol architecture with the

quivalent of multiple Access Points (AP) in different BSS realising

trusted DB infrastructure APs locally coordinate the exchange of

nformation and channel access management between the GDB (via

LSS) and the secondary usersrsquo end stations (STA)

Mode I devices are those under the control of a device that em-

loys geo-location database access while Mode II devices are those

mploying geo-location database access by themselves The informa-

ion exchange provided between the GDB and a secondary user STA

an be provided in both an open-loop (eg adopted by FCC) and a

losed-loop (eg adopted by ETSI) implementation In an open loop

mplementation daily spectrum availability information is provided

y the GDB and no feedback on the spectrum information received

s provided by STAs thus the system approach for spectrum access

s much more conservative leading to low channelsrsquo utilisation po-

ential In a closed loop implementation the STA can provide feed-

ack to the GDB and there are more communication overheads due

o the high granularity of updates however the system is more effec-

ive and reliable in the exploitation of TVWS on behalf of STAs The

ypical information provided by the GDB includes (1) the updated

hite Space Maps (WSM) of frequencies allowed for secondary use

t the timespace of the querying STA and (2) the device-dependent

ower limitations for transmission (in general conservative and ac-

urate enough to avoid relevant interference effects on primary users

dentified in the area)

The basic IEEE 80211af mechanisms regulating the communica-

ion between STA and GDB can be found in [137138]

In parallel with IEEE 80211af standards such as the IEEE 19006

ave been created that consider interfaces and data structures sup-

orting spectrum sensing information exchange applicable to spec-

rum sensing (and particularly distributed spectrum sensing) sce-

arios In particular in October 2014 the IEEE 19006 WG initiated

project for a new standard called IEEE 19006b [146] concerning

he use of spectrum sensing information to support and optimise the

ffectiveness reliability and robustness of spectrum database solu-

ions The aim is to enhance the performance and capabilities of spec-

rum databases through the use of spectrum sensing information

Co-existence Support for co-existence mechanisms so that mul-

iple technologies can effectively utilise the TVWS spectrum is im-

ortant Self co-existence between network devices of a common

echnology (eg deployed by different operators in the same area)

nd co-existence among different technologies are relevant topic of

esearch for Cognitive Radio (CR) systems and specifically for IEEE

0211af Many solutions appeared on the research scene but no one

as so far finalised as the target solution for IEEE 80211af Specific

tandardisation has been started to regulate coexistence between

ireless standards of unlicensed devices including the IEEE 802191

147] The purpose of the IEEE 802191 standard is to enable the fam-

ly of IEEE 802 Wireless Standards to most effectively use TV White

pace by providing standard coexistence methods among dissimilar

r independently operated TVWS devices Early examples of gener-

lised coexistence mechanisms included Dynamic Frequency Selec-

ion (DFS) Transmission Power Control (TPC) listen before talk (eg

or contention based IEEE 80211 80215) time division multiplex-

ng (also among different techniques such as the IEEE 80216 80220

0222) and Message-based Spectrum Contention (that is beacon-

ng messages that carry coexistence information) Opportune met-

ics must be defined to assess the measurable coexistence achieved

mong different technologies as an example the hidden node prob-

bility for a target scenario or the estimate of percentage variation

n normalised network throughput and latency (before and during

he SU transmissions) On the other hand a centralised coexistence

ontrol mechanism could be effectively realised by a central manager

or coexistence-DB like the GDB) in critical scenarios To this end

EEE 19004 [148] (a standard for heterogeneous networks in dynamic

pectrum context part of IEEE Standards Coordinating Committee 41)

ims to standardise the overall system architecture and information

xchange between the network and mobile devices which will allow

hese elements to optimally choose from available radio resources

There are three possible classification of co-existence archi-

ectures and inter-network coordination channels for CR systems

entralised coordinated and autonomous co-existence mechanisms

149] In centralised co-existence schemes the co-existence is admin-

stered by a central entity (eg like in IEEE 802191) This solution

ould be applied in both homogeneous IEEE 80211af systems (eg

ncarnated by the centralised spectrum DB entity) and in hetero-

eneous inter-networks without requiring modifications to existing

tandards under the assumption that all the involved co-existing

etwork entities would adhere to the same centralised coordination

cheme and inter-network coordination protocol On the other hand

his assumption could be hard to satisfy in some practical scenar-

os The centralised schemes are effective in providing minimisation

f inter-network interference based on the availability and quality

f centralised co-existence information In coordinated co-existence

chemes the centralised entity could be present but not taking de-

isions in general The central entity could provide a co-existence

B with required information which can be used via in-band and

ut-band signalling for co-existence decisions taken by the cluster

16 B Bellalta et al Computer Communications 75 (2016) 1ndash25

c

e

n

c

o

s

i

d

a

S

8

n

a

b

o

b

i

t

a

8

b

[

t

i

n

b

5

t

a

c

b

s

a

c

n

s

p

i

c

l

r

c

f

a

v

t

a

e

c

t

Q

a

b

s

i

c

u

n

c

i

heads of many co-existing inter-networks for implementing a proper

common coordination protocol (eg out-band busy tone signalling)

These are often hybrid solutions realising a compromise between ef-

fectiveness and suitability In autonomous co-existence schemes all the

decisions and coordination are implemented in a distributed way

by the involved network entities In these schemes many policies

can be adopted to realise a sufficient level of co-existence under a

best effort approach Solutions that could be implemented include

the use of busy tones beaconing and other signalling protocols in

dedicated control channels dynamic distributed frequency selection

schemes listen before transmit token-based and dynamic reserva-

tion schemes All the above mentioned solutions have goods and

bads and a general illustration can be found in [149] Other proac-

tive co-existence techniques try to early detect and recovermitigate

co-existence issues with relaxed inter-network coordination eg re-

alised via spectrum sensing and interference avoidancesuppression

techniques

It must be clear that different aims exist regarding the co-

existence for primary TVWS users protection against secondary

users and for secondary devices mutual interference avoidance

Many co-existence problems have been analysed in the literature

for TVWS technologies in particular between the primary (licensed)

and secondary (unlicensed) devices The coexistence management in

TVWS is complex because primary users of DVB transmissions are in-

tended to be the pure receivers rather than the broadcasters In other

words potential protected users could be everywhere and the hid-

den terminal problem could arise for these systems The co-existence

problem is also complicated by many factors that cause asymme-

try and dynamicity in the TVWS such as the mobility variable den-

sity power asymmetry and heterogeneous MACPHY layers TVWS

co-existence problems originate between the IEEE 80211af technolo-

gies and the IEEE 80222 technologies for wireless regional area net-

works (WRAN) Results of analysis of TVWS usage in Europe show

that white spaces are typically present and fragmented They are typ-

ically more abundant in rural areas where larger contiguous blocks

of unused channels are available due to broadcast network planning

giving priorities linked to population density However the exploita-

tion of TVWS in urban areas is possible eg some recent research

was realised under the assumption of exploiting shadowing effects

created in the communication environment (eg by buildings obsta-

cles) in favour of frequency reuse [150]

Since different standards for opportunistic communication in the

TV White Spaces have now been published such as IEEE 80222 [151]

IEEE 802154m [152] Weightless [153] and of course IEEE 80211af

improved methods to guarantee the coexistence of different devices

operating on several protocols in the same bands must be deployed

We highlight here some work eg [150154ndash157] that focuses on

the coexistence between different technologies operating in the same

bands In particular [155] studies the performance degradation of

IEEE 80222 when an IEEE 80211af network is operated in the same

area The problem becomes even worse when the IEEE 80211af net-

work is located near 80222 user equipment causing both networks

to perceive a strong interference from each other One proposed solu-

tion is the Coexistence Beacon Protocol [158] studied for 80222 net-

works which foresees the exchange of a periodic beacon to identify

the neighbouring and possibly interfering networks

For what concerns the co-existence problems research has ad-

dressed other important related aspects of new WLANs on TV White

Spaces For instance the higher coverage range makes them attrac-

tive for applications in the smart grid and also for M2M on TVWS

[134150159] Here the increased range compared to standard tech-

nologies in the ISM bands can have beneficial effects for indoor

mobile devices [150] and M2M [159] Specifically indoor devices

communicating on TV White Spaces offer better propagation char-

acteristics and penetration through obstacles making it easier to

realise scalable home connectivity [135] although interference and

o-existence problems could be exacerbated and must be resolved

g see [139] However studies in the literature show how the sig-

al coming from an indoor transmitter on TV White Spaces remains

onstrained inside the house making it difficult to deploy indoor-to-

utdoor networks On the other hand this shadowing limit can be

een as an opportunity lowering the interference outside the build-

ng in which the TVWS network is locally operated [134150] Un-

er this approach HDTV streaming has recently been considered as

possible use case of WLAN networks operating in the TV White

pace [160]

Non-contiguous channel bonding A relevant novel feature of IEEE

0211af is the potential for contiguous and non-contiguous chan-

el bonding which permits aggregating basic channels (also non-

djacent ones) and leveraging the possible large frequency spread

etween multiple available channels Due to the rather static nature

f primary DTV transmissions where a busy channel is unlikely to

ecome free in the near future and state changes are coarse-grained

n general it is crucial to exploit the time-locality effect and exploit

he maximum physical channel availability that could be aggregated

t any given location With the methodology inherited from IEEE

0211ac IEEE 80211af is capable of bonding together two up to four

asic channels grouped in up to two different non-contiguous chunks

137] The spectrum bandwidth of a DVB-T basic channel can be ei-

her 6 7 or 8 MHz depending on the country in which the service

s operated As an example this creates a 144 to 168 OFDM chan-

elsrsquo bandwidth potential when up to four 6ndash7ndash8 MHz channels are

onded [137]

12 Open challenges

The deployment of WLANs in the TVWS at their maximum poten-

ial still requires the resolution of open problems Generally speaking

s mentioned in previous sections the challenges for such networks

an be divided into three main categories (1) spectrum sharing

etween opportunistic (secondary) and heterogeneous devices (2)

pectrum sensingmanagement and maximum exploitation of avail-

ble spectrum (potentially enabling the spectrum-on-demand con-

ept) and (3) co-existence and interference mitigation to the primary

etwork (primary user protection) and secondary networks [154]

The overall aim of spectrum sharing techniques is to maximise

eparation to avoid overlapping or contiguous channels operations

rovided that sufficient channels are available in the TVWS spectrum

n a given spacetime scenario [154] In general important research

ontributions have to be realised for the design of proper spectrum al-

ocation methodologies and techniques satisfying multi-factorial QoS

equirements at the system-level and the user-level Novel ideas in-

lude the spectrum sharing and spectrum sensing techniques being

urther divided into cooperative and non-cooperative

Non-cooperative solutions attempt to realise spectrum sharing

nd sensing on a local basis without direct cooperation between de-

ices Examples include the ldquolisten-before-transmitrdquo approach and

ransmission power control to limit the interference and spectrum

llocation policies that are based on local node feedback only Coop-

rative solutions rely on the existence of a common communication

hannel through which devices can tentatively agree on the spec-

rum allocation that provides the desired spectrum separation and

oS Major issues include the definition of a methodology to identify

common channel (or multiple channels mutually shared in space

y pairs of heterogeneous and distributed devices) and to efficiently

hare common information For the access to the spectrum DBs the

dentification of proper common access channels is still under dis-

ussion out-of-TVWS-band cellular communications can be widely

sed given their high coverage also indoor

Trying to avoid the use of complex and resource-hungry coordi-

ation schemes and dedicated control channels rendezvous proto-

ols play a significant role both for co-existence and for establish-

ng dynamic communication channels based on spectrum sharing

B Bellalta et al Computer Communications 75 (2016) 1ndash25 17

R

n

c

t

l

o

m

e

n

i

n

w

n

a

u

s

p

b

s

m

s

a

d

t

b

a

a

m

a

t

u

8

n

t

d

d

m

f

t

n

d

n

e

r

a

h

b

f

a

c

t

u

i

f

t

d

t

s

i

f

I

l

i

a

n

c

d

I

n

o

t

i

t

e

w

t

c

m

t

a

i

t

q

b

c

a

i

l

i

a

v

t

a

a

c

t

i

t

b

t

t

a

D

p

c

s

o

q

r

p

a

g

t

e

u

d

t

m

n

t

s

f

i

1

8

f

T

endezvous is considered a fundamental problem in generalised CR

etworks at the basis of many communication processes eg in-

luding neighbour discovery routing and broadcast Rendezvous pro-

ocols attempt to establish a new link for communication on a se-

ected and agreed frequency band (channel) identified within a set

f available resources (channels) thus creating the basis for com-

unication between two or more SUs This problem is often exac-

rbated by the high (and unknown) numbers of contending SUs high

umber of available channels and their ambiguous boundaries and

dentifiers and tough co-existence requirements satisfaction in cog-

itive radio systems [161] Preliminary approaches for rendezvous

ere based on the existence of a dedicated Common Control Chan-

el (CCC) agreed among all the SUs or the existence of a centralised

gent (decision maker) assigning channels to any pairs of requesting

sers This approach could be considered in the case of IEEE 80211af

olutions based on the spectrum DB implementation however the

roblem is complicated by the fact that the common channel must

e identified and made available for all the requesting SUs at the

ame time even when these are spread over multiple collision do-

ains in space On the other hand the scalability and reliability is-

ues which are caused by the congestion of CCCs and the vulner-

bility risks of centralised DBs motivate the research of alternative

istributed solutions called blind rendezvous protocols [162] Dis-

ributed solutions have been mainly based on beaconing andor slot-

ased channel hopping algorithms exploiting randomisation sensing

nd discoverysynchronisation messages spread in slotted time over

set of candidate available channels [161] These protocols aim to

inimise the convergence time to maximise the use of resources

nd to avoid collisions among SUs by identifying and agreeing on

he use of a given common channel among N non-overlapping and

nambiguously labelled channels The main complication in IEEE

0211af systems is given by the wide availability of many different

on-contiguous spectrum bands which cannot be uniformly quan-

ised in single channels and unambiguously labelled in space in or-

er to realise a common reference domain for all the IEEE 80211af

evices In fact the knowledge of the number of channels their com-

on labels the unique IDs of SUs and the number of SUs contending

or the channel assignment are the key factors determining the effec-

iveness and good properties of blind rendezvous protocols Recently

ew classes of rendezvous protocols have been proposed which are

istributed blind and oblivious Oblivious means that available chan-

els may be identified with different labels on behalf of the differ-

nt involved SUs A number of these distributed blind and oblivious

endezvous protocols have been proposed and have been referenced

nd analysed in [162] To conclude solutions for dynamic spectrum

andoff issues (eg in case PUs suddenly appear in a region) must

e resolved in advance to minimise the latency of re-establishing a

unctional communication channel eg by implementing a proactive

lternative rendezvous channel selection in background to ongoing

ommunication processes [163]

Another challenge is to provide cooperative techniques in a dis-

ributed vs centralised implementation by analysing their convergence

nder variable conditions overheads and tradeoffs In some cases as

n vehicular networks the exploitation of scenario characteristics and

actors such as the constrained mobility could help to realise effec-

ive dissemination of sensing information over an extended and pre-

ictable horizon under a cooperative approach [164ndash166] Coopera-

ion could be helpful in the realisation of the cooperative spectrum

ensing in combination with the Geolocation DB approach Another

nteresting direction of research is the investigation of mutual ef-

ects of coexisting cooperation-based vs non-cooperative techniques

n cooperation-based devices the exchange of information could al-

ow a quick convergence to a solution and avoidance of further

nterference [145] Certainly cooperative techniques have desirable

dvantages over non-cooperative ones However tight time synchro-

isation and high coordination overheads are required The common

hannel for cooperation purposes could become a bottleneck and re-

uce the potential advantages of dynamic spectrum allocation The

EEE 80219 foresees the presence of a shared common control chan-

el (CCC) to which devices need to tune in order to gather the state

f the network Solutions such as [141] propose the use of TVWS for

he CCC

A promising direction for research is the adoption of cluster-

ng schemes for secondary devices associated with a hybrid dis-

ributedcooperative vs centralisednon-cooperative approach in an

ffort to reduce overheads while maximising advantages In this

ay secondary nodes could implement cooperation for sensing func-

ions using properly identified common channels and delegate the

entralised decisions and spectrum-DB management to well instru-

ented cluster-leader nodes This concept is incarnated to some ex-

ent by RLSS in IEEE 80211af More recently there have been propos-

ls to build distributed databases that refresh their contents period-

cally by querying the remote spectrum database [167] Here Mas-

er devices (according to Ofcom terminology) periodically cache the

uery replies from the spectrum database in order to reply to and

roadcast the spectrum availability to neighbour Slave devices

Another challenging direction for research is the definition of ac-

urate models and efficient simulation tools enabling dynamic spectrum

nalysis for both rural and urban areas Recent developments of dig-

tal maps simulation tools and propagation models theoretically al-

ow improving the capability of predicting radio propagation effects

n complex scenarios with an acceptable computation time This en-

bling technology could be used in parallel to spectrum DBs to pro-

ide more accurate forecasting of dynamic frequency allocation in

imespace scenarios Another issue to be taken into account is the

ccuracy of the remote spectrum database regarding the channel

vailability estimation Several works have already shown the inac-

uracy of propagation models due to the complexity of parameters

o be considered and differences between modelled and real scenar-

os [132134168169] Better propagation models will make it possible

o more efficiently estimate the interference between devices and

uild Radio Environment Maps (REM) or White Space Maps (WSM)

hat will possibly lead to a more accurate sharing of the radio spec-

rum [144] To this end current research could focus on how to lever-

ge simulation models in order to build more accurate frequency

Bs

In general the adoption of remote spectrum DBs methodology

ushed by national regulators has made the spectrum sensing pro-

ess optional and conservatively realised (eg power control for

ensing-only devices is limited to below 17 dBm EIRP) Thus much

f the research has focused on techniques and solutions to successfully

uery the remote database also in challenging environments such as

ural areas and indoor However since the queries must provide the

osition of the mobile device the accuracy of positioning for mobile

nd handheld devices still needs more appropriate solutions Trian-

ulation can be used to estimate the position of mobile devices with

he accuracy needed by the current regulations outdoor (50 m) How-

ver a low GPS accuracy is possible indoors [170] Certified man-

al positioning can be provided by technicians for static and indoor

evices although limiting mobility Alternatives to costly infrastruc-

ures for positioning (eg based on short range beaconing devices)

ust be deployed in indoor scenarios

Finally regarding spectrum sensing and interference mitigation

ew challenges come from the heterogeneity and different charac-

eristics that networks operating in the TV White Spaces can have

uch as different transmitting power (4 W for static devices 100 mW

or mobile and portable devices and 40 mW for communication

n channels adjacent to occupied ones) different bandwidths (5

0 15 and 20 MHz are currently foreseen for IEEE 80211af IEEE

02154m can have narrower bandwidths or wider ones) and dif-

erent medium access schemes (protocols that use either CSMA or

DMA are required to co-exist in the TV White Spaces) With all this

18 B Bellalta et al Computer Communications 75 (2016) 1ndash25

d

a

T

i

a

c

t

l

b

o

t

a

d

p

a

r

p

n

a

m

t

p

p

d

d

o

t

i

D

w

p

i

a

a

d

b

fi

d

o

t

l

b

t

b

a

i

e

o

o

c

t

t

w

A

c

M

s

c

l

O

w

g

v

t

in place the problem of determining whether another secondary de-

vice is currently transmitting in the same channel is still an open

issue

Spectrum sensing is a very hot topic in the literature with many

proposals that leverage the cooperation between devices to raise the

accuracy of the detection process eg [144171] The current litera-

ture also offers room for proposals that suggest exploiting sensing to

build more accurate spectrum DBs [164172173] Lastly regarding the

interference mitigation to the primary network regulators provide a

conservative approach based on a remote spectrum database to avoid

high risk for the primary users However in particular for the indoor

scenario the concept of secondary usersrsquo transmission in TV Grey

Spaces has been proposed TV Grey Space identifies busy channels

that are formally used by primary users in the area of interest (eg

at the rooftop level of a building for DVB-T) but which could be con-

sidered usable indoors (eg at the basement level) without causing

effective interference to the primary users [134150174] This concept

is based on the generalised spectrum underlay paradigm [175] The

remaining research challenge is to provide TV Grey space access guar-

anteeing a sufficient amount of protection to the primary user while

providing a satisfactory QoS to opportunistic devices eg through

the use of customised and validated propagation models for indoor

TVWS networks and feedback mechanisms to constantly monitor the

interference generated to the primary network

6 Emerging new trends and technologies

In this last section we review three emerging trends that in our

opinion will have a large influence in the conception of future WLANs

as they change the way WLAN protocols and functionalities are de-

veloped implemented tested and integrated with other wireless

networks

61 Programmable Wireless LANs

Especially in the enterprise environment WLAN deployments

need to support a wide range of functionalities and services This is

intrinsically difficult because of the large number of APs that must be

managed which calls for scalable solutions Typical services include

channel assignment load balancing among APs authentication au-

thorisation and accounting (AAA) policy management support for

client mobility and interference coordination Another problem is the

fact that WLAN clients autonomously take several decisions such as

which APs to associate with when to hand-over etc Therefore sup-

porting roaming clients requires the management of a large number

of association states across several APs which is a challenge if sup-

port for real-time hand-over is desired Typically such management

schemes are centralised and most of them are proprietary such as

WLAN controller solutions from Aruba [176] or Cisco [177] although

the IEEE 80211u amendment has been released to allow mobile users

to seamless roam between WiFi networks with automatic authenti-

cation and handoff [178] For example Dyson [179] enables STAs to

send information such as radio channel conditions to a centralised

controller based on a custom API (eg based on Python) As the con-

troller has a centralised view of the network it can enforce a rich

set of policies to control the network also using historical informa-

tion A demo system has been implemented along with applications

such as airtime reservations for specific clients or optimised hand-

offs However Dyson requires STAs to be modified in order to use

those new services offered by the centralised controller TRANTOR

[180] is another example of a centralised management system that

requires changes of clients in order for the infrastructure to gather

information from them in terms of eg interference measurements

In addition control commands enable the infrastructure to exercise

control such as modifying the transmit power or influencing the as-

sociation procedure Clients still use standard CSMA MAC layer for

ata transmission In contrast to DenseAP [181] or DIRAC [182] which

re also based on a centralised controller but do not modify clients

RANTOR exercises much more control on clients by the use of a ded-

cated API which enables a significantly higher gain in coordination

nd thus capacity CENTAUR [183] is another example of a centralised

ontroller that centrally schedules hidden and exposed terminals in

he downlink while using standard CSMA MAC for uplink traffic and

egacy downlink traffic To achieve good performance it uses a fixed

ack-off packet staggering techniques and a hybrid data path that

nly schedules downlink transmissions towards hidden and exposed

erminals centrally All other traffic is sent using standard DCF with

standard back-off procedure In contrast to previous approaches it

oes not require any modifications to the clients but does require data

lane centralisation and it remains questionable how scalable such

solution is for large WLAN deployments with high PHY layer data

ates beyond 100 Mbps However new approaches for fast data plane

rocessing that are currently explored in the virtualisation commu-

ity which move all the packet processing into user space may be

n option to significantly increase speed Designing such manage-

ent systems raises several interesting research questions (a) Cen-

ralised infrastructures may incur high latency but are more simple to

rogram and maintain while designing a completely distributed ap-

roach that operates close to WLAN devices may result in significant

istributed coordination and consistency problems (b) what level of

istributed control is required in order to support the flexibility needs

f future WiFi based networks supporting a high level of mobility

Traditionally WLAN APs are built on proprietary operating sys-

ems that are tightly coupled with the hardware This design makes

t hard to create new applications on top of such networking devices

espite the fact that protocols and mechanisms are available that

ould greatly improve the utility of existing networks those new

rotocols are not deployed because the closed system design makes

t very difficult to extend their functionality An important aspect for

ll such new approaches is therefore how to provide open interfaces

nd open source to speed up innovation As an example Linux-based

evices are completely open source but in order to increase flexi-

ility much more work is needed in the area of open drivers and

rmware In fact the Atheros based ath9k driver has been the main

river of innovation in the open source community because it is the

nly driver that interworks with open firmware In general solutions

hat modify the OS driver focus mainly on programmable network

evel solutions for WLANs such as channel switching or handovers

etween APs While the ath9k driver is an excellent example in how

o speed up innovation it is still very cumbersome to support flexi-

le MAC engine reprogrammability with ath9k In contrast MAClets

nd Wireless MAC processors (WMP) [184] allow a much more flex-

ble reprogramming of MAC functionalities A MAC processor is an

ntity able to execute general MAC commands that specify the MAC

perations through a software-defined state machine The behaviour

f the MAC protocol can therefore be updated at run-time by simply

hanging the sequence in which those commands are executed (ie

he MAClets) As a proof-of-concept the authors of [184] implement

he proposed MAC processor solution in a commodity WLAN hard-

are card extending the basic DCF in three directions piggy-backing

CKs a pseudo TDMA and the use of multiple channels In [185] a

ontrol framework for this WMP system is also proposed to support

AClet code mobility ie for moving loading and activating MAC

oftware programs embedded into ordinary data packets (akin to the

apsule model of traditional active networks)

The difficulty in re-programming networking hardware has also

ed to the concept of Software Defined Networks (SDNs) based on the

penFlow protocol [186] The main idea of SDNs is to extend net-

orking devices with standardised APIs that allow third-party pro-

rammers to flexibly control the data path In addition SDNs pro-

ide higher level abstractions to network designers and programmers

hrough the use of a centralised control plane offered by a network

B Bellalta et al Computer Communications 75 (2016) 1ndash25 19

Fig 11 The CLOUDMAC architecture VAP stands for Virtual Access Point and WT

stands for Wireless Termination

c

a

c

t

m

W

n

t

t

c

p

t

u

h

a

i

e

s

o

c

a

a

f

T

r

s

a

s

p

o

M

C

i

T

t

fl

A

a

a

i

c

t

s

t

e

I

t

w

H

e

b

I

l

d

o

6

r

W

s

l

a

e

e

a

s

b

c

t

(

S

p

c

F

B

t

p

T

e

t

s

u

m

i

e

o

a

A

p

d

t

t

n

p

i

t

t

f

i

ontroller such as NOX [187] which allows reuse of components such

s topology discovery or network access control for different appli-

ations The SDN concept has recently been applied to WLAN archi-

ecture in order to enable fine grained control over mobility manage-

ent and data forwarding focusing on a programmable enterprise

LAN architecture An important part of the SDN architecture is the

etwork controller which provides a centralised view of (parts of)

he SDN enabled network and uses the OpenFlow protocol to install

he forwarding rules on SDN-enabled devices (routers switches ac-

ess points cellular base stations etc)

ODIN [188] is designed to support programmability in enter-

rise WLAN architecture by separating the association state from

he physical AP They implement Light Virtual Access Points (LVAP)

sing a Split-MAC approach where the infrastructure controls the

andover procedure among different WLAN APs By managing

ssociations through SDN controllers ODIN enables proactive mobil-

ty management and load balancing within the SDN enabled WLAN

nterprise network without the need for changes in the client WLAN

tack or IEEE 80211 MAC layer While ODIN requires agents to reside

n the APs that communicate with the Odin Master within the SDN

ontroller CLOUDMAC [189] is based purely on the concept of SDN

nd virtual APs Similar to ODIN CLOUDMAC implements a Split-MAC

pproach but in addition enables the processing of WLAN MAC layer

rames within a co-located Cloud using so-called Virtual APs (VAPs)

he physical APs in CLOUDMAC are lightweight WLAN APs that are

esponsible only for sending out IEEE 80211 based MAC layer ACKs to

tandard WLAN clients and tunnel WLAN MAC layer frames through

n SDN-enabled enterprise WLAN towards the VAPs As association

tates are kept in the VAPs fast mobility is supported using sim-

le OpenFlow forwarding rules Because of the additional processing

f IEEE 80211 WLAN MAC frames in the co-located Cloud CLOUD-

AC has higher latency than a standard WLAN deployment However

LOUDMAC offers a Webservice based API to third party applications

n order to program the enterprise WLAN and enable new services

he architecture of CLOUDMAC is depicted in Fig 11 It has been ex-

ended in [190] to support QoS management and in [191] to support

exible MAC management frame prioritisation based on IEEE 80211

system based on OpenFlow has been recently proposed in [192] to

llow the station to be associated with multiple APs simultaneously

nd to switch between APs with low overhead

An important aspect to consider in order to enable programmabil-

ty is backwards compatibility Several approaches (eg [180]) require

lients to be modified to utilise the features provided This is difficult

o do in practice because it requires changes in the operating system

oftware of all clients If not all clients can utilise those APIs it is ques-

ionable how usable the new architecture will be and how much ben-

fit in terms of aggregate performance such architecture will allow

n contrast the SDN based approaches are interesting in the sense

hat they do not require changes in the client WLAN stack and work

ith the standard IEEE 80211 MAC layer deployed within the clients

owever it remains questionable how scalable such solutions are For

xample an interesting open research topic is to evaluate the scala-

ility of approaches such as CLOUDMAC [189] In addition processing

EEE 80211 MAC frames within co-located private Cloud requires low

atency support from local Cloud solutions (such as OpenStack) in or-

er to reduce the IEEE 80211 MAC processing time which is an area

f active research

2 Prototyping and testing IEEE 80211 enhancements

Most of the new proposals for next-generation WLANs are cur-

ently only evaluated using mathematical analysis and simulation

hile both analysis and simulation are necessary to characterise and

tudy those enhancements in the initial design phase or to consider

arge-scale scenarios it is difficult to consider all practical aspects of

real-world scenario This can sometimes cause significant differ-

nces between what simulations and real experiments show How-

ver real experiments are challenging because of the high complexity

nd costs of building the new hardware and software for each specific

olution to test

To mitigate implementation complexity there are several flexi-

le hardware platforms where low-level MAC mechanisms can be

ompletely implemented in software Examples based on FPGAs are

he USRP (Universal Software Radio Peripheral) [193] and the WARP

Wireless Open-Access Research Platform) [194] As an alternative

ORA (Microsoft Research Software Radio) [195] works on general

urpose computers by taking some advantage from modern multiple

ore systems

A different approach is provided by OpenFWWF [196] Open-

WWF provides an open CSMACA firmware for specific models of

roadcom chipsets so the resulting firmware can be uploaded and

ested in real commercial hardware OpenFWWF implements a sim-

le State Machine (SM) for controlling the hardware in real time

he SM evolution is driven by a main loop that reacts to events by

xecuting specific handlers When a packet originally prepared by

he Linux kernel is ready in the NIC memory handler Packet_Ready

ets up the radio hardware according to the packet metadata sched-

les the transmission and jumps back to the main loop The Trans-

ission Engine (TXE) then takes care of accessing the channel ie

t decrements the back-off counter according to the DCF rules and

ventually starts the actual transmission This triggers the execution

f the TX_frame_now event that prepares the ACK time-out clock

nd finalises the MAC header If the ACK-frame is received or if the

CK time-out expires and the maximum number of attempts for this

acket is reached handler Update_Params resets the contention win-

ow to its minimum value or otherwise doubles it Finally it loads

he back-off counter with a fresh value

In the following we will review a list of selected papers that use

he OpenFWWF firmware and the WARP or USRP platforms to test

ew proposals for WLANs in real scenarios as examples A first im-

lementation of the groupcast mechanisms defined in IEEE 80211aa

s presented in [197] The authors modify the OpenFWWF firmware

o include those functionalities and evaluate them using a 30 STAs

estbed A second example of the use of OpenFWWF is the collision-

ree MAC protocol presented in [198] In [198] the authors design and

mplement in OpenFWWF a MAC protocol that is able to achieve a

20 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 12 LTE to WIFI offloading

a

n

c

t

[

d

a

W

v

t

t

o

b

f

S

g

l

i

I

s

I

b

n

s

m

a

p

d

c

N

o

v

t

t

n

l

i

a

t

a

collision-free operation by waiting a deterministic time after success-

ful transmissions before to start a new transmission This work shows

the experimental performance of a collision-free MAC (CF-MAC) pro-

tocol for WLANs using commercial hardware The results show that

the proposed CF-MAC protocol leads to a better distribution of the

available bandwidth among users a higher throughput and lower

losses than DCF

In [199] the authors use the WARP platform to test a variant of the

DCF ndash called IEEE 80211ec ndash that substitutes control packets such as

the RTS CTS and ACK with short detectable sequences Since those se-

quences are shorter than the control packets and can be detected cor-

rectly even at lower SNRs values a significant gain in performance is

achieved MIDAS (Multiple-Input Distributed Antenna Systems) [200]

implementation using the WARP platform shows the benefits of dis-

tributing the antennas over the area to cover instead of co-locating

them at the AP in terms of capacity when MU-MIMO is employed

The authors also propose a new MAC protocol to benefit from the

spatial reuse that the DAS allows taking as a basis the IEEE 80211ac

Experimental results show that their proposal is able to achieve up

to 200 gains versus the traditional approach where all antennas are

co-located at the AP

The USRP platform has recently been proposed to develop a first

implementation of IEEE 80211ah in order to obtain a preliminary ex-

perimental performance assessment of such a technology because

no commercial off-the-shelf hardware is yet available as the amend-

ment is still in progress [201] The USRP platform is also used in

[202] to evaluate TIMO (Technology Independent Multi-Output) a so-

lution to deal with high-power non-IEEE 80211 interferers in ISM

bands

To conclude this section it is worth mentioning here also to

MAClets and WMP as given such flexibility to implement and dis-

tribute new MAC protocols and other IEEE 80211 functionalities we

believe that if a use-friendly implementation of such a MAC proces-

sors framework is provided it would significantly contribute to the

development and testing of new MAC enhancements by the research

community

63 CellularWLAN interworking

Public hotspots that offer Internet access over a WLAN using IEEE

80211 technology are now nearly ubiquitous It is forecasted that the

cumulative installed base of WiFi hotspots worldwide will amount

to 551 million by 2018 excluding private hotspots (eg WiFi access

points deployed at home) [203] The sharp increase in the availabil-

ity of public WiFi was initially perceived by mobile cellular operators

as a threat due to the additional competition from wireline Internet

service providers or emerging crowdsourced WiFi networks such as

FON5 However as cellular operators are fighting to cope with the ex-

plosion of mobile data traffic created by the rising use of multimedia

content traffic over mobile devices [15] they are also starting to use

WLANs based on the IEEE 80211 technology to offload data from their

core and access networks In general mobile data offloading refers to

the use of complementary network technologies (in licensed or un-

licensed spectrum) for delivering data originally targeted to cellular

networks Intuitively the simplest type of offloading consists of ex-

ploiting connectivity to existing co-located WiFi networks and trans-

ferring data without any delay (Fig 12) Thus this offloading tech-

nique is known as on-the-spot offloading As a consequence of this

new trend the seamless integration of cellular (eg 3GLTE) and WiFi

technologies has attracted significant research interest in recent years

(see [204] for a survey) a few solutions have already been standard-

ised [205206] and roaming between cellular and WiFi is becoming

increasingly transparent to end users CellularWLAN interworking is

5 httpscorpfoncomen

fl

t

f

lso fostered by the support in the evolving 4G standards of heteroge-

eous network deployments (HetNets) in which the existing macro

ells are complemented with a number of small low-power base sta-

ions with the goal of increasing capacity in highly congested areas

207208] It is envisaged that small cells will be based on 4G stan-

ards (eg pico and femto cells) as well as IEEE 80211 technologies

nd multimode base stations that work simultaneously with LTE and

iFi are already entering the market

It is important to point out that LTE standards already support a

ariety of mechanisms that enable data offloading However most of

he existing solutions such as Local IP Access (LIPA) and Selected In-

ernet IP Traffic Offload (SIPTO) focus on data offloading at the core

f the network [205] For instance LIPA allows an IP-enabled mo-

ile device to transfer data to another device in the same pico or

emto cell without passing through the cellular access network while

IPTO enables the routing of selected IP data flows through different

ateways The only LTE offloading mechanism that supports seam-

ess interworking with IEEE 80211-based WLANs is IP Flow Mobil-

ty and Seamless Offload (IFOM) Specifically IFOM relies on Mobile

Pv6 technologies to allow a user terminal to simultaneously route

elected IP flows over different radio access technologies [209210]

n this way a user terminal can offload selected flows to a WLAN

ased on some operator-defined policy while keeping the LTE con-

ection running However to enable such an approach it is neces-

ary to support an entity in the cellular core network that can com-

unicate with the user terminal to exchange information about the

vailability and quality of neighbouring access networks as well as to

rovide the user terminal with predefined rules to manage the han-

over process This entity in current LTE standards is called the Ac-

ess Network Discovery and Selection Function (ANDSF) server [211]

ote that the problem of selecting the best communication technol-

gy in an heterogeneous wireless network has been extensively in-

estigated in the literature both with centralised as well as decen-

ralised approaches (see [212] for a survey) However the use of mul-

iple interfaces in parallel as well as per-flow offloading are relatively

ew concepts Thus the design of scalable and efficient network se-

ection strategies for the ANDSF framework is still an open issue For

nstance three offloading methods suitable for the ANDSF framework

re proposed in [213] which are based on coverage SNR and sys-

em load A reinforcement learning approach is designed in [214] that

llows multimode base stations to autonomously steer their traffic

ows across different access technologies depending on the traffic

ype the usersrsquo QoS requirements the network load and the inter-

erence levels A number of studies have also tried to quantify the

12 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 8 IEEE 80211ah PS mode for TIM-STAs non-TIM STAs and Unscheduled STAs

a

d

w

i

a

a

c

m

i

a

e

t

s

m

t

t

u

b

i

I

s

d

f

a

b

s

a

b

S

T

c

t

w

I

4

a

a

which addresses aspects not considered in [110] such as the use of

sectorisation to avoid overlapping between multiple IEEE 80211ah

WLANs and the use of relays to further extend the IEEE 80211ah cov-

erage among other aspects Finally in [113] both IEEE 80211ah PHY

and MAC characteristics are introduced with emphasis on the ben-

efits of using the sub-1 GHz ISM band in terms of link budget the

location of the OFDM pilots for outdoors operation the use of bidi-

rectional transmission opportunities and the new packet formats to

minimise overheads among others

Enhanced power saving mechanism PS mechanisms for WLANs

were already considered in the development of the first IEEE 80211

standard with the goal of improving the lifetime of battery equipped

devices [114] In PS mode STAs keep the transceiver in sleeping mode

as much time as possible They periodically wake up to listen to the

beacons transmitted by the AP Those beacons indicate whether an

STA has packets waiting for it at the AP In the positive case that STA

remains awake and requests the delivery of those packets Otherwise

given it has nothing to receive it returns to sleep mode until the next

beacon is expected

In the IEEE 80211ah amendment time is divided into pages DTIM

(Delivery Traffic Indication Map) periods TIM (Traffic Indication Map)

periods and slots DTIM and TIM periods begin with the correspond-

ing DTIM and TIM beacons sent by the AP The functions of DTIM and

TIM beacons are described below

1 DTIM beacons They inform as to which TIM Groups (ie the

group of STAs assigned to the same TIM period) have pending

packets at the AP

2 TIM beacons Each TIM message informs a TIM Group about

which specific STA has pending data in the AP Between two

consecutive DTIMs there are as many TIM beacons as TIM

Groups

Using this DTIMTIM-based approach any STA can enter into a

power saving state if it does not have packets pending for transmis-

sion and one of two conditions is met (1) it observes in the DTIM

beacon that there is no downlink traffic addressed to its TIM Group or

(2) it observes in the DTIM beacon that there is some downlink traffic

addressed to its TIM Group but that STA does not explicitly appear in

the TIM beacon Compared to the preliminary IEEE 80211 PS mecha-

nism this approach reduces the size of the Traffic Indication Map in

each TIM beacon thus reducing the overhead and the time STAs need

to listen and process them In addition TIM periods can be organised

in pages which further increases the number of TIM groups between

two DTIM beacons

The temporal organisation of pages DTIM and TIM periods is re-

ported in Fig 8 which shows the division of TIM periods in RAW (Re-

stricted Access Window) and PRAW (Periodic RAW) The RAW is a

time interval in each TIM period where TIM stations can transmit and

receive data (see below the different types of STAs defined in IEEE

80211ah) In addition it can be divided into several downlink and

uplink slots for further granularity In the downlink the slots are as-

signed to a single STA or a group of STAs while the uplink slots are

randomly selected by the STAs with packets ready for transmission

The PRAW is the period of time in each TIM where non-TIM stations

can transmit and receive data

Types of STAs IEEE 80211ah supports three types of STAs TIM

non-TIM and Unscheduled STAs A TIM station is assigned to a TIM

Group Their data transmissions must be performed within a RAW

(Restricted Access Window) period Non-TIM stations do not have to

listen to beacons to transmit data During the association process

non-TIM devices directly negotiate with the AP to obtain a trans-

mission time allocated in a PRAW The following channel access can

either be renegotiated or occurs periodically depending on the re-

quirements set by the STA Unscheduled STAs do not need to listen to

any beacons similar to non-TIM stations Even inside any restricted

access window unscheduled STAs can send a poll frame to the AP

sking for immediate access to the channel The response frame in-

icates an interval (outside both restricted access windows) during

hich unscheduled stations can access the channel This procedure

s meant for STAs that transmit data very sporadically

Hierarchical station organisation To support a large number of STAs

nd their organisation in pages DTIM and TIM periods IEEE 80211ah

ssigns to each associated STA a unique identifier of 13 bits which is

alled the Association Identifier (AID) Using this new AID the maxi-

um number of supported STAs is increased from the original 2007

n IEEE 80211 to 8191 (= 213 minus 1) in IEEE 80211ah However it also

llows categorising STAs according to the type of application they are

xecuting their power level or even their desired QoS by assigning

hem to different TIM groups

Long sleeping periods IEEE 80211ah offers TIM Non-TIM and Un-

cheduled STAs the possibility to set very long doze times (up to

onths) The corresponding clock drift produced by such long doze

imes must be taken into consideration however as the higher the

ime an STA has been asleep the further in advance it should wake

p to avoid possible synchronisation problems with the network

Efficient small data transmission Three new enhancements have

een proposed to reduce the overhead when the data packet size

s small First while IEEE 80211 contains a 28-byte MAC header

EEE 80211ah proposes a short 18-byte version by using AIDs in-

tead of MAC addresses Second IEEE 80211 has defined several null

ata packet (NDP) frames which consist only of a PHY header These

rames can be used to create short ACKs short Block ACKs short CTSs

nd short PS-Polls Finally a Fast Frame Exchange mechanism has

een developed so that if an STA has data to transmit it can notify a

uccessful reception by transmitting its data frame instead of an ACK

Sectorisation Since the PHY layer is based on the IEEE 80211ac

mendment single and multi-user beam-forming are also supported

y IEEE 80211ah This allows the transmission of data to multiple

TAs simultaneously in the downlink increasing the system capacity

he use of the beam-forming capability of IEEE 80211ah APs is also

onsidered as a means to group STAs into different independent an-

enna sectors with the main goal of reducing interference issues This

ould be particularly useful in the case of overlapping with other

EEE 80211ah WLANs or in the presence of hidden nodes

12 Open challenges

A first open challenge is to understand the coverage and achiev-

ble transmission rates in IEEE 80211ah WLANs in both indoor

nd outdoor scenarios Propagation models for WLANs working at

B Bellalta et al Computer Communications 75 (2016) 1ndash25 13

f

c

8

p

R

p

i

m

l

a

m

c

p

p

f

m

t

a

e

p

t

m

a

l

T

m

a

i

n

t

t

t

u

p

c

w

f

a

t

8

a

m

a

s

l

a

m

b

m

d

S

i

a

o

a

R

t

n

G

fi

s

t

p

t

c

u

c

w

i

o

p

g

a

c

i

f

c

(

d

m

l

i

R

t

l

t

L

m

a

m

o

e

f

a

I

i

t

w

i

t

a

c

w

p

f

t

d

5

i

s

e

s

w

5

a

c

a

n

requencies lower than 1 GHz are evaluated in [115] The authors

ompare two path loss propagation models proposed by the IEEE

0211ah Task Group (one for macro and one for picohotzone de-

loyments) [116] with Lee and HatandashOkumura propagation models

esults show that the IEEE 80211ah channel models underestimate

ath loss with respect to Lee and Hata models Moreover in a compar-

son with empirical data it is observed that the IEEE 80211ah channel

odels also underestimate the initial loss and the slope of the path-

oss function A new model parameterisation is thus proposed by the

uthors In [117] the feasibility of an IEEE 80211ah WLAN deploy-

ent is also evaluated in terms of the achievable range and bit rate

omputed on the basis of the link budget using the same path loss

ropagation models as in [115] Results show that the transmission

ower limitations in the uplink can limit the overall network per-

ormance Finally in [118] the authors evaluate the achievable trans-

ission range for the different transmission rates and the achievable

hroughput for different combination of transmission power values

nd transmission rates Further studies are required due to the het-

rogeneity of scenarios in which IEEE 80211ah WLANs can be de-

loyed to characterise the effective coverage ranges with special at-

ention to mixed indoors and outdoors scenarios and in presence of

obile nodes

Due to the wide coverage IEEE 80211ah WLANs may be severely

ffected by the presence of hidden terminals To mitigate such a prob-

em smart solutions to guarantee that the STAs assigned to the same

IM Group are inside each other transmission range or the employ-

ent of RTSCTS-based solutions may contribute to mitigate such

problem An overview of long-range scenarios for IEEE 80211ah

s presented in [119] including a detailed description of the hidden

ode problem in them

Single-hop uplink transmissions from distant STAs require high

ransmission power to reach the AP Since not all STAs may be able

o transmit at the required power a solution could be the introduc-

ion of nodes able to relay transmissions from them However the

se of relays has to be efficiently harmonised with the operation in

ower saving mode allowing temporal periods in which the relays

an gather the data from their associated STAs and then periods in

hich they forward the data to the AP which is still an open challenge

or IEEE 80211ah WLANs In case relays are used data prediction and

ggregation techniques could be implemented to make more efficient

ransmissions [120]

Another challenge is the assessment of the efficiency of IEEE

0211ah PS mechanisms including their optimisation In [111] the

uthors evaluate the capacity of the TIM and page segmentation

echanisms in terms of the maximum number of nodes supported

nd the energy consumed given a certain network traffic profile Re-

ults confirm that a large number of STAs can be supported with

ow energy consumption The impact of the number of nodes oper-

ting in PS mode on the energy consumption and the delay perfor-

ance is analysed in [121] A detailed analysis of the TIM Group-

ased channel access adopted by the IEEE 80211ah task group is

ade in [122] where the STAs assigned to each TIM are uniformly

istributed between the slots of the RAW Since only one group of

TAs is allowed to transmit in each RAW slot the channel contention

s minimised The number of downlink and uplink slots in a RAW

s well as their duration is optimised on the basis of the number

f active STAs and the network traffic profile in [123] In [124] the

uthors propose an algorithm to determine the optimal size of the

AW interval for uplink transmissions based on the estimation of

he number of STAs transmitting and the duration of the RAW Fi-

ally in [125] the authors address the contention problems in Smart

rid communication networks with many nodes and periodic traf-

c when using an IEEE 80211ah WLAN It is still a challenge to con-

ider smart systems able to adapt the IEEE 80211ah parameters to

he instantaneous system state in order to save as much energy as

ossible

Since the PS mechanisms affect only the operation of TIM STAs

he performance of non-TIM and Unscheduled STAs has not yet been

onsidered in the literature To identify the scenarios in which the

se of non-TIM and Unscheduled STAs are of interest is still an open

hallenge as further investigating how non-TIM STAs negotiate over

hich PRAWs they can use to transmit and receive data Co-existence

ssues between the three types of STAs in a single network is also an

pen challenge that requires further work to guarantee a minimum

erformance level for all of them

The design of strategies to distribute the STAs between all the TIM

roups based on their specific traffic profile position battery level

nd application priority is another important challenge that is still

ompletely open EDCA is the default channel access scheme included

n the IEEE 80211ah amendment and provides some basic traffic dif-

erentiation capabilities at the packet level Besides the different ac-

ess categories (ACs) needing to be renamed and their parameters

Arbitration Inter-Frame Spacing or AIFS CWmin TXOP duration) up-

ated to fit the specific traffic profiles of M2M communications other

echanisms can be implemented A mechanism to assign the down-

ink RAW slots when packets from multiple priority levels are wait-

ng for transmission at the AP is required for example Also uplink

AW slots are currently selected randomly by the STAs that want to

ransmit a packet A mechanism to reserve some slots at each up-

ink RAW for high priority STAs may therefore be an option although

his may severely degrade the performance of the low priority STAs

astly different repetition patterns for different priority TIM groups

ay also allow priority STAs assigned to those priority TIM groups to

ccess the channel more often We expect this challenge will receive

uch attention in the upcoming years because of the heterogeneity

f sensors that will be connected though a single IEEE 80211ah AP

specially in urban scenarios

Similar to what has been discussed for 4G cellular networks used

or M2M communications [126] the development of mechanisms to

void and resolve congestion situations is another open challenge for

EEE 80211ah WLANs with many nodes For instance a mechanism

n which STAs have to wait a random number of DTIM periods before

hey can start a transmission may be implemented This approach

ould effectively distribute the traffic over a longer period of time

n overload conditions but otherwise would unnecessarily increase

he access delay Adaptive and load-aware solutions may play again

n important role to keep the optimal WLAN operation in all possible

ircumstances

Finally since IEEE 80211ah will compete with 4G5G cellular net-

orks and WSNs to provide M2M connectivity in many different ap-

lication domains such as smart cities or e-health comparative per-

ormance studies to determine the strong and weak points of each

echnology are required including also aspects such as the cost of the

evices and the system reliability

Cognitive radio technology for TV White Spaces

This section gives an overview of the IEEE 80211af amendment

ntroducing its most relevant features and open challenges Fig 9

hows the basic components and features considered for WLANs op-

rating in the TV White Spaces (TVWS) which include local spectrum

ensing by the AP and STAs and the use of geolocation data bases

ith information on channel availability

1 The IEEE 80211af amendment

Thanks to the transition from analog TV to digital TV several VHF

nd UHF spectrum channels used for decades for analog TV broad-

asting are now unused or are under-utilised in many geographical

reas [127ndash129] This ldquodigital dividendrdquo of spectrum has suggested

ational regulators such as the FCC in the US and Ofcom in the UK

14 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 9 WLAN operating in TVWS basic elements and functionalities

Fig 10 Hidden terminal problem in TVWS when the secondary network uses only in-

stantaneous channel sensing to decide whether or not a TV channel is occupied When

the STA is not able to detect the TV signal due to the presence of obstacles it may

initiate a transmission and create interference to the primary users

s

c

5

a

i

p

t

o

o

t

e

c

u

l

m

b

a

m

t

e

t

t

t

m

D

c

I

n

b

u

8

c

a

t

i

t

w

f

to discuss how to reuse these channels for unlicensed devicesrsquo com-

munications [130131] These potentially vacant channels in the VHF

and UHF bands are referred to as TVWS [132] and include spectrum

portions like the 470ndash790 MHz in Europe and non-contiguous 54ndash72

76ndash88 MHz 174ndash216 MHz 470ndash698 MHz and 698ndash806 MHz in USA

A snapshot of the TV spectrum occupancy in the city of Barcelona in

2012 is shown in [133]

The attractive characteristics of TVWS (not only for WLANs) in-

clude the ability to penetrate through walls and other obstacles much

more effectively than other widely used spectrum bands such as the

24 and 57 GHz ISM bands [134ndash137] This fact along with a progres-

sive spectrum scarcity in ISM bands has suggested the birth of the

IEEE 80211af amendment published in February 2014 [138] which

provides the IEEE 80211 operational characteristics for TVWS access

of unlicensed White Space Devices (WSD) A good summary of IEEE

80211af can be found in [137] The main advantage of operating IEEE

80211 WLANs in the TVWS comes from an increased coverage range

which can reach up to 1 km in rural areas and open fields [139]

and less energy needed to transmit However this comes at the cost

of an increased interference risk to other WSDs which creates co-

existence problems and demands new PHY and MAC layers to ef-

ficiently support channel access and operations preserving licensed

usersrsquo devices

WLANs operated in the TVWS could cover a number of interest-

ing and emerging use cases and scenarios eg Internet access in ru-

ral or sparsely populated areas Smart Grid sensor aggregation me-

tering and control Internet of Things advanced WLAN operations

and TVWS traffic offloading in indoor environments [140141] Nev-

ertheless the concept could be realised to create or extend com-

mercial or municipality WiFi services offered to citizens with cover-

age at the whole city level realised with reasonably limited network

infrastructures [142143]

IEEE 80211af will use the PHY and MAC layers derived from IEEE

80211ac It will adopt concepts such as the OFDM multi-user beam-

forming contiguous and non-contiguous channel bonding and packet

aggregation Among the mandatory and innovative behavioural and

operational parameters the most notable one is the channel acqui-

sition support realised through remote geolocation-based spectrum

allocation databases which maintain the channelsrsquo availability infor-

mation in any given area and time of day providing upon request the

list of free channels available for use

In the following section we highlight and summarise three main

novelties that IEEE 80211af introduces In Section 511 we explain

how the access infrastructure to the remote spectrum database is de-

signed and what the requirements are we discuss the coexistence is-

sues and methodologies adopted in IEEE 80211af and we also discuss

the novel concept of non-contiguous channel bonding The following

ections also contain an illustration of related works Open research

hallenges will be summarised in Section 512

11 Novel features

This section provides an illustration of novel features introduced

nd discussed in the path to IEEE 80211af and the directions provided

n related works and standardisation initiatives to resolve classical

roblems properly characterised in the new framework of TVWS

echnologies

Channel acquisition spectrum database and channel sensing Much

f the attention in operating in the TVWS is given to the protection

f the primary licensed users in the TVWS spectrum band In general

he primary users were considered as the Digital TV (DTV) broadcast-

rs and receivers In fact receivers would suffer the wasteful effect of

ollisions during the TV broadcast reception in the case of secondary

sersrsquo (SU) transmission interference (Fig 10)

It is a well-known problem that the definition of a coherent wire-

ess channel status (idlebusy) from the viewpoint of distributed pri-

ary users (eg DTV receivers) is made difficult by the difference

etween transmitters and receivers interpretation of signals in a vari-

ble time-space collision domain resulting in hiddenexposed ter-

inals concepts Solutions proposed to reduce the hiddenexposed

erminals included the RTSCTS mechanisms and beaconing How-

ver it is difficult to implement solutions that would effectively de-

ermine the channel status of TVWS channels based on localised dis-

ributed sensing activities performed by multiple secondary users in

he same collision domain On the other hand the use of RTSCTS

echanism would not be effective in this case because most of the

TV primary users (and specifically the receivers) were not originally

onceived to transmit signals or implement the RTSCTS handshake

n conclusion the actions to identify free (and busy) TVWS chan-

els should not be based on primary usersrsquo involvement but should

e almost totally implemented by specifically designed secondary

sersrsquo methodologies

There has been much debate on the methodology that IEEE

0211af secondary devices should adopt to get knowledge of free

hannels at a given time on a given target communication area in

n effort to guarantee verified optimised and reliable TVWS spec-

rum use The three main methodologies discussed include sens-

ng solutions geolocation databases (DB) and beaconing Moreover

wo approaches are possible in general distributed and centralised

ith their well known tradeoffs in terms of effectiveness required in-

rastructure deployment and coordination overhead The distributed

B Bellalta et al Computer Communications 75 (2016) 1ndash25 15

m

b

s

c

t

u

a

t

o

l

t

w

t

t

a

o

c

c

a

(

t

o

t

t

D

t

t

o

t

f

a

t

c

i

a

c

d

t

s

t

b

I

l

(

e

a

i

R

p

e

t

c

c

i

b

i

i

t

b

t

t

t

W

a

p

c

i

t

h

p

t

n

a

t

e

t

t

t

p

t

a

r

8

w

s

w

[

i

S

o

a

t

f

i

8

i

r

a

a

i

t

c

(

I

s

a

e

t

t

c

[

i

c

i

g

s

n

s

t

i

o

o

s

c

D

o

ethodology that is receiving more attention in the literature is

ased on distributed channel sensing solutions [127144145] These

olutions include the observation of spectrum use and possible

ooperative aggregation of spectrum sensing information to increase

he accuracy of detection and reduce the vulnerability of primary

sersrsquo transmissions The spectrum sensing capability is required in

ll the secondary devices with a minus114 dBm sensitivity Dynamic

ransmission power control must be provided and the upper limit

n emissions is 100 mW EIRP (20 dBm) for portable devices (further

imited in the case of adjacent channel use to reduce out-of-band in-

erference) However database (DB) coordination of TVWS combined

ith spectrum sensing is considered the most promising and effec-

ive technique compared with spectrum sensing alone The DB spec-

rum information has more chances to be effective (ie identifying

ll free channels opportunities) and reliable (ie not prone to attacks

r erroneous interpretation of channels status) In fact regulators de-

ided to push for the conservative solution of a remotely accessible

entralised spectrum DB that maintains the information on the avail-

bility of all TVWS channels at any given point in time and location

with a target accuracy around 50 m) under their responsibility For

his reason the spectrum databases have become a mandatory part

f many spectrum sharing systems and the dominant technical solu-

ion to support TV white spaces

The IEEE 80211af standard adheres to this vision and also has

he role of a common regulatory framework for different spectrum

B implementations defining a generalised coordination architec-

ure protocols and interfaces for spectrum queries and local spec-

rum information management The protocol explicitly considers out

f scope the communication protocol and technology adopted in

he background leaving the freedom to those players responsible

or the deployment to select any suitable Internet-based and local

ccess network technology Secondary devices that need to access

he TVWS spectrum should get information on which channels they

an effectively use and which parameters to adopt without impact-

ng primary usersrsquo transmissions (and receptions) by querying the

vailable spectrum DB The query must contain the transmission

haracteristics and accurate geographical position of the secondary

evice which can be determined by means of a geo-positioning sys-

em (GPS) regularly updated in the case of movement (or manually

et for static devices) The IEEE 80211af reference system architec-

ure for DB spectrum access is composed of multiple Geolocation-

ased Spectrum information Databases (GDBs) entities connected via

nternet to Registered Location Secure Servers (RLSS) RLSSs work as

ocal proxies of the GDB for a localised group of Basic Service Sets

BSS) RLSS are connected via a secure protocol architecture with the

quivalent of multiple Access Points (AP) in different BSS realising

trusted DB infrastructure APs locally coordinate the exchange of

nformation and channel access management between the GDB (via

LSS) and the secondary usersrsquo end stations (STA)

Mode I devices are those under the control of a device that em-

loys geo-location database access while Mode II devices are those

mploying geo-location database access by themselves The informa-

ion exchange provided between the GDB and a secondary user STA

an be provided in both an open-loop (eg adopted by FCC) and a

losed-loop (eg adopted by ETSI) implementation In an open loop

mplementation daily spectrum availability information is provided

y the GDB and no feedback on the spectrum information received

s provided by STAs thus the system approach for spectrum access

s much more conservative leading to low channelsrsquo utilisation po-

ential In a closed loop implementation the STA can provide feed-

ack to the GDB and there are more communication overheads due

o the high granularity of updates however the system is more effec-

ive and reliable in the exploitation of TVWS on behalf of STAs The

ypical information provided by the GDB includes (1) the updated

hite Space Maps (WSM) of frequencies allowed for secondary use

t the timespace of the querying STA and (2) the device-dependent

ower limitations for transmission (in general conservative and ac-

urate enough to avoid relevant interference effects on primary users

dentified in the area)

The basic IEEE 80211af mechanisms regulating the communica-

ion between STA and GDB can be found in [137138]

In parallel with IEEE 80211af standards such as the IEEE 19006

ave been created that consider interfaces and data structures sup-

orting spectrum sensing information exchange applicable to spec-

rum sensing (and particularly distributed spectrum sensing) sce-

arios In particular in October 2014 the IEEE 19006 WG initiated

project for a new standard called IEEE 19006b [146] concerning

he use of spectrum sensing information to support and optimise the

ffectiveness reliability and robustness of spectrum database solu-

ions The aim is to enhance the performance and capabilities of spec-

rum databases through the use of spectrum sensing information

Co-existence Support for co-existence mechanisms so that mul-

iple technologies can effectively utilise the TVWS spectrum is im-

ortant Self co-existence between network devices of a common

echnology (eg deployed by different operators in the same area)

nd co-existence among different technologies are relevant topic of

esearch for Cognitive Radio (CR) systems and specifically for IEEE

0211af Many solutions appeared on the research scene but no one

as so far finalised as the target solution for IEEE 80211af Specific

tandardisation has been started to regulate coexistence between

ireless standards of unlicensed devices including the IEEE 802191

147] The purpose of the IEEE 802191 standard is to enable the fam-

ly of IEEE 802 Wireless Standards to most effectively use TV White

pace by providing standard coexistence methods among dissimilar

r independently operated TVWS devices Early examples of gener-

lised coexistence mechanisms included Dynamic Frequency Selec-

ion (DFS) Transmission Power Control (TPC) listen before talk (eg

or contention based IEEE 80211 80215) time division multiplex-

ng (also among different techniques such as the IEEE 80216 80220

0222) and Message-based Spectrum Contention (that is beacon-

ng messages that carry coexistence information) Opportune met-

ics must be defined to assess the measurable coexistence achieved

mong different technologies as an example the hidden node prob-

bility for a target scenario or the estimate of percentage variation

n normalised network throughput and latency (before and during

he SU transmissions) On the other hand a centralised coexistence

ontrol mechanism could be effectively realised by a central manager

or coexistence-DB like the GDB) in critical scenarios To this end

EEE 19004 [148] (a standard for heterogeneous networks in dynamic

pectrum context part of IEEE Standards Coordinating Committee 41)

ims to standardise the overall system architecture and information

xchange between the network and mobile devices which will allow

hese elements to optimally choose from available radio resources

There are three possible classification of co-existence archi-

ectures and inter-network coordination channels for CR systems

entralised coordinated and autonomous co-existence mechanisms

149] In centralised co-existence schemes the co-existence is admin-

stered by a central entity (eg like in IEEE 802191) This solution

ould be applied in both homogeneous IEEE 80211af systems (eg

ncarnated by the centralised spectrum DB entity) and in hetero-

eneous inter-networks without requiring modifications to existing

tandards under the assumption that all the involved co-existing

etwork entities would adhere to the same centralised coordination

cheme and inter-network coordination protocol On the other hand

his assumption could be hard to satisfy in some practical scenar-

os The centralised schemes are effective in providing minimisation

f inter-network interference based on the availability and quality

f centralised co-existence information In coordinated co-existence

chemes the centralised entity could be present but not taking de-

isions in general The central entity could provide a co-existence

B with required information which can be used via in-band and

ut-band signalling for co-existence decisions taken by the cluster

16 B Bellalta et al Computer Communications 75 (2016) 1ndash25

c

e

n

c

o

s

i

d

a

S

8

n

a

b

o

b

i

t

a

8

b

[

t

i

n

b

5

t

a

c

b

s

a

c

n

s

p

i

c

l

r

c

f

a

v

t

a

e

c

t

Q

a

b

s

i

c

u

n

c

i

heads of many co-existing inter-networks for implementing a proper

common coordination protocol (eg out-band busy tone signalling)

These are often hybrid solutions realising a compromise between ef-

fectiveness and suitability In autonomous co-existence schemes all the

decisions and coordination are implemented in a distributed way

by the involved network entities In these schemes many policies

can be adopted to realise a sufficient level of co-existence under a

best effort approach Solutions that could be implemented include

the use of busy tones beaconing and other signalling protocols in

dedicated control channels dynamic distributed frequency selection

schemes listen before transmit token-based and dynamic reserva-

tion schemes All the above mentioned solutions have goods and

bads and a general illustration can be found in [149] Other proac-

tive co-existence techniques try to early detect and recovermitigate

co-existence issues with relaxed inter-network coordination eg re-

alised via spectrum sensing and interference avoidancesuppression

techniques

It must be clear that different aims exist regarding the co-

existence for primary TVWS users protection against secondary

users and for secondary devices mutual interference avoidance

Many co-existence problems have been analysed in the literature

for TVWS technologies in particular between the primary (licensed)

and secondary (unlicensed) devices The coexistence management in

TVWS is complex because primary users of DVB transmissions are in-

tended to be the pure receivers rather than the broadcasters In other

words potential protected users could be everywhere and the hid-

den terminal problem could arise for these systems The co-existence

problem is also complicated by many factors that cause asymme-

try and dynamicity in the TVWS such as the mobility variable den-

sity power asymmetry and heterogeneous MACPHY layers TVWS

co-existence problems originate between the IEEE 80211af technolo-

gies and the IEEE 80222 technologies for wireless regional area net-

works (WRAN) Results of analysis of TVWS usage in Europe show

that white spaces are typically present and fragmented They are typ-

ically more abundant in rural areas where larger contiguous blocks

of unused channels are available due to broadcast network planning

giving priorities linked to population density However the exploita-

tion of TVWS in urban areas is possible eg some recent research

was realised under the assumption of exploiting shadowing effects

created in the communication environment (eg by buildings obsta-

cles) in favour of frequency reuse [150]

Since different standards for opportunistic communication in the

TV White Spaces have now been published such as IEEE 80222 [151]

IEEE 802154m [152] Weightless [153] and of course IEEE 80211af

improved methods to guarantee the coexistence of different devices

operating on several protocols in the same bands must be deployed

We highlight here some work eg [150154ndash157] that focuses on

the coexistence between different technologies operating in the same

bands In particular [155] studies the performance degradation of

IEEE 80222 when an IEEE 80211af network is operated in the same

area The problem becomes even worse when the IEEE 80211af net-

work is located near 80222 user equipment causing both networks

to perceive a strong interference from each other One proposed solu-

tion is the Coexistence Beacon Protocol [158] studied for 80222 net-

works which foresees the exchange of a periodic beacon to identify

the neighbouring and possibly interfering networks

For what concerns the co-existence problems research has ad-

dressed other important related aspects of new WLANs on TV White

Spaces For instance the higher coverage range makes them attrac-

tive for applications in the smart grid and also for M2M on TVWS

[134150159] Here the increased range compared to standard tech-

nologies in the ISM bands can have beneficial effects for indoor

mobile devices [150] and M2M [159] Specifically indoor devices

communicating on TV White Spaces offer better propagation char-

acteristics and penetration through obstacles making it easier to

realise scalable home connectivity [135] although interference and

o-existence problems could be exacerbated and must be resolved

g see [139] However studies in the literature show how the sig-

al coming from an indoor transmitter on TV White Spaces remains

onstrained inside the house making it difficult to deploy indoor-to-

utdoor networks On the other hand this shadowing limit can be

een as an opportunity lowering the interference outside the build-

ng in which the TVWS network is locally operated [134150] Un-

er this approach HDTV streaming has recently been considered as

possible use case of WLAN networks operating in the TV White

pace [160]

Non-contiguous channel bonding A relevant novel feature of IEEE

0211af is the potential for contiguous and non-contiguous chan-

el bonding which permits aggregating basic channels (also non-

djacent ones) and leveraging the possible large frequency spread

etween multiple available channels Due to the rather static nature

f primary DTV transmissions where a busy channel is unlikely to

ecome free in the near future and state changes are coarse-grained

n general it is crucial to exploit the time-locality effect and exploit

he maximum physical channel availability that could be aggregated

t any given location With the methodology inherited from IEEE

0211ac IEEE 80211af is capable of bonding together two up to four

asic channels grouped in up to two different non-contiguous chunks

137] The spectrum bandwidth of a DVB-T basic channel can be ei-

her 6 7 or 8 MHz depending on the country in which the service

s operated As an example this creates a 144 to 168 OFDM chan-

elsrsquo bandwidth potential when up to four 6ndash7ndash8 MHz channels are

onded [137]

12 Open challenges

The deployment of WLANs in the TVWS at their maximum poten-

ial still requires the resolution of open problems Generally speaking

s mentioned in previous sections the challenges for such networks

an be divided into three main categories (1) spectrum sharing

etween opportunistic (secondary) and heterogeneous devices (2)

pectrum sensingmanagement and maximum exploitation of avail-

ble spectrum (potentially enabling the spectrum-on-demand con-

ept) and (3) co-existence and interference mitigation to the primary

etwork (primary user protection) and secondary networks [154]

The overall aim of spectrum sharing techniques is to maximise

eparation to avoid overlapping or contiguous channels operations

rovided that sufficient channels are available in the TVWS spectrum

n a given spacetime scenario [154] In general important research

ontributions have to be realised for the design of proper spectrum al-

ocation methodologies and techniques satisfying multi-factorial QoS

equirements at the system-level and the user-level Novel ideas in-

lude the spectrum sharing and spectrum sensing techniques being

urther divided into cooperative and non-cooperative

Non-cooperative solutions attempt to realise spectrum sharing

nd sensing on a local basis without direct cooperation between de-

ices Examples include the ldquolisten-before-transmitrdquo approach and

ransmission power control to limit the interference and spectrum

llocation policies that are based on local node feedback only Coop-

rative solutions rely on the existence of a common communication

hannel through which devices can tentatively agree on the spec-

rum allocation that provides the desired spectrum separation and

oS Major issues include the definition of a methodology to identify

common channel (or multiple channels mutually shared in space

y pairs of heterogeneous and distributed devices) and to efficiently

hare common information For the access to the spectrum DBs the

dentification of proper common access channels is still under dis-

ussion out-of-TVWS-band cellular communications can be widely

sed given their high coverage also indoor

Trying to avoid the use of complex and resource-hungry coordi-

ation schemes and dedicated control channels rendezvous proto-

ols play a significant role both for co-existence and for establish-

ng dynamic communication channels based on spectrum sharing

B Bellalta et al Computer Communications 75 (2016) 1ndash25 17

R

n

c

t

l

o

m

e

n

i

n

w

n

a

u

s

p

b

s

m

s

a

d

t

b

a

a

m

a

t

u

8

n

t

d

d

m

f

t

n

d

n

e

r

a

h

b

f

a

c

t

u

i

f

t

d

t

s

i

f

I

l

i

a

n

c

d

I

n

o

t

i

t

e

w

t

c

m

t

a

i

t

q

b

c

a

i

l

i

a

v

t

a

a

c

t

i

t

b

t

t

a

D

p

c

s

o

q

r

p

a

g

t

e

u

d

t

m

n

t

s

f

i

1

8

f

T

endezvous is considered a fundamental problem in generalised CR

etworks at the basis of many communication processes eg in-

luding neighbour discovery routing and broadcast Rendezvous pro-

ocols attempt to establish a new link for communication on a se-

ected and agreed frequency band (channel) identified within a set

f available resources (channels) thus creating the basis for com-

unication between two or more SUs This problem is often exac-

rbated by the high (and unknown) numbers of contending SUs high

umber of available channels and their ambiguous boundaries and

dentifiers and tough co-existence requirements satisfaction in cog-

itive radio systems [161] Preliminary approaches for rendezvous

ere based on the existence of a dedicated Common Control Chan-

el (CCC) agreed among all the SUs or the existence of a centralised

gent (decision maker) assigning channels to any pairs of requesting

sers This approach could be considered in the case of IEEE 80211af

olutions based on the spectrum DB implementation however the

roblem is complicated by the fact that the common channel must

e identified and made available for all the requesting SUs at the

ame time even when these are spread over multiple collision do-

ains in space On the other hand the scalability and reliability is-

ues which are caused by the congestion of CCCs and the vulner-

bility risks of centralised DBs motivate the research of alternative

istributed solutions called blind rendezvous protocols [162] Dis-

ributed solutions have been mainly based on beaconing andor slot-

ased channel hopping algorithms exploiting randomisation sensing

nd discoverysynchronisation messages spread in slotted time over

set of candidate available channels [161] These protocols aim to

inimise the convergence time to maximise the use of resources

nd to avoid collisions among SUs by identifying and agreeing on

he use of a given common channel among N non-overlapping and

nambiguously labelled channels The main complication in IEEE

0211af systems is given by the wide availability of many different

on-contiguous spectrum bands which cannot be uniformly quan-

ised in single channels and unambiguously labelled in space in or-

er to realise a common reference domain for all the IEEE 80211af

evices In fact the knowledge of the number of channels their com-

on labels the unique IDs of SUs and the number of SUs contending

or the channel assignment are the key factors determining the effec-

iveness and good properties of blind rendezvous protocols Recently

ew classes of rendezvous protocols have been proposed which are

istributed blind and oblivious Oblivious means that available chan-

els may be identified with different labels on behalf of the differ-

nt involved SUs A number of these distributed blind and oblivious

endezvous protocols have been proposed and have been referenced

nd analysed in [162] To conclude solutions for dynamic spectrum

andoff issues (eg in case PUs suddenly appear in a region) must

e resolved in advance to minimise the latency of re-establishing a

unctional communication channel eg by implementing a proactive

lternative rendezvous channel selection in background to ongoing

ommunication processes [163]

Another challenge is to provide cooperative techniques in a dis-

ributed vs centralised implementation by analysing their convergence

nder variable conditions overheads and tradeoffs In some cases as

n vehicular networks the exploitation of scenario characteristics and

actors such as the constrained mobility could help to realise effec-

ive dissemination of sensing information over an extended and pre-

ictable horizon under a cooperative approach [164ndash166] Coopera-

ion could be helpful in the realisation of the cooperative spectrum

ensing in combination with the Geolocation DB approach Another

nteresting direction of research is the investigation of mutual ef-

ects of coexisting cooperation-based vs non-cooperative techniques

n cooperation-based devices the exchange of information could al-

ow a quick convergence to a solution and avoidance of further

nterference [145] Certainly cooperative techniques have desirable

dvantages over non-cooperative ones However tight time synchro-

isation and high coordination overheads are required The common

hannel for cooperation purposes could become a bottleneck and re-

uce the potential advantages of dynamic spectrum allocation The

EEE 80219 foresees the presence of a shared common control chan-

el (CCC) to which devices need to tune in order to gather the state

f the network Solutions such as [141] propose the use of TVWS for

he CCC

A promising direction for research is the adoption of cluster-

ng schemes for secondary devices associated with a hybrid dis-

ributedcooperative vs centralisednon-cooperative approach in an

ffort to reduce overheads while maximising advantages In this

ay secondary nodes could implement cooperation for sensing func-

ions using properly identified common channels and delegate the

entralised decisions and spectrum-DB management to well instru-

ented cluster-leader nodes This concept is incarnated to some ex-

ent by RLSS in IEEE 80211af More recently there have been propos-

ls to build distributed databases that refresh their contents period-

cally by querying the remote spectrum database [167] Here Mas-

er devices (according to Ofcom terminology) periodically cache the

uery replies from the spectrum database in order to reply to and

roadcast the spectrum availability to neighbour Slave devices

Another challenging direction for research is the definition of ac-

urate models and efficient simulation tools enabling dynamic spectrum

nalysis for both rural and urban areas Recent developments of dig-

tal maps simulation tools and propagation models theoretically al-

ow improving the capability of predicting radio propagation effects

n complex scenarios with an acceptable computation time This en-

bling technology could be used in parallel to spectrum DBs to pro-

ide more accurate forecasting of dynamic frequency allocation in

imespace scenarios Another issue to be taken into account is the

ccuracy of the remote spectrum database regarding the channel

vailability estimation Several works have already shown the inac-

uracy of propagation models due to the complexity of parameters

o be considered and differences between modelled and real scenar-

os [132134168169] Better propagation models will make it possible

o more efficiently estimate the interference between devices and

uild Radio Environment Maps (REM) or White Space Maps (WSM)

hat will possibly lead to a more accurate sharing of the radio spec-

rum [144] To this end current research could focus on how to lever-

ge simulation models in order to build more accurate frequency

Bs

In general the adoption of remote spectrum DBs methodology

ushed by national regulators has made the spectrum sensing pro-

ess optional and conservatively realised (eg power control for

ensing-only devices is limited to below 17 dBm EIRP) Thus much

f the research has focused on techniques and solutions to successfully

uery the remote database also in challenging environments such as

ural areas and indoor However since the queries must provide the

osition of the mobile device the accuracy of positioning for mobile

nd handheld devices still needs more appropriate solutions Trian-

ulation can be used to estimate the position of mobile devices with

he accuracy needed by the current regulations outdoor (50 m) How-

ver a low GPS accuracy is possible indoors [170] Certified man-

al positioning can be provided by technicians for static and indoor

evices although limiting mobility Alternatives to costly infrastruc-

ures for positioning (eg based on short range beaconing devices)

ust be deployed in indoor scenarios

Finally regarding spectrum sensing and interference mitigation

ew challenges come from the heterogeneity and different charac-

eristics that networks operating in the TV White Spaces can have

uch as different transmitting power (4 W for static devices 100 mW

or mobile and portable devices and 40 mW for communication

n channels adjacent to occupied ones) different bandwidths (5

0 15 and 20 MHz are currently foreseen for IEEE 80211af IEEE

02154m can have narrower bandwidths or wider ones) and dif-

erent medium access schemes (protocols that use either CSMA or

DMA are required to co-exist in the TV White Spaces) With all this

18 B Bellalta et al Computer Communications 75 (2016) 1ndash25

d

a

T

i

a

c

t

l

b

o

t

a

d

p

a

r

p

n

a

m

t

p

p

d

d

o

t

i

D

w

p

i

a

a

d

b

fi

d

o

t

l

b

t

b

a

i

e

o

o

c

t

t

w

A

c

M

s

c

l

O

w

g

v

t

in place the problem of determining whether another secondary de-

vice is currently transmitting in the same channel is still an open

issue

Spectrum sensing is a very hot topic in the literature with many

proposals that leverage the cooperation between devices to raise the

accuracy of the detection process eg [144171] The current litera-

ture also offers room for proposals that suggest exploiting sensing to

build more accurate spectrum DBs [164172173] Lastly regarding the

interference mitigation to the primary network regulators provide a

conservative approach based on a remote spectrum database to avoid

high risk for the primary users However in particular for the indoor

scenario the concept of secondary usersrsquo transmission in TV Grey

Spaces has been proposed TV Grey Space identifies busy channels

that are formally used by primary users in the area of interest (eg

at the rooftop level of a building for DVB-T) but which could be con-

sidered usable indoors (eg at the basement level) without causing

effective interference to the primary users [134150174] This concept

is based on the generalised spectrum underlay paradigm [175] The

remaining research challenge is to provide TV Grey space access guar-

anteeing a sufficient amount of protection to the primary user while

providing a satisfactory QoS to opportunistic devices eg through

the use of customised and validated propagation models for indoor

TVWS networks and feedback mechanisms to constantly monitor the

interference generated to the primary network

6 Emerging new trends and technologies

In this last section we review three emerging trends that in our

opinion will have a large influence in the conception of future WLANs

as they change the way WLAN protocols and functionalities are de-

veloped implemented tested and integrated with other wireless

networks

61 Programmable Wireless LANs

Especially in the enterprise environment WLAN deployments

need to support a wide range of functionalities and services This is

intrinsically difficult because of the large number of APs that must be

managed which calls for scalable solutions Typical services include

channel assignment load balancing among APs authentication au-

thorisation and accounting (AAA) policy management support for

client mobility and interference coordination Another problem is the

fact that WLAN clients autonomously take several decisions such as

which APs to associate with when to hand-over etc Therefore sup-

porting roaming clients requires the management of a large number

of association states across several APs which is a challenge if sup-

port for real-time hand-over is desired Typically such management

schemes are centralised and most of them are proprietary such as

WLAN controller solutions from Aruba [176] or Cisco [177] although

the IEEE 80211u amendment has been released to allow mobile users

to seamless roam between WiFi networks with automatic authenti-

cation and handoff [178] For example Dyson [179] enables STAs to

send information such as radio channel conditions to a centralised

controller based on a custom API (eg based on Python) As the con-

troller has a centralised view of the network it can enforce a rich

set of policies to control the network also using historical informa-

tion A demo system has been implemented along with applications

such as airtime reservations for specific clients or optimised hand-

offs However Dyson requires STAs to be modified in order to use

those new services offered by the centralised controller TRANTOR

[180] is another example of a centralised management system that

requires changes of clients in order for the infrastructure to gather

information from them in terms of eg interference measurements

In addition control commands enable the infrastructure to exercise

control such as modifying the transmit power or influencing the as-

sociation procedure Clients still use standard CSMA MAC layer for

ata transmission In contrast to DenseAP [181] or DIRAC [182] which

re also based on a centralised controller but do not modify clients

RANTOR exercises much more control on clients by the use of a ded-

cated API which enables a significantly higher gain in coordination

nd thus capacity CENTAUR [183] is another example of a centralised

ontroller that centrally schedules hidden and exposed terminals in

he downlink while using standard CSMA MAC for uplink traffic and

egacy downlink traffic To achieve good performance it uses a fixed

ack-off packet staggering techniques and a hybrid data path that

nly schedules downlink transmissions towards hidden and exposed

erminals centrally All other traffic is sent using standard DCF with

standard back-off procedure In contrast to previous approaches it

oes not require any modifications to the clients but does require data

lane centralisation and it remains questionable how scalable such

solution is for large WLAN deployments with high PHY layer data

ates beyond 100 Mbps However new approaches for fast data plane

rocessing that are currently explored in the virtualisation commu-

ity which move all the packet processing into user space may be

n option to significantly increase speed Designing such manage-

ent systems raises several interesting research questions (a) Cen-

ralised infrastructures may incur high latency but are more simple to

rogram and maintain while designing a completely distributed ap-

roach that operates close to WLAN devices may result in significant

istributed coordination and consistency problems (b) what level of

istributed control is required in order to support the flexibility needs

f future WiFi based networks supporting a high level of mobility

Traditionally WLAN APs are built on proprietary operating sys-

ems that are tightly coupled with the hardware This design makes

t hard to create new applications on top of such networking devices

espite the fact that protocols and mechanisms are available that

ould greatly improve the utility of existing networks those new

rotocols are not deployed because the closed system design makes

t very difficult to extend their functionality An important aspect for

ll such new approaches is therefore how to provide open interfaces

nd open source to speed up innovation As an example Linux-based

evices are completely open source but in order to increase flexi-

ility much more work is needed in the area of open drivers and

rmware In fact the Atheros based ath9k driver has been the main

river of innovation in the open source community because it is the

nly driver that interworks with open firmware In general solutions

hat modify the OS driver focus mainly on programmable network

evel solutions for WLANs such as channel switching or handovers

etween APs While the ath9k driver is an excellent example in how

o speed up innovation it is still very cumbersome to support flexi-

le MAC engine reprogrammability with ath9k In contrast MAClets

nd Wireless MAC processors (WMP) [184] allow a much more flex-

ble reprogramming of MAC functionalities A MAC processor is an

ntity able to execute general MAC commands that specify the MAC

perations through a software-defined state machine The behaviour

f the MAC protocol can therefore be updated at run-time by simply

hanging the sequence in which those commands are executed (ie

he MAClets) As a proof-of-concept the authors of [184] implement

he proposed MAC processor solution in a commodity WLAN hard-

are card extending the basic DCF in three directions piggy-backing

CKs a pseudo TDMA and the use of multiple channels In [185] a

ontrol framework for this WMP system is also proposed to support

AClet code mobility ie for moving loading and activating MAC

oftware programs embedded into ordinary data packets (akin to the

apsule model of traditional active networks)

The difficulty in re-programming networking hardware has also

ed to the concept of Software Defined Networks (SDNs) based on the

penFlow protocol [186] The main idea of SDNs is to extend net-

orking devices with standardised APIs that allow third-party pro-

rammers to flexibly control the data path In addition SDNs pro-

ide higher level abstractions to network designers and programmers

hrough the use of a centralised control plane offered by a network

B Bellalta et al Computer Communications 75 (2016) 1ndash25 19

Fig 11 The CLOUDMAC architecture VAP stands for Virtual Access Point and WT

stands for Wireless Termination

c

a

c

t

m

W

n

t

t

c

p

t

u

h

a

i

e

s

o

c

a

a

f

T

r

s

a

s

p

o

M

C

i

T

t

fl

A

a

a

i

c

t

s

t

e

I

t

w

H

e

b

I

l

d

o

6

r

W

s

l

a

e

e

a

s

b

c

t

(

S

p

c

F

B

t

p

T

e

t

s

u

m

i

e

o

a

A

p

d

t

t

n

p

i

t

t

f

i

ontroller such as NOX [187] which allows reuse of components such

s topology discovery or network access control for different appli-

ations The SDN concept has recently been applied to WLAN archi-

ecture in order to enable fine grained control over mobility manage-

ent and data forwarding focusing on a programmable enterprise

LAN architecture An important part of the SDN architecture is the

etwork controller which provides a centralised view of (parts of)

he SDN enabled network and uses the OpenFlow protocol to install

he forwarding rules on SDN-enabled devices (routers switches ac-

ess points cellular base stations etc)

ODIN [188] is designed to support programmability in enter-

rise WLAN architecture by separating the association state from

he physical AP They implement Light Virtual Access Points (LVAP)

sing a Split-MAC approach where the infrastructure controls the

andover procedure among different WLAN APs By managing

ssociations through SDN controllers ODIN enables proactive mobil-

ty management and load balancing within the SDN enabled WLAN

nterprise network without the need for changes in the client WLAN

tack or IEEE 80211 MAC layer While ODIN requires agents to reside

n the APs that communicate with the Odin Master within the SDN

ontroller CLOUDMAC [189] is based purely on the concept of SDN

nd virtual APs Similar to ODIN CLOUDMAC implements a Split-MAC

pproach but in addition enables the processing of WLAN MAC layer

rames within a co-located Cloud using so-called Virtual APs (VAPs)

he physical APs in CLOUDMAC are lightweight WLAN APs that are

esponsible only for sending out IEEE 80211 based MAC layer ACKs to

tandard WLAN clients and tunnel WLAN MAC layer frames through

n SDN-enabled enterprise WLAN towards the VAPs As association

tates are kept in the VAPs fast mobility is supported using sim-

le OpenFlow forwarding rules Because of the additional processing

f IEEE 80211 WLAN MAC frames in the co-located Cloud CLOUD-

AC has higher latency than a standard WLAN deployment However

LOUDMAC offers a Webservice based API to third party applications

n order to program the enterprise WLAN and enable new services

he architecture of CLOUDMAC is depicted in Fig 11 It has been ex-

ended in [190] to support QoS management and in [191] to support

exible MAC management frame prioritisation based on IEEE 80211

system based on OpenFlow has been recently proposed in [192] to

llow the station to be associated with multiple APs simultaneously

nd to switch between APs with low overhead

An important aspect to consider in order to enable programmabil-

ty is backwards compatibility Several approaches (eg [180]) require

lients to be modified to utilise the features provided This is difficult

o do in practice because it requires changes in the operating system

oftware of all clients If not all clients can utilise those APIs it is ques-

ionable how usable the new architecture will be and how much ben-

fit in terms of aggregate performance such architecture will allow

n contrast the SDN based approaches are interesting in the sense

hat they do not require changes in the client WLAN stack and work

ith the standard IEEE 80211 MAC layer deployed within the clients

owever it remains questionable how scalable such solutions are For

xample an interesting open research topic is to evaluate the scala-

ility of approaches such as CLOUDMAC [189] In addition processing

EEE 80211 MAC frames within co-located private Cloud requires low

atency support from local Cloud solutions (such as OpenStack) in or-

er to reduce the IEEE 80211 MAC processing time which is an area

f active research

2 Prototyping and testing IEEE 80211 enhancements

Most of the new proposals for next-generation WLANs are cur-

ently only evaluated using mathematical analysis and simulation

hile both analysis and simulation are necessary to characterise and

tudy those enhancements in the initial design phase or to consider

arge-scale scenarios it is difficult to consider all practical aspects of

real-world scenario This can sometimes cause significant differ-

nces between what simulations and real experiments show How-

ver real experiments are challenging because of the high complexity

nd costs of building the new hardware and software for each specific

olution to test

To mitigate implementation complexity there are several flexi-

le hardware platforms where low-level MAC mechanisms can be

ompletely implemented in software Examples based on FPGAs are

he USRP (Universal Software Radio Peripheral) [193] and the WARP

Wireless Open-Access Research Platform) [194] As an alternative

ORA (Microsoft Research Software Radio) [195] works on general

urpose computers by taking some advantage from modern multiple

ore systems

A different approach is provided by OpenFWWF [196] Open-

WWF provides an open CSMACA firmware for specific models of

roadcom chipsets so the resulting firmware can be uploaded and

ested in real commercial hardware OpenFWWF implements a sim-

le State Machine (SM) for controlling the hardware in real time

he SM evolution is driven by a main loop that reacts to events by

xecuting specific handlers When a packet originally prepared by

he Linux kernel is ready in the NIC memory handler Packet_Ready

ets up the radio hardware according to the packet metadata sched-

les the transmission and jumps back to the main loop The Trans-

ission Engine (TXE) then takes care of accessing the channel ie

t decrements the back-off counter according to the DCF rules and

ventually starts the actual transmission This triggers the execution

f the TX_frame_now event that prepares the ACK time-out clock

nd finalises the MAC header If the ACK-frame is received or if the

CK time-out expires and the maximum number of attempts for this

acket is reached handler Update_Params resets the contention win-

ow to its minimum value or otherwise doubles it Finally it loads

he back-off counter with a fresh value

In the following we will review a list of selected papers that use

he OpenFWWF firmware and the WARP or USRP platforms to test

ew proposals for WLANs in real scenarios as examples A first im-

lementation of the groupcast mechanisms defined in IEEE 80211aa

s presented in [197] The authors modify the OpenFWWF firmware

o include those functionalities and evaluate them using a 30 STAs

estbed A second example of the use of OpenFWWF is the collision-

ree MAC protocol presented in [198] In [198] the authors design and

mplement in OpenFWWF a MAC protocol that is able to achieve a

20 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 12 LTE to WIFI offloading

a

n

c

t

[

d

a

W

v

t

t

o

b

f

S

g

l

i

I

s

I

b

n

s

m

a

p

d

c

N

o

v

t

t

n

l

i

a

t

a

collision-free operation by waiting a deterministic time after success-

ful transmissions before to start a new transmission This work shows

the experimental performance of a collision-free MAC (CF-MAC) pro-

tocol for WLANs using commercial hardware The results show that

the proposed CF-MAC protocol leads to a better distribution of the

available bandwidth among users a higher throughput and lower

losses than DCF

In [199] the authors use the WARP platform to test a variant of the

DCF ndash called IEEE 80211ec ndash that substitutes control packets such as

the RTS CTS and ACK with short detectable sequences Since those se-

quences are shorter than the control packets and can be detected cor-

rectly even at lower SNRs values a significant gain in performance is

achieved MIDAS (Multiple-Input Distributed Antenna Systems) [200]

implementation using the WARP platform shows the benefits of dis-

tributing the antennas over the area to cover instead of co-locating

them at the AP in terms of capacity when MU-MIMO is employed

The authors also propose a new MAC protocol to benefit from the

spatial reuse that the DAS allows taking as a basis the IEEE 80211ac

Experimental results show that their proposal is able to achieve up

to 200 gains versus the traditional approach where all antennas are

co-located at the AP

The USRP platform has recently been proposed to develop a first

implementation of IEEE 80211ah in order to obtain a preliminary ex-

perimental performance assessment of such a technology because

no commercial off-the-shelf hardware is yet available as the amend-

ment is still in progress [201] The USRP platform is also used in

[202] to evaluate TIMO (Technology Independent Multi-Output) a so-

lution to deal with high-power non-IEEE 80211 interferers in ISM

bands

To conclude this section it is worth mentioning here also to

MAClets and WMP as given such flexibility to implement and dis-

tribute new MAC protocols and other IEEE 80211 functionalities we

believe that if a use-friendly implementation of such a MAC proces-

sors framework is provided it would significantly contribute to the

development and testing of new MAC enhancements by the research

community

63 CellularWLAN interworking

Public hotspots that offer Internet access over a WLAN using IEEE

80211 technology are now nearly ubiquitous It is forecasted that the

cumulative installed base of WiFi hotspots worldwide will amount

to 551 million by 2018 excluding private hotspots (eg WiFi access

points deployed at home) [203] The sharp increase in the availabil-

ity of public WiFi was initially perceived by mobile cellular operators

as a threat due to the additional competition from wireline Internet

service providers or emerging crowdsourced WiFi networks such as

FON5 However as cellular operators are fighting to cope with the ex-

plosion of mobile data traffic created by the rising use of multimedia

content traffic over mobile devices [15] they are also starting to use

WLANs based on the IEEE 80211 technology to offload data from their

core and access networks In general mobile data offloading refers to

the use of complementary network technologies (in licensed or un-

licensed spectrum) for delivering data originally targeted to cellular

networks Intuitively the simplest type of offloading consists of ex-

ploiting connectivity to existing co-located WiFi networks and trans-

ferring data without any delay (Fig 12) Thus this offloading tech-

nique is known as on-the-spot offloading As a consequence of this

new trend the seamless integration of cellular (eg 3GLTE) and WiFi

technologies has attracted significant research interest in recent years

(see [204] for a survey) a few solutions have already been standard-

ised [205206] and roaming between cellular and WiFi is becoming

increasingly transparent to end users CellularWLAN interworking is

5 httpscorpfoncomen

fl

t

f

lso fostered by the support in the evolving 4G standards of heteroge-

eous network deployments (HetNets) in which the existing macro

ells are complemented with a number of small low-power base sta-

ions with the goal of increasing capacity in highly congested areas

207208] It is envisaged that small cells will be based on 4G stan-

ards (eg pico and femto cells) as well as IEEE 80211 technologies

nd multimode base stations that work simultaneously with LTE and

iFi are already entering the market

It is important to point out that LTE standards already support a

ariety of mechanisms that enable data offloading However most of

he existing solutions such as Local IP Access (LIPA) and Selected In-

ernet IP Traffic Offload (SIPTO) focus on data offloading at the core

f the network [205] For instance LIPA allows an IP-enabled mo-

ile device to transfer data to another device in the same pico or

emto cell without passing through the cellular access network while

IPTO enables the routing of selected IP data flows through different

ateways The only LTE offloading mechanism that supports seam-

ess interworking with IEEE 80211-based WLANs is IP Flow Mobil-

ty and Seamless Offload (IFOM) Specifically IFOM relies on Mobile

Pv6 technologies to allow a user terminal to simultaneously route

elected IP flows over different radio access technologies [209210]

n this way a user terminal can offload selected flows to a WLAN

ased on some operator-defined policy while keeping the LTE con-

ection running However to enable such an approach it is neces-

ary to support an entity in the cellular core network that can com-

unicate with the user terminal to exchange information about the

vailability and quality of neighbouring access networks as well as to

rovide the user terminal with predefined rules to manage the han-

over process This entity in current LTE standards is called the Ac-

ess Network Discovery and Selection Function (ANDSF) server [211]

ote that the problem of selecting the best communication technol-

gy in an heterogeneous wireless network has been extensively in-

estigated in the literature both with centralised as well as decen-

ralised approaches (see [212] for a survey) However the use of mul-

iple interfaces in parallel as well as per-flow offloading are relatively

ew concepts Thus the design of scalable and efficient network se-

ection strategies for the ANDSF framework is still an open issue For

nstance three offloading methods suitable for the ANDSF framework

re proposed in [213] which are based on coverage SNR and sys-

em load A reinforcement learning approach is designed in [214] that

llows multimode base stations to autonomously steer their traffic

ows across different access technologies depending on the traffic

ype the usersrsquo QoS requirements the network load and the inter-

erence levels A number of studies have also tried to quantify the

B Bellalta et al Computer Communications 75 (2016) 1ndash25 13

f

c

8

p

R

p

i

m

l

a

m

c

p

p

f

m

t

a

e

p

t

m

a

l

T

m

a

i

n

t

t

t

u

p

c

w

f

a

t

8

a

m

a

s

l

a

m

b

m

d

S

i

a

o

a

R

t

n

G

fi

s

t

p

t

c

u

c

w

i

o

p

g

a

c

i

f

c

(

d

m

l

i

R

t

l

t

L

m

a

m

o

e

f

a

I

i

t

w

i

t

a

c

w

p

f

t

d

5

i

s

e

s

w

5

a

c

a

n

requencies lower than 1 GHz are evaluated in [115] The authors

ompare two path loss propagation models proposed by the IEEE

0211ah Task Group (one for macro and one for picohotzone de-

loyments) [116] with Lee and HatandashOkumura propagation models

esults show that the IEEE 80211ah channel models underestimate

ath loss with respect to Lee and Hata models Moreover in a compar-

son with empirical data it is observed that the IEEE 80211ah channel

odels also underestimate the initial loss and the slope of the path-

oss function A new model parameterisation is thus proposed by the

uthors In [117] the feasibility of an IEEE 80211ah WLAN deploy-

ent is also evaluated in terms of the achievable range and bit rate

omputed on the basis of the link budget using the same path loss

ropagation models as in [115] Results show that the transmission

ower limitations in the uplink can limit the overall network per-

ormance Finally in [118] the authors evaluate the achievable trans-

ission range for the different transmission rates and the achievable

hroughput for different combination of transmission power values

nd transmission rates Further studies are required due to the het-

rogeneity of scenarios in which IEEE 80211ah WLANs can be de-

loyed to characterise the effective coverage ranges with special at-

ention to mixed indoors and outdoors scenarios and in presence of

obile nodes

Due to the wide coverage IEEE 80211ah WLANs may be severely

ffected by the presence of hidden terminals To mitigate such a prob-

em smart solutions to guarantee that the STAs assigned to the same

IM Group are inside each other transmission range or the employ-

ent of RTSCTS-based solutions may contribute to mitigate such

problem An overview of long-range scenarios for IEEE 80211ah

s presented in [119] including a detailed description of the hidden

ode problem in them

Single-hop uplink transmissions from distant STAs require high

ransmission power to reach the AP Since not all STAs may be able

o transmit at the required power a solution could be the introduc-

ion of nodes able to relay transmissions from them However the

se of relays has to be efficiently harmonised with the operation in

ower saving mode allowing temporal periods in which the relays

an gather the data from their associated STAs and then periods in

hich they forward the data to the AP which is still an open challenge

or IEEE 80211ah WLANs In case relays are used data prediction and

ggregation techniques could be implemented to make more efficient

ransmissions [120]

Another challenge is the assessment of the efficiency of IEEE

0211ah PS mechanisms including their optimisation In [111] the

uthors evaluate the capacity of the TIM and page segmentation

echanisms in terms of the maximum number of nodes supported

nd the energy consumed given a certain network traffic profile Re-

ults confirm that a large number of STAs can be supported with

ow energy consumption The impact of the number of nodes oper-

ting in PS mode on the energy consumption and the delay perfor-

ance is analysed in [121] A detailed analysis of the TIM Group-

ased channel access adopted by the IEEE 80211ah task group is

ade in [122] where the STAs assigned to each TIM are uniformly

istributed between the slots of the RAW Since only one group of

TAs is allowed to transmit in each RAW slot the channel contention

s minimised The number of downlink and uplink slots in a RAW

s well as their duration is optimised on the basis of the number

f active STAs and the network traffic profile in [123] In [124] the

uthors propose an algorithm to determine the optimal size of the

AW interval for uplink transmissions based on the estimation of

he number of STAs transmitting and the duration of the RAW Fi-

ally in [125] the authors address the contention problems in Smart

rid communication networks with many nodes and periodic traf-

c when using an IEEE 80211ah WLAN It is still a challenge to con-

ider smart systems able to adapt the IEEE 80211ah parameters to

he instantaneous system state in order to save as much energy as

ossible

Since the PS mechanisms affect only the operation of TIM STAs

he performance of non-TIM and Unscheduled STAs has not yet been

onsidered in the literature To identify the scenarios in which the

se of non-TIM and Unscheduled STAs are of interest is still an open

hallenge as further investigating how non-TIM STAs negotiate over

hich PRAWs they can use to transmit and receive data Co-existence

ssues between the three types of STAs in a single network is also an

pen challenge that requires further work to guarantee a minimum

erformance level for all of them

The design of strategies to distribute the STAs between all the TIM

roups based on their specific traffic profile position battery level

nd application priority is another important challenge that is still

ompletely open EDCA is the default channel access scheme included

n the IEEE 80211ah amendment and provides some basic traffic dif-

erentiation capabilities at the packet level Besides the different ac-

ess categories (ACs) needing to be renamed and their parameters

Arbitration Inter-Frame Spacing or AIFS CWmin TXOP duration) up-

ated to fit the specific traffic profiles of M2M communications other

echanisms can be implemented A mechanism to assign the down-

ink RAW slots when packets from multiple priority levels are wait-

ng for transmission at the AP is required for example Also uplink

AW slots are currently selected randomly by the STAs that want to

ransmit a packet A mechanism to reserve some slots at each up-

ink RAW for high priority STAs may therefore be an option although

his may severely degrade the performance of the low priority STAs

astly different repetition patterns for different priority TIM groups

ay also allow priority STAs assigned to those priority TIM groups to

ccess the channel more often We expect this challenge will receive

uch attention in the upcoming years because of the heterogeneity

f sensors that will be connected though a single IEEE 80211ah AP

specially in urban scenarios

Similar to what has been discussed for 4G cellular networks used

or M2M communications [126] the development of mechanisms to

void and resolve congestion situations is another open challenge for

EEE 80211ah WLANs with many nodes For instance a mechanism

n which STAs have to wait a random number of DTIM periods before

hey can start a transmission may be implemented This approach

ould effectively distribute the traffic over a longer period of time

n overload conditions but otherwise would unnecessarily increase

he access delay Adaptive and load-aware solutions may play again

n important role to keep the optimal WLAN operation in all possible

ircumstances

Finally since IEEE 80211ah will compete with 4G5G cellular net-

orks and WSNs to provide M2M connectivity in many different ap-

lication domains such as smart cities or e-health comparative per-

ormance studies to determine the strong and weak points of each

echnology are required including also aspects such as the cost of the

evices and the system reliability

Cognitive radio technology for TV White Spaces

This section gives an overview of the IEEE 80211af amendment

ntroducing its most relevant features and open challenges Fig 9

hows the basic components and features considered for WLANs op-

rating in the TV White Spaces (TVWS) which include local spectrum

ensing by the AP and STAs and the use of geolocation data bases

ith information on channel availability

1 The IEEE 80211af amendment

Thanks to the transition from analog TV to digital TV several VHF

nd UHF spectrum channels used for decades for analog TV broad-

asting are now unused or are under-utilised in many geographical

reas [127ndash129] This ldquodigital dividendrdquo of spectrum has suggested

ational regulators such as the FCC in the US and Ofcom in the UK

14 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 9 WLAN operating in TVWS basic elements and functionalities

Fig 10 Hidden terminal problem in TVWS when the secondary network uses only in-

stantaneous channel sensing to decide whether or not a TV channel is occupied When

the STA is not able to detect the TV signal due to the presence of obstacles it may

initiate a transmission and create interference to the primary users

s

c

5

a

i

p

t

o

o

t

e

c

u

l

m

b

a

m

t

e

t

t

t

m

D

c

I

n

b

u

8

c

a

t

i

t

w

f

to discuss how to reuse these channels for unlicensed devicesrsquo com-

munications [130131] These potentially vacant channels in the VHF

and UHF bands are referred to as TVWS [132] and include spectrum

portions like the 470ndash790 MHz in Europe and non-contiguous 54ndash72

76ndash88 MHz 174ndash216 MHz 470ndash698 MHz and 698ndash806 MHz in USA

A snapshot of the TV spectrum occupancy in the city of Barcelona in

2012 is shown in [133]

The attractive characteristics of TVWS (not only for WLANs) in-

clude the ability to penetrate through walls and other obstacles much

more effectively than other widely used spectrum bands such as the

24 and 57 GHz ISM bands [134ndash137] This fact along with a progres-

sive spectrum scarcity in ISM bands has suggested the birth of the

IEEE 80211af amendment published in February 2014 [138] which

provides the IEEE 80211 operational characteristics for TVWS access

of unlicensed White Space Devices (WSD) A good summary of IEEE

80211af can be found in [137] The main advantage of operating IEEE

80211 WLANs in the TVWS comes from an increased coverage range

which can reach up to 1 km in rural areas and open fields [139]

and less energy needed to transmit However this comes at the cost

of an increased interference risk to other WSDs which creates co-

existence problems and demands new PHY and MAC layers to ef-

ficiently support channel access and operations preserving licensed

usersrsquo devices

WLANs operated in the TVWS could cover a number of interest-

ing and emerging use cases and scenarios eg Internet access in ru-

ral or sparsely populated areas Smart Grid sensor aggregation me-

tering and control Internet of Things advanced WLAN operations

and TVWS traffic offloading in indoor environments [140141] Nev-

ertheless the concept could be realised to create or extend com-

mercial or municipality WiFi services offered to citizens with cover-

age at the whole city level realised with reasonably limited network

infrastructures [142143]

IEEE 80211af will use the PHY and MAC layers derived from IEEE

80211ac It will adopt concepts such as the OFDM multi-user beam-

forming contiguous and non-contiguous channel bonding and packet

aggregation Among the mandatory and innovative behavioural and

operational parameters the most notable one is the channel acqui-

sition support realised through remote geolocation-based spectrum

allocation databases which maintain the channelsrsquo availability infor-

mation in any given area and time of day providing upon request the

list of free channels available for use

In the following section we highlight and summarise three main

novelties that IEEE 80211af introduces In Section 511 we explain

how the access infrastructure to the remote spectrum database is de-

signed and what the requirements are we discuss the coexistence is-

sues and methodologies adopted in IEEE 80211af and we also discuss

the novel concept of non-contiguous channel bonding The following

ections also contain an illustration of related works Open research

hallenges will be summarised in Section 512

11 Novel features

This section provides an illustration of novel features introduced

nd discussed in the path to IEEE 80211af and the directions provided

n related works and standardisation initiatives to resolve classical

roblems properly characterised in the new framework of TVWS

echnologies

Channel acquisition spectrum database and channel sensing Much

f the attention in operating in the TVWS is given to the protection

f the primary licensed users in the TVWS spectrum band In general

he primary users were considered as the Digital TV (DTV) broadcast-

rs and receivers In fact receivers would suffer the wasteful effect of

ollisions during the TV broadcast reception in the case of secondary

sersrsquo (SU) transmission interference (Fig 10)

It is a well-known problem that the definition of a coherent wire-

ess channel status (idlebusy) from the viewpoint of distributed pri-

ary users (eg DTV receivers) is made difficult by the difference

etween transmitters and receivers interpretation of signals in a vari-

ble time-space collision domain resulting in hiddenexposed ter-

inals concepts Solutions proposed to reduce the hiddenexposed

erminals included the RTSCTS mechanisms and beaconing How-

ver it is difficult to implement solutions that would effectively de-

ermine the channel status of TVWS channels based on localised dis-

ributed sensing activities performed by multiple secondary users in

he same collision domain On the other hand the use of RTSCTS

echanism would not be effective in this case because most of the

TV primary users (and specifically the receivers) were not originally

onceived to transmit signals or implement the RTSCTS handshake

n conclusion the actions to identify free (and busy) TVWS chan-

els should not be based on primary usersrsquo involvement but should

e almost totally implemented by specifically designed secondary

sersrsquo methodologies

There has been much debate on the methodology that IEEE

0211af secondary devices should adopt to get knowledge of free

hannels at a given time on a given target communication area in

n effort to guarantee verified optimised and reliable TVWS spec-

rum use The three main methodologies discussed include sens-

ng solutions geolocation databases (DB) and beaconing Moreover

wo approaches are possible in general distributed and centralised

ith their well known tradeoffs in terms of effectiveness required in-

rastructure deployment and coordination overhead The distributed

B Bellalta et al Computer Communications 75 (2016) 1ndash25 15

m

b

s

c

t

u

a

t

o

l

t

w

t

t

a

o

c

c

a

(

t

o

t

t

D

t

t

o

t

f

a

t

c

i

a

c

d

t

s

t

b

I

l

(

e

a

i

R

p

e

t

c

c

i

b

i

i

t

b

t

t

t

W

a

p

c

i

t

h

p

t

n

a

t

e

t

t

t

p

t

a

r

8

w

s

w

[

i

S

o

a

t

f

i

8

i

r

a

a

i

t

c

(

I

s

a

e

t

t

c

[

i

c

i

g

s

n

s

t

i

o

o

s

c

D

o

ethodology that is receiving more attention in the literature is

ased on distributed channel sensing solutions [127144145] These

olutions include the observation of spectrum use and possible

ooperative aggregation of spectrum sensing information to increase

he accuracy of detection and reduce the vulnerability of primary

sersrsquo transmissions The spectrum sensing capability is required in

ll the secondary devices with a minus114 dBm sensitivity Dynamic

ransmission power control must be provided and the upper limit

n emissions is 100 mW EIRP (20 dBm) for portable devices (further

imited in the case of adjacent channel use to reduce out-of-band in-

erference) However database (DB) coordination of TVWS combined

ith spectrum sensing is considered the most promising and effec-

ive technique compared with spectrum sensing alone The DB spec-

rum information has more chances to be effective (ie identifying

ll free channels opportunities) and reliable (ie not prone to attacks

r erroneous interpretation of channels status) In fact regulators de-

ided to push for the conservative solution of a remotely accessible

entralised spectrum DB that maintains the information on the avail-

bility of all TVWS channels at any given point in time and location

with a target accuracy around 50 m) under their responsibility For

his reason the spectrum databases have become a mandatory part

f many spectrum sharing systems and the dominant technical solu-

ion to support TV white spaces

The IEEE 80211af standard adheres to this vision and also has

he role of a common regulatory framework for different spectrum

B implementations defining a generalised coordination architec-

ure protocols and interfaces for spectrum queries and local spec-

rum information management The protocol explicitly considers out

f scope the communication protocol and technology adopted in

he background leaving the freedom to those players responsible

or the deployment to select any suitable Internet-based and local

ccess network technology Secondary devices that need to access

he TVWS spectrum should get information on which channels they

an effectively use and which parameters to adopt without impact-

ng primary usersrsquo transmissions (and receptions) by querying the

vailable spectrum DB The query must contain the transmission

haracteristics and accurate geographical position of the secondary

evice which can be determined by means of a geo-positioning sys-

em (GPS) regularly updated in the case of movement (or manually

et for static devices) The IEEE 80211af reference system architec-

ure for DB spectrum access is composed of multiple Geolocation-

ased Spectrum information Databases (GDBs) entities connected via

nternet to Registered Location Secure Servers (RLSS) RLSSs work as

ocal proxies of the GDB for a localised group of Basic Service Sets

BSS) RLSS are connected via a secure protocol architecture with the

quivalent of multiple Access Points (AP) in different BSS realising

trusted DB infrastructure APs locally coordinate the exchange of

nformation and channel access management between the GDB (via

LSS) and the secondary usersrsquo end stations (STA)

Mode I devices are those under the control of a device that em-

loys geo-location database access while Mode II devices are those

mploying geo-location database access by themselves The informa-

ion exchange provided between the GDB and a secondary user STA

an be provided in both an open-loop (eg adopted by FCC) and a

losed-loop (eg adopted by ETSI) implementation In an open loop

mplementation daily spectrum availability information is provided

y the GDB and no feedback on the spectrum information received

s provided by STAs thus the system approach for spectrum access

s much more conservative leading to low channelsrsquo utilisation po-

ential In a closed loop implementation the STA can provide feed-

ack to the GDB and there are more communication overheads due

o the high granularity of updates however the system is more effec-

ive and reliable in the exploitation of TVWS on behalf of STAs The

ypical information provided by the GDB includes (1) the updated

hite Space Maps (WSM) of frequencies allowed for secondary use

t the timespace of the querying STA and (2) the device-dependent

ower limitations for transmission (in general conservative and ac-

urate enough to avoid relevant interference effects on primary users

dentified in the area)

The basic IEEE 80211af mechanisms regulating the communica-

ion between STA and GDB can be found in [137138]

In parallel with IEEE 80211af standards such as the IEEE 19006

ave been created that consider interfaces and data structures sup-

orting spectrum sensing information exchange applicable to spec-

rum sensing (and particularly distributed spectrum sensing) sce-

arios In particular in October 2014 the IEEE 19006 WG initiated

project for a new standard called IEEE 19006b [146] concerning

he use of spectrum sensing information to support and optimise the

ffectiveness reliability and robustness of spectrum database solu-

ions The aim is to enhance the performance and capabilities of spec-

rum databases through the use of spectrum sensing information

Co-existence Support for co-existence mechanisms so that mul-

iple technologies can effectively utilise the TVWS spectrum is im-

ortant Self co-existence between network devices of a common

echnology (eg deployed by different operators in the same area)

nd co-existence among different technologies are relevant topic of

esearch for Cognitive Radio (CR) systems and specifically for IEEE

0211af Many solutions appeared on the research scene but no one

as so far finalised as the target solution for IEEE 80211af Specific

tandardisation has been started to regulate coexistence between

ireless standards of unlicensed devices including the IEEE 802191

147] The purpose of the IEEE 802191 standard is to enable the fam-

ly of IEEE 802 Wireless Standards to most effectively use TV White

pace by providing standard coexistence methods among dissimilar

r independently operated TVWS devices Early examples of gener-

lised coexistence mechanisms included Dynamic Frequency Selec-

ion (DFS) Transmission Power Control (TPC) listen before talk (eg

or contention based IEEE 80211 80215) time division multiplex-

ng (also among different techniques such as the IEEE 80216 80220

0222) and Message-based Spectrum Contention (that is beacon-

ng messages that carry coexistence information) Opportune met-

ics must be defined to assess the measurable coexistence achieved

mong different technologies as an example the hidden node prob-

bility for a target scenario or the estimate of percentage variation

n normalised network throughput and latency (before and during

he SU transmissions) On the other hand a centralised coexistence

ontrol mechanism could be effectively realised by a central manager

or coexistence-DB like the GDB) in critical scenarios To this end

EEE 19004 [148] (a standard for heterogeneous networks in dynamic

pectrum context part of IEEE Standards Coordinating Committee 41)

ims to standardise the overall system architecture and information

xchange between the network and mobile devices which will allow

hese elements to optimally choose from available radio resources

There are three possible classification of co-existence archi-

ectures and inter-network coordination channels for CR systems

entralised coordinated and autonomous co-existence mechanisms

149] In centralised co-existence schemes the co-existence is admin-

stered by a central entity (eg like in IEEE 802191) This solution

ould be applied in both homogeneous IEEE 80211af systems (eg

ncarnated by the centralised spectrum DB entity) and in hetero-

eneous inter-networks without requiring modifications to existing

tandards under the assumption that all the involved co-existing

etwork entities would adhere to the same centralised coordination

cheme and inter-network coordination protocol On the other hand

his assumption could be hard to satisfy in some practical scenar-

os The centralised schemes are effective in providing minimisation

f inter-network interference based on the availability and quality

f centralised co-existence information In coordinated co-existence

chemes the centralised entity could be present but not taking de-

isions in general The central entity could provide a co-existence

B with required information which can be used via in-band and

ut-band signalling for co-existence decisions taken by the cluster

16 B Bellalta et al Computer Communications 75 (2016) 1ndash25

c

e

n

c

o

s

i

d

a

S

8

n

a

b

o

b

i

t

a

8

b

[

t

i

n

b

5

t

a

c

b

s

a

c

n

s

p

i

c

l

r

c

f

a

v

t

a

e

c

t

Q

a

b

s

i

c

u

n

c

i

heads of many co-existing inter-networks for implementing a proper

common coordination protocol (eg out-band busy tone signalling)

These are often hybrid solutions realising a compromise between ef-

fectiveness and suitability In autonomous co-existence schemes all the

decisions and coordination are implemented in a distributed way

by the involved network entities In these schemes many policies

can be adopted to realise a sufficient level of co-existence under a

best effort approach Solutions that could be implemented include

the use of busy tones beaconing and other signalling protocols in

dedicated control channels dynamic distributed frequency selection

schemes listen before transmit token-based and dynamic reserva-

tion schemes All the above mentioned solutions have goods and

bads and a general illustration can be found in [149] Other proac-

tive co-existence techniques try to early detect and recovermitigate

co-existence issues with relaxed inter-network coordination eg re-

alised via spectrum sensing and interference avoidancesuppression

techniques

It must be clear that different aims exist regarding the co-

existence for primary TVWS users protection against secondary

users and for secondary devices mutual interference avoidance

Many co-existence problems have been analysed in the literature

for TVWS technologies in particular between the primary (licensed)

and secondary (unlicensed) devices The coexistence management in

TVWS is complex because primary users of DVB transmissions are in-

tended to be the pure receivers rather than the broadcasters In other

words potential protected users could be everywhere and the hid-

den terminal problem could arise for these systems The co-existence

problem is also complicated by many factors that cause asymme-

try and dynamicity in the TVWS such as the mobility variable den-

sity power asymmetry and heterogeneous MACPHY layers TVWS

co-existence problems originate between the IEEE 80211af technolo-

gies and the IEEE 80222 technologies for wireless regional area net-

works (WRAN) Results of analysis of TVWS usage in Europe show

that white spaces are typically present and fragmented They are typ-

ically more abundant in rural areas where larger contiguous blocks

of unused channels are available due to broadcast network planning

giving priorities linked to population density However the exploita-

tion of TVWS in urban areas is possible eg some recent research

was realised under the assumption of exploiting shadowing effects

created in the communication environment (eg by buildings obsta-

cles) in favour of frequency reuse [150]

Since different standards for opportunistic communication in the

TV White Spaces have now been published such as IEEE 80222 [151]

IEEE 802154m [152] Weightless [153] and of course IEEE 80211af

improved methods to guarantee the coexistence of different devices

operating on several protocols in the same bands must be deployed

We highlight here some work eg [150154ndash157] that focuses on

the coexistence between different technologies operating in the same

bands In particular [155] studies the performance degradation of

IEEE 80222 when an IEEE 80211af network is operated in the same

area The problem becomes even worse when the IEEE 80211af net-

work is located near 80222 user equipment causing both networks

to perceive a strong interference from each other One proposed solu-

tion is the Coexistence Beacon Protocol [158] studied for 80222 net-

works which foresees the exchange of a periodic beacon to identify

the neighbouring and possibly interfering networks

For what concerns the co-existence problems research has ad-

dressed other important related aspects of new WLANs on TV White

Spaces For instance the higher coverage range makes them attrac-

tive for applications in the smart grid and also for M2M on TVWS

[134150159] Here the increased range compared to standard tech-

nologies in the ISM bands can have beneficial effects for indoor

mobile devices [150] and M2M [159] Specifically indoor devices

communicating on TV White Spaces offer better propagation char-

acteristics and penetration through obstacles making it easier to

realise scalable home connectivity [135] although interference and

o-existence problems could be exacerbated and must be resolved

g see [139] However studies in the literature show how the sig-

al coming from an indoor transmitter on TV White Spaces remains

onstrained inside the house making it difficult to deploy indoor-to-

utdoor networks On the other hand this shadowing limit can be

een as an opportunity lowering the interference outside the build-

ng in which the TVWS network is locally operated [134150] Un-

er this approach HDTV streaming has recently been considered as

possible use case of WLAN networks operating in the TV White

pace [160]

Non-contiguous channel bonding A relevant novel feature of IEEE

0211af is the potential for contiguous and non-contiguous chan-

el bonding which permits aggregating basic channels (also non-

djacent ones) and leveraging the possible large frequency spread

etween multiple available channels Due to the rather static nature

f primary DTV transmissions where a busy channel is unlikely to

ecome free in the near future and state changes are coarse-grained

n general it is crucial to exploit the time-locality effect and exploit

he maximum physical channel availability that could be aggregated

t any given location With the methodology inherited from IEEE

0211ac IEEE 80211af is capable of bonding together two up to four

asic channels grouped in up to two different non-contiguous chunks

137] The spectrum bandwidth of a DVB-T basic channel can be ei-

her 6 7 or 8 MHz depending on the country in which the service

s operated As an example this creates a 144 to 168 OFDM chan-

elsrsquo bandwidth potential when up to four 6ndash7ndash8 MHz channels are

onded [137]

12 Open challenges

The deployment of WLANs in the TVWS at their maximum poten-

ial still requires the resolution of open problems Generally speaking

s mentioned in previous sections the challenges for such networks

an be divided into three main categories (1) spectrum sharing

etween opportunistic (secondary) and heterogeneous devices (2)

pectrum sensingmanagement and maximum exploitation of avail-

ble spectrum (potentially enabling the spectrum-on-demand con-

ept) and (3) co-existence and interference mitigation to the primary

etwork (primary user protection) and secondary networks [154]

The overall aim of spectrum sharing techniques is to maximise

eparation to avoid overlapping or contiguous channels operations

rovided that sufficient channels are available in the TVWS spectrum

n a given spacetime scenario [154] In general important research

ontributions have to be realised for the design of proper spectrum al-

ocation methodologies and techniques satisfying multi-factorial QoS

equirements at the system-level and the user-level Novel ideas in-

lude the spectrum sharing and spectrum sensing techniques being

urther divided into cooperative and non-cooperative

Non-cooperative solutions attempt to realise spectrum sharing

nd sensing on a local basis without direct cooperation between de-

ices Examples include the ldquolisten-before-transmitrdquo approach and

ransmission power control to limit the interference and spectrum

llocation policies that are based on local node feedback only Coop-

rative solutions rely on the existence of a common communication

hannel through which devices can tentatively agree on the spec-

rum allocation that provides the desired spectrum separation and

oS Major issues include the definition of a methodology to identify

common channel (or multiple channels mutually shared in space

y pairs of heterogeneous and distributed devices) and to efficiently

hare common information For the access to the spectrum DBs the

dentification of proper common access channels is still under dis-

ussion out-of-TVWS-band cellular communications can be widely

sed given their high coverage also indoor

Trying to avoid the use of complex and resource-hungry coordi-

ation schemes and dedicated control channels rendezvous proto-

ols play a significant role both for co-existence and for establish-

ng dynamic communication channels based on spectrum sharing

B Bellalta et al Computer Communications 75 (2016) 1ndash25 17

R

n

c

t

l

o

m

e

n

i

n

w

n

a

u

s

p

b

s

m

s

a

d

t

b

a

a

m

a

t

u

8

n

t

d

d

m

f

t

n

d

n

e

r

a

h

b

f

a

c

t

u

i

f

t

d

t

s

i

f

I

l

i

a

n

c

d

I

n

o

t

i

t

e

w

t

c

m

t

a

i

t

q

b

c

a

i

l

i

a

v

t

a

a

c

t

i

t

b

t

t

a

D

p

c

s

o

q

r

p

a

g

t

e

u

d

t

m

n

t

s

f

i

1

8

f

T

endezvous is considered a fundamental problem in generalised CR

etworks at the basis of many communication processes eg in-

luding neighbour discovery routing and broadcast Rendezvous pro-

ocols attempt to establish a new link for communication on a se-

ected and agreed frequency band (channel) identified within a set

f available resources (channels) thus creating the basis for com-

unication between two or more SUs This problem is often exac-

rbated by the high (and unknown) numbers of contending SUs high

umber of available channels and their ambiguous boundaries and

dentifiers and tough co-existence requirements satisfaction in cog-

itive radio systems [161] Preliminary approaches for rendezvous

ere based on the existence of a dedicated Common Control Chan-

el (CCC) agreed among all the SUs or the existence of a centralised

gent (decision maker) assigning channels to any pairs of requesting

sers This approach could be considered in the case of IEEE 80211af

olutions based on the spectrum DB implementation however the

roblem is complicated by the fact that the common channel must

e identified and made available for all the requesting SUs at the

ame time even when these are spread over multiple collision do-

ains in space On the other hand the scalability and reliability is-

ues which are caused by the congestion of CCCs and the vulner-

bility risks of centralised DBs motivate the research of alternative

istributed solutions called blind rendezvous protocols [162] Dis-

ributed solutions have been mainly based on beaconing andor slot-

ased channel hopping algorithms exploiting randomisation sensing

nd discoverysynchronisation messages spread in slotted time over

set of candidate available channels [161] These protocols aim to

inimise the convergence time to maximise the use of resources

nd to avoid collisions among SUs by identifying and agreeing on

he use of a given common channel among N non-overlapping and

nambiguously labelled channels The main complication in IEEE

0211af systems is given by the wide availability of many different

on-contiguous spectrum bands which cannot be uniformly quan-

ised in single channels and unambiguously labelled in space in or-

er to realise a common reference domain for all the IEEE 80211af

evices In fact the knowledge of the number of channels their com-

on labels the unique IDs of SUs and the number of SUs contending

or the channel assignment are the key factors determining the effec-

iveness and good properties of blind rendezvous protocols Recently

ew classes of rendezvous protocols have been proposed which are

istributed blind and oblivious Oblivious means that available chan-

els may be identified with different labels on behalf of the differ-

nt involved SUs A number of these distributed blind and oblivious

endezvous protocols have been proposed and have been referenced

nd analysed in [162] To conclude solutions for dynamic spectrum

andoff issues (eg in case PUs suddenly appear in a region) must

e resolved in advance to minimise the latency of re-establishing a

unctional communication channel eg by implementing a proactive

lternative rendezvous channel selection in background to ongoing

ommunication processes [163]

Another challenge is to provide cooperative techniques in a dis-

ributed vs centralised implementation by analysing their convergence

nder variable conditions overheads and tradeoffs In some cases as

n vehicular networks the exploitation of scenario characteristics and

actors such as the constrained mobility could help to realise effec-

ive dissemination of sensing information over an extended and pre-

ictable horizon under a cooperative approach [164ndash166] Coopera-

ion could be helpful in the realisation of the cooperative spectrum

ensing in combination with the Geolocation DB approach Another

nteresting direction of research is the investigation of mutual ef-

ects of coexisting cooperation-based vs non-cooperative techniques

n cooperation-based devices the exchange of information could al-

ow a quick convergence to a solution and avoidance of further

nterference [145] Certainly cooperative techniques have desirable

dvantages over non-cooperative ones However tight time synchro-

isation and high coordination overheads are required The common

hannel for cooperation purposes could become a bottleneck and re-

uce the potential advantages of dynamic spectrum allocation The

EEE 80219 foresees the presence of a shared common control chan-

el (CCC) to which devices need to tune in order to gather the state

f the network Solutions such as [141] propose the use of TVWS for

he CCC

A promising direction for research is the adoption of cluster-

ng schemes for secondary devices associated with a hybrid dis-

ributedcooperative vs centralisednon-cooperative approach in an

ffort to reduce overheads while maximising advantages In this

ay secondary nodes could implement cooperation for sensing func-

ions using properly identified common channels and delegate the

entralised decisions and spectrum-DB management to well instru-

ented cluster-leader nodes This concept is incarnated to some ex-

ent by RLSS in IEEE 80211af More recently there have been propos-

ls to build distributed databases that refresh their contents period-

cally by querying the remote spectrum database [167] Here Mas-

er devices (according to Ofcom terminology) periodically cache the

uery replies from the spectrum database in order to reply to and

roadcast the spectrum availability to neighbour Slave devices

Another challenging direction for research is the definition of ac-

urate models and efficient simulation tools enabling dynamic spectrum

nalysis for both rural and urban areas Recent developments of dig-

tal maps simulation tools and propagation models theoretically al-

ow improving the capability of predicting radio propagation effects

n complex scenarios with an acceptable computation time This en-

bling technology could be used in parallel to spectrum DBs to pro-

ide more accurate forecasting of dynamic frequency allocation in

imespace scenarios Another issue to be taken into account is the

ccuracy of the remote spectrum database regarding the channel

vailability estimation Several works have already shown the inac-

uracy of propagation models due to the complexity of parameters

o be considered and differences between modelled and real scenar-

os [132134168169] Better propagation models will make it possible

o more efficiently estimate the interference between devices and

uild Radio Environment Maps (REM) or White Space Maps (WSM)

hat will possibly lead to a more accurate sharing of the radio spec-

rum [144] To this end current research could focus on how to lever-

ge simulation models in order to build more accurate frequency

Bs

In general the adoption of remote spectrum DBs methodology

ushed by national regulators has made the spectrum sensing pro-

ess optional and conservatively realised (eg power control for

ensing-only devices is limited to below 17 dBm EIRP) Thus much

f the research has focused on techniques and solutions to successfully

uery the remote database also in challenging environments such as

ural areas and indoor However since the queries must provide the

osition of the mobile device the accuracy of positioning for mobile

nd handheld devices still needs more appropriate solutions Trian-

ulation can be used to estimate the position of mobile devices with

he accuracy needed by the current regulations outdoor (50 m) How-

ver a low GPS accuracy is possible indoors [170] Certified man-

al positioning can be provided by technicians for static and indoor

evices although limiting mobility Alternatives to costly infrastruc-

ures for positioning (eg based on short range beaconing devices)

ust be deployed in indoor scenarios

Finally regarding spectrum sensing and interference mitigation

ew challenges come from the heterogeneity and different charac-

eristics that networks operating in the TV White Spaces can have

uch as different transmitting power (4 W for static devices 100 mW

or mobile and portable devices and 40 mW for communication

n channels adjacent to occupied ones) different bandwidths (5

0 15 and 20 MHz are currently foreseen for IEEE 80211af IEEE

02154m can have narrower bandwidths or wider ones) and dif-

erent medium access schemes (protocols that use either CSMA or

DMA are required to co-exist in the TV White Spaces) With all this

18 B Bellalta et al Computer Communications 75 (2016) 1ndash25

d

a

T

i

a

c

t

l

b

o

t

a

d

p

a

r

p

n

a

m

t

p

p

d

d

o

t

i

D

w

p

i

a

a

d

b

fi

d

o

t

l

b

t

b

a

i

e

o

o

c

t

t

w

A

c

M

s

c

l

O

w

g

v

t

in place the problem of determining whether another secondary de-

vice is currently transmitting in the same channel is still an open

issue

Spectrum sensing is a very hot topic in the literature with many

proposals that leverage the cooperation between devices to raise the

accuracy of the detection process eg [144171] The current litera-

ture also offers room for proposals that suggest exploiting sensing to

build more accurate spectrum DBs [164172173] Lastly regarding the

interference mitigation to the primary network regulators provide a

conservative approach based on a remote spectrum database to avoid

high risk for the primary users However in particular for the indoor

scenario the concept of secondary usersrsquo transmission in TV Grey

Spaces has been proposed TV Grey Space identifies busy channels

that are formally used by primary users in the area of interest (eg

at the rooftop level of a building for DVB-T) but which could be con-

sidered usable indoors (eg at the basement level) without causing

effective interference to the primary users [134150174] This concept

is based on the generalised spectrum underlay paradigm [175] The

remaining research challenge is to provide TV Grey space access guar-

anteeing a sufficient amount of protection to the primary user while

providing a satisfactory QoS to opportunistic devices eg through

the use of customised and validated propagation models for indoor

TVWS networks and feedback mechanisms to constantly monitor the

interference generated to the primary network

6 Emerging new trends and technologies

In this last section we review three emerging trends that in our

opinion will have a large influence in the conception of future WLANs

as they change the way WLAN protocols and functionalities are de-

veloped implemented tested and integrated with other wireless

networks

61 Programmable Wireless LANs

Especially in the enterprise environment WLAN deployments

need to support a wide range of functionalities and services This is

intrinsically difficult because of the large number of APs that must be

managed which calls for scalable solutions Typical services include

channel assignment load balancing among APs authentication au-

thorisation and accounting (AAA) policy management support for

client mobility and interference coordination Another problem is the

fact that WLAN clients autonomously take several decisions such as

which APs to associate with when to hand-over etc Therefore sup-

porting roaming clients requires the management of a large number

of association states across several APs which is a challenge if sup-

port for real-time hand-over is desired Typically such management

schemes are centralised and most of them are proprietary such as

WLAN controller solutions from Aruba [176] or Cisco [177] although

the IEEE 80211u amendment has been released to allow mobile users

to seamless roam between WiFi networks with automatic authenti-

cation and handoff [178] For example Dyson [179] enables STAs to

send information such as radio channel conditions to a centralised

controller based on a custom API (eg based on Python) As the con-

troller has a centralised view of the network it can enforce a rich

set of policies to control the network also using historical informa-

tion A demo system has been implemented along with applications

such as airtime reservations for specific clients or optimised hand-

offs However Dyson requires STAs to be modified in order to use

those new services offered by the centralised controller TRANTOR

[180] is another example of a centralised management system that

requires changes of clients in order for the infrastructure to gather

information from them in terms of eg interference measurements

In addition control commands enable the infrastructure to exercise

control such as modifying the transmit power or influencing the as-

sociation procedure Clients still use standard CSMA MAC layer for

ata transmission In contrast to DenseAP [181] or DIRAC [182] which

re also based on a centralised controller but do not modify clients

RANTOR exercises much more control on clients by the use of a ded-

cated API which enables a significantly higher gain in coordination

nd thus capacity CENTAUR [183] is another example of a centralised

ontroller that centrally schedules hidden and exposed terminals in

he downlink while using standard CSMA MAC for uplink traffic and

egacy downlink traffic To achieve good performance it uses a fixed

ack-off packet staggering techniques and a hybrid data path that

nly schedules downlink transmissions towards hidden and exposed

erminals centrally All other traffic is sent using standard DCF with

standard back-off procedure In contrast to previous approaches it

oes not require any modifications to the clients but does require data

lane centralisation and it remains questionable how scalable such

solution is for large WLAN deployments with high PHY layer data

ates beyond 100 Mbps However new approaches for fast data plane

rocessing that are currently explored in the virtualisation commu-

ity which move all the packet processing into user space may be

n option to significantly increase speed Designing such manage-

ent systems raises several interesting research questions (a) Cen-

ralised infrastructures may incur high latency but are more simple to

rogram and maintain while designing a completely distributed ap-

roach that operates close to WLAN devices may result in significant

istributed coordination and consistency problems (b) what level of

istributed control is required in order to support the flexibility needs

f future WiFi based networks supporting a high level of mobility

Traditionally WLAN APs are built on proprietary operating sys-

ems that are tightly coupled with the hardware This design makes

t hard to create new applications on top of such networking devices

espite the fact that protocols and mechanisms are available that

ould greatly improve the utility of existing networks those new

rotocols are not deployed because the closed system design makes

t very difficult to extend their functionality An important aspect for

ll such new approaches is therefore how to provide open interfaces

nd open source to speed up innovation As an example Linux-based

evices are completely open source but in order to increase flexi-

ility much more work is needed in the area of open drivers and

rmware In fact the Atheros based ath9k driver has been the main

river of innovation in the open source community because it is the

nly driver that interworks with open firmware In general solutions

hat modify the OS driver focus mainly on programmable network

evel solutions for WLANs such as channel switching or handovers

etween APs While the ath9k driver is an excellent example in how

o speed up innovation it is still very cumbersome to support flexi-

le MAC engine reprogrammability with ath9k In contrast MAClets

nd Wireless MAC processors (WMP) [184] allow a much more flex-

ble reprogramming of MAC functionalities A MAC processor is an

ntity able to execute general MAC commands that specify the MAC

perations through a software-defined state machine The behaviour

f the MAC protocol can therefore be updated at run-time by simply

hanging the sequence in which those commands are executed (ie

he MAClets) As a proof-of-concept the authors of [184] implement

he proposed MAC processor solution in a commodity WLAN hard-

are card extending the basic DCF in three directions piggy-backing

CKs a pseudo TDMA and the use of multiple channels In [185] a

ontrol framework for this WMP system is also proposed to support

AClet code mobility ie for moving loading and activating MAC

oftware programs embedded into ordinary data packets (akin to the

apsule model of traditional active networks)

The difficulty in re-programming networking hardware has also

ed to the concept of Software Defined Networks (SDNs) based on the

penFlow protocol [186] The main idea of SDNs is to extend net-

orking devices with standardised APIs that allow third-party pro-

rammers to flexibly control the data path In addition SDNs pro-

ide higher level abstractions to network designers and programmers

hrough the use of a centralised control plane offered by a network

B Bellalta et al Computer Communications 75 (2016) 1ndash25 19

Fig 11 The CLOUDMAC architecture VAP stands for Virtual Access Point and WT

stands for Wireless Termination

c

a

c

t

m

W

n

t

t

c

p

t

u

h

a

i

e

s

o

c

a

a

f

T

r

s

a

s

p

o

M

C

i

T

t

fl

A

a

a

i

c

t

s

t

e

I

t

w

H

e

b

I

l

d

o

6

r

W

s

l

a

e

e

a

s

b

c

t

(

S

p

c

F

B

t

p

T

e

t

s

u

m

i

e

o

a

A

p

d

t

t

n

p

i

t

t

f

i

ontroller such as NOX [187] which allows reuse of components such

s topology discovery or network access control for different appli-

ations The SDN concept has recently been applied to WLAN archi-

ecture in order to enable fine grained control over mobility manage-

ent and data forwarding focusing on a programmable enterprise

LAN architecture An important part of the SDN architecture is the

etwork controller which provides a centralised view of (parts of)

he SDN enabled network and uses the OpenFlow protocol to install

he forwarding rules on SDN-enabled devices (routers switches ac-

ess points cellular base stations etc)

ODIN [188] is designed to support programmability in enter-

rise WLAN architecture by separating the association state from

he physical AP They implement Light Virtual Access Points (LVAP)

sing a Split-MAC approach where the infrastructure controls the

andover procedure among different WLAN APs By managing

ssociations through SDN controllers ODIN enables proactive mobil-

ty management and load balancing within the SDN enabled WLAN

nterprise network without the need for changes in the client WLAN

tack or IEEE 80211 MAC layer While ODIN requires agents to reside

n the APs that communicate with the Odin Master within the SDN

ontroller CLOUDMAC [189] is based purely on the concept of SDN

nd virtual APs Similar to ODIN CLOUDMAC implements a Split-MAC

pproach but in addition enables the processing of WLAN MAC layer

rames within a co-located Cloud using so-called Virtual APs (VAPs)

he physical APs in CLOUDMAC are lightweight WLAN APs that are

esponsible only for sending out IEEE 80211 based MAC layer ACKs to

tandard WLAN clients and tunnel WLAN MAC layer frames through

n SDN-enabled enterprise WLAN towards the VAPs As association

tates are kept in the VAPs fast mobility is supported using sim-

le OpenFlow forwarding rules Because of the additional processing

f IEEE 80211 WLAN MAC frames in the co-located Cloud CLOUD-

AC has higher latency than a standard WLAN deployment However

LOUDMAC offers a Webservice based API to third party applications

n order to program the enterprise WLAN and enable new services

he architecture of CLOUDMAC is depicted in Fig 11 It has been ex-

ended in [190] to support QoS management and in [191] to support

exible MAC management frame prioritisation based on IEEE 80211

system based on OpenFlow has been recently proposed in [192] to

llow the station to be associated with multiple APs simultaneously

nd to switch between APs with low overhead

An important aspect to consider in order to enable programmabil-

ty is backwards compatibility Several approaches (eg [180]) require

lients to be modified to utilise the features provided This is difficult

o do in practice because it requires changes in the operating system

oftware of all clients If not all clients can utilise those APIs it is ques-

ionable how usable the new architecture will be and how much ben-

fit in terms of aggregate performance such architecture will allow

n contrast the SDN based approaches are interesting in the sense

hat they do not require changes in the client WLAN stack and work

ith the standard IEEE 80211 MAC layer deployed within the clients

owever it remains questionable how scalable such solutions are For

xample an interesting open research topic is to evaluate the scala-

ility of approaches such as CLOUDMAC [189] In addition processing

EEE 80211 MAC frames within co-located private Cloud requires low

atency support from local Cloud solutions (such as OpenStack) in or-

er to reduce the IEEE 80211 MAC processing time which is an area

f active research

2 Prototyping and testing IEEE 80211 enhancements

Most of the new proposals for next-generation WLANs are cur-

ently only evaluated using mathematical analysis and simulation

hile both analysis and simulation are necessary to characterise and

tudy those enhancements in the initial design phase or to consider

arge-scale scenarios it is difficult to consider all practical aspects of

real-world scenario This can sometimes cause significant differ-

nces between what simulations and real experiments show How-

ver real experiments are challenging because of the high complexity

nd costs of building the new hardware and software for each specific

olution to test

To mitigate implementation complexity there are several flexi-

le hardware platforms where low-level MAC mechanisms can be

ompletely implemented in software Examples based on FPGAs are

he USRP (Universal Software Radio Peripheral) [193] and the WARP

Wireless Open-Access Research Platform) [194] As an alternative

ORA (Microsoft Research Software Radio) [195] works on general

urpose computers by taking some advantage from modern multiple

ore systems

A different approach is provided by OpenFWWF [196] Open-

WWF provides an open CSMACA firmware for specific models of

roadcom chipsets so the resulting firmware can be uploaded and

ested in real commercial hardware OpenFWWF implements a sim-

le State Machine (SM) for controlling the hardware in real time

he SM evolution is driven by a main loop that reacts to events by

xecuting specific handlers When a packet originally prepared by

he Linux kernel is ready in the NIC memory handler Packet_Ready

ets up the radio hardware according to the packet metadata sched-

les the transmission and jumps back to the main loop The Trans-

ission Engine (TXE) then takes care of accessing the channel ie

t decrements the back-off counter according to the DCF rules and

ventually starts the actual transmission This triggers the execution

f the TX_frame_now event that prepares the ACK time-out clock

nd finalises the MAC header If the ACK-frame is received or if the

CK time-out expires and the maximum number of attempts for this

acket is reached handler Update_Params resets the contention win-

ow to its minimum value or otherwise doubles it Finally it loads

he back-off counter with a fresh value

In the following we will review a list of selected papers that use

he OpenFWWF firmware and the WARP or USRP platforms to test

ew proposals for WLANs in real scenarios as examples A first im-

lementation of the groupcast mechanisms defined in IEEE 80211aa

s presented in [197] The authors modify the OpenFWWF firmware

o include those functionalities and evaluate them using a 30 STAs

estbed A second example of the use of OpenFWWF is the collision-

ree MAC protocol presented in [198] In [198] the authors design and

mplement in OpenFWWF a MAC protocol that is able to achieve a

20 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 12 LTE to WIFI offloading

a

n

c

t

[

d

a

W

v

t

t

o

b

f

S

g

l

i

I

s

I

b

n

s

m

a

p

d

c

N

o

v

t

t

n

l

i

a

t

a

collision-free operation by waiting a deterministic time after success-

ful transmissions before to start a new transmission This work shows

the experimental performance of a collision-free MAC (CF-MAC) pro-

tocol for WLANs using commercial hardware The results show that

the proposed CF-MAC protocol leads to a better distribution of the

available bandwidth among users a higher throughput and lower

losses than DCF

In [199] the authors use the WARP platform to test a variant of the

DCF ndash called IEEE 80211ec ndash that substitutes control packets such as

the RTS CTS and ACK with short detectable sequences Since those se-

quences are shorter than the control packets and can be detected cor-

rectly even at lower SNRs values a significant gain in performance is

achieved MIDAS (Multiple-Input Distributed Antenna Systems) [200]

implementation using the WARP platform shows the benefits of dis-

tributing the antennas over the area to cover instead of co-locating

them at the AP in terms of capacity when MU-MIMO is employed

The authors also propose a new MAC protocol to benefit from the

spatial reuse that the DAS allows taking as a basis the IEEE 80211ac

Experimental results show that their proposal is able to achieve up

to 200 gains versus the traditional approach where all antennas are

co-located at the AP

The USRP platform has recently been proposed to develop a first

implementation of IEEE 80211ah in order to obtain a preliminary ex-

perimental performance assessment of such a technology because

no commercial off-the-shelf hardware is yet available as the amend-

ment is still in progress [201] The USRP platform is also used in

[202] to evaluate TIMO (Technology Independent Multi-Output) a so-

lution to deal with high-power non-IEEE 80211 interferers in ISM

bands

To conclude this section it is worth mentioning here also to

MAClets and WMP as given such flexibility to implement and dis-

tribute new MAC protocols and other IEEE 80211 functionalities we

believe that if a use-friendly implementation of such a MAC proces-

sors framework is provided it would significantly contribute to the

development and testing of new MAC enhancements by the research

community

63 CellularWLAN interworking

Public hotspots that offer Internet access over a WLAN using IEEE

80211 technology are now nearly ubiquitous It is forecasted that the

cumulative installed base of WiFi hotspots worldwide will amount

to 551 million by 2018 excluding private hotspots (eg WiFi access

points deployed at home) [203] The sharp increase in the availabil-

ity of public WiFi was initially perceived by mobile cellular operators

as a threat due to the additional competition from wireline Internet

service providers or emerging crowdsourced WiFi networks such as

FON5 However as cellular operators are fighting to cope with the ex-

plosion of mobile data traffic created by the rising use of multimedia

content traffic over mobile devices [15] they are also starting to use

WLANs based on the IEEE 80211 technology to offload data from their

core and access networks In general mobile data offloading refers to

the use of complementary network technologies (in licensed or un-

licensed spectrum) for delivering data originally targeted to cellular

networks Intuitively the simplest type of offloading consists of ex-

ploiting connectivity to existing co-located WiFi networks and trans-

ferring data without any delay (Fig 12) Thus this offloading tech-

nique is known as on-the-spot offloading As a consequence of this

new trend the seamless integration of cellular (eg 3GLTE) and WiFi

technologies has attracted significant research interest in recent years

(see [204] for a survey) a few solutions have already been standard-

ised [205206] and roaming between cellular and WiFi is becoming

increasingly transparent to end users CellularWLAN interworking is

5 httpscorpfoncomen

fl

t

f

lso fostered by the support in the evolving 4G standards of heteroge-

eous network deployments (HetNets) in which the existing macro

ells are complemented with a number of small low-power base sta-

ions with the goal of increasing capacity in highly congested areas

207208] It is envisaged that small cells will be based on 4G stan-

ards (eg pico and femto cells) as well as IEEE 80211 technologies

nd multimode base stations that work simultaneously with LTE and

iFi are already entering the market

It is important to point out that LTE standards already support a

ariety of mechanisms that enable data offloading However most of

he existing solutions such as Local IP Access (LIPA) and Selected In-

ernet IP Traffic Offload (SIPTO) focus on data offloading at the core

f the network [205] For instance LIPA allows an IP-enabled mo-

ile device to transfer data to another device in the same pico or

emto cell without passing through the cellular access network while

IPTO enables the routing of selected IP data flows through different

ateways The only LTE offloading mechanism that supports seam-

ess interworking with IEEE 80211-based WLANs is IP Flow Mobil-

ty and Seamless Offload (IFOM) Specifically IFOM relies on Mobile

Pv6 technologies to allow a user terminal to simultaneously route

elected IP flows over different radio access technologies [209210]

n this way a user terminal can offload selected flows to a WLAN

ased on some operator-defined policy while keeping the LTE con-

ection running However to enable such an approach it is neces-

ary to support an entity in the cellular core network that can com-

unicate with the user terminal to exchange information about the

vailability and quality of neighbouring access networks as well as to

rovide the user terminal with predefined rules to manage the han-

over process This entity in current LTE standards is called the Ac-

ess Network Discovery and Selection Function (ANDSF) server [211]

ote that the problem of selecting the best communication technol-

gy in an heterogeneous wireless network has been extensively in-

estigated in the literature both with centralised as well as decen-

ralised approaches (see [212] for a survey) However the use of mul-

iple interfaces in parallel as well as per-flow offloading are relatively

ew concepts Thus the design of scalable and efficient network se-

ection strategies for the ANDSF framework is still an open issue For

nstance three offloading methods suitable for the ANDSF framework

re proposed in [213] which are based on coverage SNR and sys-

em load A reinforcement learning approach is designed in [214] that

llows multimode base stations to autonomously steer their traffic

ows across different access technologies depending on the traffic

ype the usersrsquo QoS requirements the network load and the inter-

erence levels A number of studies have also tried to quantify the

14 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 9 WLAN operating in TVWS basic elements and functionalities

Fig 10 Hidden terminal problem in TVWS when the secondary network uses only in-

stantaneous channel sensing to decide whether or not a TV channel is occupied When

the STA is not able to detect the TV signal due to the presence of obstacles it may

initiate a transmission and create interference to the primary users

s

c

5

a

i

p

t

o

o

t

e

c

u

l

m

b

a

m

t

e

t

t

t

m

D

c

I

n

b

u

8

c

a

t

i

t

w

f

to discuss how to reuse these channels for unlicensed devicesrsquo com-

munications [130131] These potentially vacant channels in the VHF

and UHF bands are referred to as TVWS [132] and include spectrum

portions like the 470ndash790 MHz in Europe and non-contiguous 54ndash72

76ndash88 MHz 174ndash216 MHz 470ndash698 MHz and 698ndash806 MHz in USA

A snapshot of the TV spectrum occupancy in the city of Barcelona in

2012 is shown in [133]

The attractive characteristics of TVWS (not only for WLANs) in-

clude the ability to penetrate through walls and other obstacles much

more effectively than other widely used spectrum bands such as the

24 and 57 GHz ISM bands [134ndash137] This fact along with a progres-

sive spectrum scarcity in ISM bands has suggested the birth of the

IEEE 80211af amendment published in February 2014 [138] which

provides the IEEE 80211 operational characteristics for TVWS access

of unlicensed White Space Devices (WSD) A good summary of IEEE

80211af can be found in [137] The main advantage of operating IEEE

80211 WLANs in the TVWS comes from an increased coverage range

which can reach up to 1 km in rural areas and open fields [139]

and less energy needed to transmit However this comes at the cost

of an increased interference risk to other WSDs which creates co-

existence problems and demands new PHY and MAC layers to ef-

ficiently support channel access and operations preserving licensed

usersrsquo devices

WLANs operated in the TVWS could cover a number of interest-

ing and emerging use cases and scenarios eg Internet access in ru-

ral or sparsely populated areas Smart Grid sensor aggregation me-

tering and control Internet of Things advanced WLAN operations

and TVWS traffic offloading in indoor environments [140141] Nev-

ertheless the concept could be realised to create or extend com-

mercial or municipality WiFi services offered to citizens with cover-

age at the whole city level realised with reasonably limited network

infrastructures [142143]

IEEE 80211af will use the PHY and MAC layers derived from IEEE

80211ac It will adopt concepts such as the OFDM multi-user beam-

forming contiguous and non-contiguous channel bonding and packet

aggregation Among the mandatory and innovative behavioural and

operational parameters the most notable one is the channel acqui-

sition support realised through remote geolocation-based spectrum

allocation databases which maintain the channelsrsquo availability infor-

mation in any given area and time of day providing upon request the

list of free channels available for use

In the following section we highlight and summarise three main

novelties that IEEE 80211af introduces In Section 511 we explain

how the access infrastructure to the remote spectrum database is de-

signed and what the requirements are we discuss the coexistence is-

sues and methodologies adopted in IEEE 80211af and we also discuss

the novel concept of non-contiguous channel bonding The following

ections also contain an illustration of related works Open research

hallenges will be summarised in Section 512

11 Novel features

This section provides an illustration of novel features introduced

nd discussed in the path to IEEE 80211af and the directions provided

n related works and standardisation initiatives to resolve classical

roblems properly characterised in the new framework of TVWS

echnologies

Channel acquisition spectrum database and channel sensing Much

f the attention in operating in the TVWS is given to the protection

f the primary licensed users in the TVWS spectrum band In general

he primary users were considered as the Digital TV (DTV) broadcast-

rs and receivers In fact receivers would suffer the wasteful effect of

ollisions during the TV broadcast reception in the case of secondary

sersrsquo (SU) transmission interference (Fig 10)

It is a well-known problem that the definition of a coherent wire-

ess channel status (idlebusy) from the viewpoint of distributed pri-

ary users (eg DTV receivers) is made difficult by the difference

etween transmitters and receivers interpretation of signals in a vari-

ble time-space collision domain resulting in hiddenexposed ter-

inals concepts Solutions proposed to reduce the hiddenexposed

erminals included the RTSCTS mechanisms and beaconing How-

ver it is difficult to implement solutions that would effectively de-

ermine the channel status of TVWS channels based on localised dis-

ributed sensing activities performed by multiple secondary users in

he same collision domain On the other hand the use of RTSCTS

echanism would not be effective in this case because most of the

TV primary users (and specifically the receivers) were not originally

onceived to transmit signals or implement the RTSCTS handshake

n conclusion the actions to identify free (and busy) TVWS chan-

els should not be based on primary usersrsquo involvement but should

e almost totally implemented by specifically designed secondary

sersrsquo methodologies

There has been much debate on the methodology that IEEE

0211af secondary devices should adopt to get knowledge of free

hannels at a given time on a given target communication area in

n effort to guarantee verified optimised and reliable TVWS spec-

rum use The three main methodologies discussed include sens-

ng solutions geolocation databases (DB) and beaconing Moreover

wo approaches are possible in general distributed and centralised

ith their well known tradeoffs in terms of effectiveness required in-

rastructure deployment and coordination overhead The distributed

B Bellalta et al Computer Communications 75 (2016) 1ndash25 15

m

b

s

c

t

u

a

t

o

l

t

w

t

t

a

o

c

c

a

(

t

o

t

t

D

t

t

o

t

f

a

t

c

i

a

c

d

t

s

t

b

I

l

(

e

a

i

R

p

e

t

c

c

i

b

i

i

t

b

t

t

t

W

a

p

c

i

t

h

p

t

n

a

t

e

t

t

t

p

t

a

r

8

w

s

w

[

i

S

o

a

t

f

i

8

i

r

a

a

i

t

c

(

I

s

a

e

t

t

c

[

i

c

i

g

s

n

s

t

i

o

o

s

c

D

o

ethodology that is receiving more attention in the literature is

ased on distributed channel sensing solutions [127144145] These

olutions include the observation of spectrum use and possible

ooperative aggregation of spectrum sensing information to increase

he accuracy of detection and reduce the vulnerability of primary

sersrsquo transmissions The spectrum sensing capability is required in

ll the secondary devices with a minus114 dBm sensitivity Dynamic

ransmission power control must be provided and the upper limit

n emissions is 100 mW EIRP (20 dBm) for portable devices (further

imited in the case of adjacent channel use to reduce out-of-band in-

erference) However database (DB) coordination of TVWS combined

ith spectrum sensing is considered the most promising and effec-

ive technique compared with spectrum sensing alone The DB spec-

rum information has more chances to be effective (ie identifying

ll free channels opportunities) and reliable (ie not prone to attacks

r erroneous interpretation of channels status) In fact regulators de-

ided to push for the conservative solution of a remotely accessible

entralised spectrum DB that maintains the information on the avail-

bility of all TVWS channels at any given point in time and location

with a target accuracy around 50 m) under their responsibility For

his reason the spectrum databases have become a mandatory part

f many spectrum sharing systems and the dominant technical solu-

ion to support TV white spaces

The IEEE 80211af standard adheres to this vision and also has

he role of a common regulatory framework for different spectrum

B implementations defining a generalised coordination architec-

ure protocols and interfaces for spectrum queries and local spec-

rum information management The protocol explicitly considers out

f scope the communication protocol and technology adopted in

he background leaving the freedom to those players responsible

or the deployment to select any suitable Internet-based and local

ccess network technology Secondary devices that need to access

he TVWS spectrum should get information on which channels they

an effectively use and which parameters to adopt without impact-

ng primary usersrsquo transmissions (and receptions) by querying the

vailable spectrum DB The query must contain the transmission

haracteristics and accurate geographical position of the secondary

evice which can be determined by means of a geo-positioning sys-

em (GPS) regularly updated in the case of movement (or manually

et for static devices) The IEEE 80211af reference system architec-

ure for DB spectrum access is composed of multiple Geolocation-

ased Spectrum information Databases (GDBs) entities connected via

nternet to Registered Location Secure Servers (RLSS) RLSSs work as

ocal proxies of the GDB for a localised group of Basic Service Sets

BSS) RLSS are connected via a secure protocol architecture with the

quivalent of multiple Access Points (AP) in different BSS realising

trusted DB infrastructure APs locally coordinate the exchange of

nformation and channel access management between the GDB (via

LSS) and the secondary usersrsquo end stations (STA)

Mode I devices are those under the control of a device that em-

loys geo-location database access while Mode II devices are those

mploying geo-location database access by themselves The informa-

ion exchange provided between the GDB and a secondary user STA

an be provided in both an open-loop (eg adopted by FCC) and a

losed-loop (eg adopted by ETSI) implementation In an open loop

mplementation daily spectrum availability information is provided

y the GDB and no feedback on the spectrum information received

s provided by STAs thus the system approach for spectrum access

s much more conservative leading to low channelsrsquo utilisation po-

ential In a closed loop implementation the STA can provide feed-

ack to the GDB and there are more communication overheads due

o the high granularity of updates however the system is more effec-

ive and reliable in the exploitation of TVWS on behalf of STAs The

ypical information provided by the GDB includes (1) the updated

hite Space Maps (WSM) of frequencies allowed for secondary use

t the timespace of the querying STA and (2) the device-dependent

ower limitations for transmission (in general conservative and ac-

urate enough to avoid relevant interference effects on primary users

dentified in the area)

The basic IEEE 80211af mechanisms regulating the communica-

ion between STA and GDB can be found in [137138]

In parallel with IEEE 80211af standards such as the IEEE 19006

ave been created that consider interfaces and data structures sup-

orting spectrum sensing information exchange applicable to spec-

rum sensing (and particularly distributed spectrum sensing) sce-

arios In particular in October 2014 the IEEE 19006 WG initiated

project for a new standard called IEEE 19006b [146] concerning

he use of spectrum sensing information to support and optimise the

ffectiveness reliability and robustness of spectrum database solu-

ions The aim is to enhance the performance and capabilities of spec-

rum databases through the use of spectrum sensing information

Co-existence Support for co-existence mechanisms so that mul-

iple technologies can effectively utilise the TVWS spectrum is im-

ortant Self co-existence between network devices of a common

echnology (eg deployed by different operators in the same area)

nd co-existence among different technologies are relevant topic of

esearch for Cognitive Radio (CR) systems and specifically for IEEE

0211af Many solutions appeared on the research scene but no one

as so far finalised as the target solution for IEEE 80211af Specific

tandardisation has been started to regulate coexistence between

ireless standards of unlicensed devices including the IEEE 802191

147] The purpose of the IEEE 802191 standard is to enable the fam-

ly of IEEE 802 Wireless Standards to most effectively use TV White

pace by providing standard coexistence methods among dissimilar

r independently operated TVWS devices Early examples of gener-

lised coexistence mechanisms included Dynamic Frequency Selec-

ion (DFS) Transmission Power Control (TPC) listen before talk (eg

or contention based IEEE 80211 80215) time division multiplex-

ng (also among different techniques such as the IEEE 80216 80220

0222) and Message-based Spectrum Contention (that is beacon-

ng messages that carry coexistence information) Opportune met-

ics must be defined to assess the measurable coexistence achieved

mong different technologies as an example the hidden node prob-

bility for a target scenario or the estimate of percentage variation

n normalised network throughput and latency (before and during

he SU transmissions) On the other hand a centralised coexistence

ontrol mechanism could be effectively realised by a central manager

or coexistence-DB like the GDB) in critical scenarios To this end

EEE 19004 [148] (a standard for heterogeneous networks in dynamic

pectrum context part of IEEE Standards Coordinating Committee 41)

ims to standardise the overall system architecture and information

xchange between the network and mobile devices which will allow

hese elements to optimally choose from available radio resources

There are three possible classification of co-existence archi-

ectures and inter-network coordination channels for CR systems

entralised coordinated and autonomous co-existence mechanisms

149] In centralised co-existence schemes the co-existence is admin-

stered by a central entity (eg like in IEEE 802191) This solution

ould be applied in both homogeneous IEEE 80211af systems (eg

ncarnated by the centralised spectrum DB entity) and in hetero-

eneous inter-networks without requiring modifications to existing

tandards under the assumption that all the involved co-existing

etwork entities would adhere to the same centralised coordination

cheme and inter-network coordination protocol On the other hand

his assumption could be hard to satisfy in some practical scenar-

os The centralised schemes are effective in providing minimisation

f inter-network interference based on the availability and quality

f centralised co-existence information In coordinated co-existence

chemes the centralised entity could be present but not taking de-

isions in general The central entity could provide a co-existence

B with required information which can be used via in-band and

ut-band signalling for co-existence decisions taken by the cluster

16 B Bellalta et al Computer Communications 75 (2016) 1ndash25

c

e

n

c

o

s

i

d

a

S

8

n

a

b

o

b

i

t

a

8

b

[

t

i

n

b

5

t

a

c

b

s

a

c

n

s

p

i

c

l

r

c

f

a

v

t

a

e

c

t

Q

a

b

s

i

c

u

n

c

i

heads of many co-existing inter-networks for implementing a proper

common coordination protocol (eg out-band busy tone signalling)

These are often hybrid solutions realising a compromise between ef-

fectiveness and suitability In autonomous co-existence schemes all the

decisions and coordination are implemented in a distributed way

by the involved network entities In these schemes many policies

can be adopted to realise a sufficient level of co-existence under a

best effort approach Solutions that could be implemented include

the use of busy tones beaconing and other signalling protocols in

dedicated control channels dynamic distributed frequency selection

schemes listen before transmit token-based and dynamic reserva-

tion schemes All the above mentioned solutions have goods and

bads and a general illustration can be found in [149] Other proac-

tive co-existence techniques try to early detect and recovermitigate

co-existence issues with relaxed inter-network coordination eg re-

alised via spectrum sensing and interference avoidancesuppression

techniques

It must be clear that different aims exist regarding the co-

existence for primary TVWS users protection against secondary

users and for secondary devices mutual interference avoidance

Many co-existence problems have been analysed in the literature

for TVWS technologies in particular between the primary (licensed)

and secondary (unlicensed) devices The coexistence management in

TVWS is complex because primary users of DVB transmissions are in-

tended to be the pure receivers rather than the broadcasters In other

words potential protected users could be everywhere and the hid-

den terminal problem could arise for these systems The co-existence

problem is also complicated by many factors that cause asymme-

try and dynamicity in the TVWS such as the mobility variable den-

sity power asymmetry and heterogeneous MACPHY layers TVWS

co-existence problems originate between the IEEE 80211af technolo-

gies and the IEEE 80222 technologies for wireless regional area net-

works (WRAN) Results of analysis of TVWS usage in Europe show

that white spaces are typically present and fragmented They are typ-

ically more abundant in rural areas where larger contiguous blocks

of unused channels are available due to broadcast network planning

giving priorities linked to population density However the exploita-

tion of TVWS in urban areas is possible eg some recent research

was realised under the assumption of exploiting shadowing effects

created in the communication environment (eg by buildings obsta-

cles) in favour of frequency reuse [150]

Since different standards for opportunistic communication in the

TV White Spaces have now been published such as IEEE 80222 [151]

IEEE 802154m [152] Weightless [153] and of course IEEE 80211af

improved methods to guarantee the coexistence of different devices

operating on several protocols in the same bands must be deployed

We highlight here some work eg [150154ndash157] that focuses on

the coexistence between different technologies operating in the same

bands In particular [155] studies the performance degradation of

IEEE 80222 when an IEEE 80211af network is operated in the same

area The problem becomes even worse when the IEEE 80211af net-

work is located near 80222 user equipment causing both networks

to perceive a strong interference from each other One proposed solu-

tion is the Coexistence Beacon Protocol [158] studied for 80222 net-

works which foresees the exchange of a periodic beacon to identify

the neighbouring and possibly interfering networks

For what concerns the co-existence problems research has ad-

dressed other important related aspects of new WLANs on TV White

Spaces For instance the higher coverage range makes them attrac-

tive for applications in the smart grid and also for M2M on TVWS

[134150159] Here the increased range compared to standard tech-

nologies in the ISM bands can have beneficial effects for indoor

mobile devices [150] and M2M [159] Specifically indoor devices

communicating on TV White Spaces offer better propagation char-

acteristics and penetration through obstacles making it easier to

realise scalable home connectivity [135] although interference and

o-existence problems could be exacerbated and must be resolved

g see [139] However studies in the literature show how the sig-

al coming from an indoor transmitter on TV White Spaces remains

onstrained inside the house making it difficult to deploy indoor-to-

utdoor networks On the other hand this shadowing limit can be

een as an opportunity lowering the interference outside the build-

ng in which the TVWS network is locally operated [134150] Un-

er this approach HDTV streaming has recently been considered as

possible use case of WLAN networks operating in the TV White

pace [160]

Non-contiguous channel bonding A relevant novel feature of IEEE

0211af is the potential for contiguous and non-contiguous chan-

el bonding which permits aggregating basic channels (also non-

djacent ones) and leveraging the possible large frequency spread

etween multiple available channels Due to the rather static nature

f primary DTV transmissions where a busy channel is unlikely to

ecome free in the near future and state changes are coarse-grained

n general it is crucial to exploit the time-locality effect and exploit

he maximum physical channel availability that could be aggregated

t any given location With the methodology inherited from IEEE

0211ac IEEE 80211af is capable of bonding together two up to four

asic channels grouped in up to two different non-contiguous chunks

137] The spectrum bandwidth of a DVB-T basic channel can be ei-

her 6 7 or 8 MHz depending on the country in which the service

s operated As an example this creates a 144 to 168 OFDM chan-

elsrsquo bandwidth potential when up to four 6ndash7ndash8 MHz channels are

onded [137]

12 Open challenges

The deployment of WLANs in the TVWS at their maximum poten-

ial still requires the resolution of open problems Generally speaking

s mentioned in previous sections the challenges for such networks

an be divided into three main categories (1) spectrum sharing

etween opportunistic (secondary) and heterogeneous devices (2)

pectrum sensingmanagement and maximum exploitation of avail-

ble spectrum (potentially enabling the spectrum-on-demand con-

ept) and (3) co-existence and interference mitigation to the primary

etwork (primary user protection) and secondary networks [154]

The overall aim of spectrum sharing techniques is to maximise

eparation to avoid overlapping or contiguous channels operations

rovided that sufficient channels are available in the TVWS spectrum

n a given spacetime scenario [154] In general important research

ontributions have to be realised for the design of proper spectrum al-

ocation methodologies and techniques satisfying multi-factorial QoS

equirements at the system-level and the user-level Novel ideas in-

lude the spectrum sharing and spectrum sensing techniques being

urther divided into cooperative and non-cooperative

Non-cooperative solutions attempt to realise spectrum sharing

nd sensing on a local basis without direct cooperation between de-

ices Examples include the ldquolisten-before-transmitrdquo approach and

ransmission power control to limit the interference and spectrum

llocation policies that are based on local node feedback only Coop-

rative solutions rely on the existence of a common communication

hannel through which devices can tentatively agree on the spec-

rum allocation that provides the desired spectrum separation and

oS Major issues include the definition of a methodology to identify

common channel (or multiple channels mutually shared in space

y pairs of heterogeneous and distributed devices) and to efficiently

hare common information For the access to the spectrum DBs the

dentification of proper common access channels is still under dis-

ussion out-of-TVWS-band cellular communications can be widely

sed given their high coverage also indoor

Trying to avoid the use of complex and resource-hungry coordi-

ation schemes and dedicated control channels rendezvous proto-

ols play a significant role both for co-existence and for establish-

ng dynamic communication channels based on spectrum sharing

B Bellalta et al Computer Communications 75 (2016) 1ndash25 17

R

n

c

t

l

o

m

e

n

i

n

w

n

a

u

s

p

b

s

m

s

a

d

t

b

a

a

m

a

t

u

8

n

t

d

d

m

f

t

n

d

n

e

r

a

h

b

f

a

c

t

u

i

f

t

d

t

s

i

f

I

l

i

a

n

c

d

I

n

o

t

i

t

e

w

t

c

m

t

a

i

t

q

b

c

a

i

l

i

a

v

t

a

a

c

t

i

t

b

t

t

a

D

p

c

s

o

q

r

p

a

g

t

e

u

d

t

m

n

t

s

f

i

1

8

f

T

endezvous is considered a fundamental problem in generalised CR

etworks at the basis of many communication processes eg in-

luding neighbour discovery routing and broadcast Rendezvous pro-

ocols attempt to establish a new link for communication on a se-

ected and agreed frequency band (channel) identified within a set

f available resources (channels) thus creating the basis for com-

unication between two or more SUs This problem is often exac-

rbated by the high (and unknown) numbers of contending SUs high

umber of available channels and their ambiguous boundaries and

dentifiers and tough co-existence requirements satisfaction in cog-

itive radio systems [161] Preliminary approaches for rendezvous

ere based on the existence of a dedicated Common Control Chan-

el (CCC) agreed among all the SUs or the existence of a centralised

gent (decision maker) assigning channels to any pairs of requesting

sers This approach could be considered in the case of IEEE 80211af

olutions based on the spectrum DB implementation however the

roblem is complicated by the fact that the common channel must

e identified and made available for all the requesting SUs at the

ame time even when these are spread over multiple collision do-

ains in space On the other hand the scalability and reliability is-

ues which are caused by the congestion of CCCs and the vulner-

bility risks of centralised DBs motivate the research of alternative

istributed solutions called blind rendezvous protocols [162] Dis-

ributed solutions have been mainly based on beaconing andor slot-

ased channel hopping algorithms exploiting randomisation sensing

nd discoverysynchronisation messages spread in slotted time over

set of candidate available channels [161] These protocols aim to

inimise the convergence time to maximise the use of resources

nd to avoid collisions among SUs by identifying and agreeing on

he use of a given common channel among N non-overlapping and

nambiguously labelled channels The main complication in IEEE

0211af systems is given by the wide availability of many different

on-contiguous spectrum bands which cannot be uniformly quan-

ised in single channels and unambiguously labelled in space in or-

er to realise a common reference domain for all the IEEE 80211af

evices In fact the knowledge of the number of channels their com-

on labels the unique IDs of SUs and the number of SUs contending

or the channel assignment are the key factors determining the effec-

iveness and good properties of blind rendezvous protocols Recently

ew classes of rendezvous protocols have been proposed which are

istributed blind and oblivious Oblivious means that available chan-

els may be identified with different labels on behalf of the differ-

nt involved SUs A number of these distributed blind and oblivious

endezvous protocols have been proposed and have been referenced

nd analysed in [162] To conclude solutions for dynamic spectrum

andoff issues (eg in case PUs suddenly appear in a region) must

e resolved in advance to minimise the latency of re-establishing a

unctional communication channel eg by implementing a proactive

lternative rendezvous channel selection in background to ongoing

ommunication processes [163]

Another challenge is to provide cooperative techniques in a dis-

ributed vs centralised implementation by analysing their convergence

nder variable conditions overheads and tradeoffs In some cases as

n vehicular networks the exploitation of scenario characteristics and

actors such as the constrained mobility could help to realise effec-

ive dissemination of sensing information over an extended and pre-

ictable horizon under a cooperative approach [164ndash166] Coopera-

ion could be helpful in the realisation of the cooperative spectrum

ensing in combination with the Geolocation DB approach Another

nteresting direction of research is the investigation of mutual ef-

ects of coexisting cooperation-based vs non-cooperative techniques

n cooperation-based devices the exchange of information could al-

ow a quick convergence to a solution and avoidance of further

nterference [145] Certainly cooperative techniques have desirable

dvantages over non-cooperative ones However tight time synchro-

isation and high coordination overheads are required The common

hannel for cooperation purposes could become a bottleneck and re-

uce the potential advantages of dynamic spectrum allocation The

EEE 80219 foresees the presence of a shared common control chan-

el (CCC) to which devices need to tune in order to gather the state

f the network Solutions such as [141] propose the use of TVWS for

he CCC

A promising direction for research is the adoption of cluster-

ng schemes for secondary devices associated with a hybrid dis-

ributedcooperative vs centralisednon-cooperative approach in an

ffort to reduce overheads while maximising advantages In this

ay secondary nodes could implement cooperation for sensing func-

ions using properly identified common channels and delegate the

entralised decisions and spectrum-DB management to well instru-

ented cluster-leader nodes This concept is incarnated to some ex-

ent by RLSS in IEEE 80211af More recently there have been propos-

ls to build distributed databases that refresh their contents period-

cally by querying the remote spectrum database [167] Here Mas-

er devices (according to Ofcom terminology) periodically cache the

uery replies from the spectrum database in order to reply to and

roadcast the spectrum availability to neighbour Slave devices

Another challenging direction for research is the definition of ac-

urate models and efficient simulation tools enabling dynamic spectrum

nalysis for both rural and urban areas Recent developments of dig-

tal maps simulation tools and propagation models theoretically al-

ow improving the capability of predicting radio propagation effects

n complex scenarios with an acceptable computation time This en-

bling technology could be used in parallel to spectrum DBs to pro-

ide more accurate forecasting of dynamic frequency allocation in

imespace scenarios Another issue to be taken into account is the

ccuracy of the remote spectrum database regarding the channel

vailability estimation Several works have already shown the inac-

uracy of propagation models due to the complexity of parameters

o be considered and differences between modelled and real scenar-

os [132134168169] Better propagation models will make it possible

o more efficiently estimate the interference between devices and

uild Radio Environment Maps (REM) or White Space Maps (WSM)

hat will possibly lead to a more accurate sharing of the radio spec-

rum [144] To this end current research could focus on how to lever-

ge simulation models in order to build more accurate frequency

Bs

In general the adoption of remote spectrum DBs methodology

ushed by national regulators has made the spectrum sensing pro-

ess optional and conservatively realised (eg power control for

ensing-only devices is limited to below 17 dBm EIRP) Thus much

f the research has focused on techniques and solutions to successfully

uery the remote database also in challenging environments such as

ural areas and indoor However since the queries must provide the

osition of the mobile device the accuracy of positioning for mobile

nd handheld devices still needs more appropriate solutions Trian-

ulation can be used to estimate the position of mobile devices with

he accuracy needed by the current regulations outdoor (50 m) How-

ver a low GPS accuracy is possible indoors [170] Certified man-

al positioning can be provided by technicians for static and indoor

evices although limiting mobility Alternatives to costly infrastruc-

ures for positioning (eg based on short range beaconing devices)

ust be deployed in indoor scenarios

Finally regarding spectrum sensing and interference mitigation

ew challenges come from the heterogeneity and different charac-

eristics that networks operating in the TV White Spaces can have

uch as different transmitting power (4 W for static devices 100 mW

or mobile and portable devices and 40 mW for communication

n channels adjacent to occupied ones) different bandwidths (5

0 15 and 20 MHz are currently foreseen for IEEE 80211af IEEE

02154m can have narrower bandwidths or wider ones) and dif-

erent medium access schemes (protocols that use either CSMA or

DMA are required to co-exist in the TV White Spaces) With all this

18 B Bellalta et al Computer Communications 75 (2016) 1ndash25

d

a

T

i

a

c

t

l

b

o

t

a

d

p

a

r

p

n

a

m

t

p

p

d

d

o

t

i

D

w

p

i

a

a

d

b

fi

d

o

t

l

b

t

b

a

i

e

o

o

c

t

t

w

A

c

M

s

c

l

O

w

g

v

t

in place the problem of determining whether another secondary de-

vice is currently transmitting in the same channel is still an open

issue

Spectrum sensing is a very hot topic in the literature with many

proposals that leverage the cooperation between devices to raise the

accuracy of the detection process eg [144171] The current litera-

ture also offers room for proposals that suggest exploiting sensing to

build more accurate spectrum DBs [164172173] Lastly regarding the

interference mitigation to the primary network regulators provide a

conservative approach based on a remote spectrum database to avoid

high risk for the primary users However in particular for the indoor

scenario the concept of secondary usersrsquo transmission in TV Grey

Spaces has been proposed TV Grey Space identifies busy channels

that are formally used by primary users in the area of interest (eg

at the rooftop level of a building for DVB-T) but which could be con-

sidered usable indoors (eg at the basement level) without causing

effective interference to the primary users [134150174] This concept

is based on the generalised spectrum underlay paradigm [175] The

remaining research challenge is to provide TV Grey space access guar-

anteeing a sufficient amount of protection to the primary user while

providing a satisfactory QoS to opportunistic devices eg through

the use of customised and validated propagation models for indoor

TVWS networks and feedback mechanisms to constantly monitor the

interference generated to the primary network

6 Emerging new trends and technologies

In this last section we review three emerging trends that in our

opinion will have a large influence in the conception of future WLANs

as they change the way WLAN protocols and functionalities are de-

veloped implemented tested and integrated with other wireless

networks

61 Programmable Wireless LANs

Especially in the enterprise environment WLAN deployments

need to support a wide range of functionalities and services This is

intrinsically difficult because of the large number of APs that must be

managed which calls for scalable solutions Typical services include

channel assignment load balancing among APs authentication au-

thorisation and accounting (AAA) policy management support for

client mobility and interference coordination Another problem is the

fact that WLAN clients autonomously take several decisions such as

which APs to associate with when to hand-over etc Therefore sup-

porting roaming clients requires the management of a large number

of association states across several APs which is a challenge if sup-

port for real-time hand-over is desired Typically such management

schemes are centralised and most of them are proprietary such as

WLAN controller solutions from Aruba [176] or Cisco [177] although

the IEEE 80211u amendment has been released to allow mobile users

to seamless roam between WiFi networks with automatic authenti-

cation and handoff [178] For example Dyson [179] enables STAs to

send information such as radio channel conditions to a centralised

controller based on a custom API (eg based on Python) As the con-

troller has a centralised view of the network it can enforce a rich

set of policies to control the network also using historical informa-

tion A demo system has been implemented along with applications

such as airtime reservations for specific clients or optimised hand-

offs However Dyson requires STAs to be modified in order to use

those new services offered by the centralised controller TRANTOR

[180] is another example of a centralised management system that

requires changes of clients in order for the infrastructure to gather

information from them in terms of eg interference measurements

In addition control commands enable the infrastructure to exercise

control such as modifying the transmit power or influencing the as-

sociation procedure Clients still use standard CSMA MAC layer for

ata transmission In contrast to DenseAP [181] or DIRAC [182] which

re also based on a centralised controller but do not modify clients

RANTOR exercises much more control on clients by the use of a ded-

cated API which enables a significantly higher gain in coordination

nd thus capacity CENTAUR [183] is another example of a centralised

ontroller that centrally schedules hidden and exposed terminals in

he downlink while using standard CSMA MAC for uplink traffic and

egacy downlink traffic To achieve good performance it uses a fixed

ack-off packet staggering techniques and a hybrid data path that

nly schedules downlink transmissions towards hidden and exposed

erminals centrally All other traffic is sent using standard DCF with

standard back-off procedure In contrast to previous approaches it

oes not require any modifications to the clients but does require data

lane centralisation and it remains questionable how scalable such

solution is for large WLAN deployments with high PHY layer data

ates beyond 100 Mbps However new approaches for fast data plane

rocessing that are currently explored in the virtualisation commu-

ity which move all the packet processing into user space may be

n option to significantly increase speed Designing such manage-

ent systems raises several interesting research questions (a) Cen-

ralised infrastructures may incur high latency but are more simple to

rogram and maintain while designing a completely distributed ap-

roach that operates close to WLAN devices may result in significant

istributed coordination and consistency problems (b) what level of

istributed control is required in order to support the flexibility needs

f future WiFi based networks supporting a high level of mobility

Traditionally WLAN APs are built on proprietary operating sys-

ems that are tightly coupled with the hardware This design makes

t hard to create new applications on top of such networking devices

espite the fact that protocols and mechanisms are available that

ould greatly improve the utility of existing networks those new

rotocols are not deployed because the closed system design makes

t very difficult to extend their functionality An important aspect for

ll such new approaches is therefore how to provide open interfaces

nd open source to speed up innovation As an example Linux-based

evices are completely open source but in order to increase flexi-

ility much more work is needed in the area of open drivers and

rmware In fact the Atheros based ath9k driver has been the main

river of innovation in the open source community because it is the

nly driver that interworks with open firmware In general solutions

hat modify the OS driver focus mainly on programmable network

evel solutions for WLANs such as channel switching or handovers

etween APs While the ath9k driver is an excellent example in how

o speed up innovation it is still very cumbersome to support flexi-

le MAC engine reprogrammability with ath9k In contrast MAClets

nd Wireless MAC processors (WMP) [184] allow a much more flex-

ble reprogramming of MAC functionalities A MAC processor is an

ntity able to execute general MAC commands that specify the MAC

perations through a software-defined state machine The behaviour

f the MAC protocol can therefore be updated at run-time by simply

hanging the sequence in which those commands are executed (ie

he MAClets) As a proof-of-concept the authors of [184] implement

he proposed MAC processor solution in a commodity WLAN hard-

are card extending the basic DCF in three directions piggy-backing

CKs a pseudo TDMA and the use of multiple channels In [185] a

ontrol framework for this WMP system is also proposed to support

AClet code mobility ie for moving loading and activating MAC

oftware programs embedded into ordinary data packets (akin to the

apsule model of traditional active networks)

The difficulty in re-programming networking hardware has also

ed to the concept of Software Defined Networks (SDNs) based on the

penFlow protocol [186] The main idea of SDNs is to extend net-

orking devices with standardised APIs that allow third-party pro-

rammers to flexibly control the data path In addition SDNs pro-

ide higher level abstractions to network designers and programmers

hrough the use of a centralised control plane offered by a network

B Bellalta et al Computer Communications 75 (2016) 1ndash25 19

Fig 11 The CLOUDMAC architecture VAP stands for Virtual Access Point and WT

stands for Wireless Termination

c

a

c

t

m

W

n

t

t

c

p

t

u

h

a

i

e

s

o

c

a

a

f

T

r

s

a

s

p

o

M

C

i

T

t

fl

A

a

a

i

c

t

s

t

e

I

t

w

H

e

b

I

l

d

o

6

r

W

s

l

a

e

e

a

s

b

c

t

(

S

p

c

F

B

t

p

T

e

t

s

u

m

i

e

o

a

A

p

d

t

t

n

p

i

t

t

f

i

ontroller such as NOX [187] which allows reuse of components such

s topology discovery or network access control for different appli-

ations The SDN concept has recently been applied to WLAN archi-

ecture in order to enable fine grained control over mobility manage-

ent and data forwarding focusing on a programmable enterprise

LAN architecture An important part of the SDN architecture is the

etwork controller which provides a centralised view of (parts of)

he SDN enabled network and uses the OpenFlow protocol to install

he forwarding rules on SDN-enabled devices (routers switches ac-

ess points cellular base stations etc)

ODIN [188] is designed to support programmability in enter-

rise WLAN architecture by separating the association state from

he physical AP They implement Light Virtual Access Points (LVAP)

sing a Split-MAC approach where the infrastructure controls the

andover procedure among different WLAN APs By managing

ssociations through SDN controllers ODIN enables proactive mobil-

ty management and load balancing within the SDN enabled WLAN

nterprise network without the need for changes in the client WLAN

tack or IEEE 80211 MAC layer While ODIN requires agents to reside

n the APs that communicate with the Odin Master within the SDN

ontroller CLOUDMAC [189] is based purely on the concept of SDN

nd virtual APs Similar to ODIN CLOUDMAC implements a Split-MAC

pproach but in addition enables the processing of WLAN MAC layer

rames within a co-located Cloud using so-called Virtual APs (VAPs)

he physical APs in CLOUDMAC are lightweight WLAN APs that are

esponsible only for sending out IEEE 80211 based MAC layer ACKs to

tandard WLAN clients and tunnel WLAN MAC layer frames through

n SDN-enabled enterprise WLAN towards the VAPs As association

tates are kept in the VAPs fast mobility is supported using sim-

le OpenFlow forwarding rules Because of the additional processing

f IEEE 80211 WLAN MAC frames in the co-located Cloud CLOUD-

AC has higher latency than a standard WLAN deployment However

LOUDMAC offers a Webservice based API to third party applications

n order to program the enterprise WLAN and enable new services

he architecture of CLOUDMAC is depicted in Fig 11 It has been ex-

ended in [190] to support QoS management and in [191] to support

exible MAC management frame prioritisation based on IEEE 80211

system based on OpenFlow has been recently proposed in [192] to

llow the station to be associated with multiple APs simultaneously

nd to switch between APs with low overhead

An important aspect to consider in order to enable programmabil-

ty is backwards compatibility Several approaches (eg [180]) require

lients to be modified to utilise the features provided This is difficult

o do in practice because it requires changes in the operating system

oftware of all clients If not all clients can utilise those APIs it is ques-

ionable how usable the new architecture will be and how much ben-

fit in terms of aggregate performance such architecture will allow

n contrast the SDN based approaches are interesting in the sense

hat they do not require changes in the client WLAN stack and work

ith the standard IEEE 80211 MAC layer deployed within the clients

owever it remains questionable how scalable such solutions are For

xample an interesting open research topic is to evaluate the scala-

ility of approaches such as CLOUDMAC [189] In addition processing

EEE 80211 MAC frames within co-located private Cloud requires low

atency support from local Cloud solutions (such as OpenStack) in or-

er to reduce the IEEE 80211 MAC processing time which is an area

f active research

2 Prototyping and testing IEEE 80211 enhancements

Most of the new proposals for next-generation WLANs are cur-

ently only evaluated using mathematical analysis and simulation

hile both analysis and simulation are necessary to characterise and

tudy those enhancements in the initial design phase or to consider

arge-scale scenarios it is difficult to consider all practical aspects of

real-world scenario This can sometimes cause significant differ-

nces between what simulations and real experiments show How-

ver real experiments are challenging because of the high complexity

nd costs of building the new hardware and software for each specific

olution to test

To mitigate implementation complexity there are several flexi-

le hardware platforms where low-level MAC mechanisms can be

ompletely implemented in software Examples based on FPGAs are

he USRP (Universal Software Radio Peripheral) [193] and the WARP

Wireless Open-Access Research Platform) [194] As an alternative

ORA (Microsoft Research Software Radio) [195] works on general

urpose computers by taking some advantage from modern multiple

ore systems

A different approach is provided by OpenFWWF [196] Open-

WWF provides an open CSMACA firmware for specific models of

roadcom chipsets so the resulting firmware can be uploaded and

ested in real commercial hardware OpenFWWF implements a sim-

le State Machine (SM) for controlling the hardware in real time

he SM evolution is driven by a main loop that reacts to events by

xecuting specific handlers When a packet originally prepared by

he Linux kernel is ready in the NIC memory handler Packet_Ready

ets up the radio hardware according to the packet metadata sched-

les the transmission and jumps back to the main loop The Trans-

ission Engine (TXE) then takes care of accessing the channel ie

t decrements the back-off counter according to the DCF rules and

ventually starts the actual transmission This triggers the execution

f the TX_frame_now event that prepares the ACK time-out clock

nd finalises the MAC header If the ACK-frame is received or if the

CK time-out expires and the maximum number of attempts for this

acket is reached handler Update_Params resets the contention win-

ow to its minimum value or otherwise doubles it Finally it loads

he back-off counter with a fresh value

In the following we will review a list of selected papers that use

he OpenFWWF firmware and the WARP or USRP platforms to test

ew proposals for WLANs in real scenarios as examples A first im-

lementation of the groupcast mechanisms defined in IEEE 80211aa

s presented in [197] The authors modify the OpenFWWF firmware

o include those functionalities and evaluate them using a 30 STAs

estbed A second example of the use of OpenFWWF is the collision-

ree MAC protocol presented in [198] In [198] the authors design and

mplement in OpenFWWF a MAC protocol that is able to achieve a

20 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 12 LTE to WIFI offloading

a

n

c

t

[

d

a

W

v

t

t

o

b

f

S

g

l

i

I

s

I

b

n

s

m

a

p

d

c

N

o

v

t

t

n

l

i

a

t

a

collision-free operation by waiting a deterministic time after success-

ful transmissions before to start a new transmission This work shows

the experimental performance of a collision-free MAC (CF-MAC) pro-

tocol for WLANs using commercial hardware The results show that

the proposed CF-MAC protocol leads to a better distribution of the

available bandwidth among users a higher throughput and lower

losses than DCF

In [199] the authors use the WARP platform to test a variant of the

DCF ndash called IEEE 80211ec ndash that substitutes control packets such as

the RTS CTS and ACK with short detectable sequences Since those se-

quences are shorter than the control packets and can be detected cor-

rectly even at lower SNRs values a significant gain in performance is

achieved MIDAS (Multiple-Input Distributed Antenna Systems) [200]

implementation using the WARP platform shows the benefits of dis-

tributing the antennas over the area to cover instead of co-locating

them at the AP in terms of capacity when MU-MIMO is employed

The authors also propose a new MAC protocol to benefit from the

spatial reuse that the DAS allows taking as a basis the IEEE 80211ac

Experimental results show that their proposal is able to achieve up

to 200 gains versus the traditional approach where all antennas are

co-located at the AP

The USRP platform has recently been proposed to develop a first

implementation of IEEE 80211ah in order to obtain a preliminary ex-

perimental performance assessment of such a technology because

no commercial off-the-shelf hardware is yet available as the amend-

ment is still in progress [201] The USRP platform is also used in

[202] to evaluate TIMO (Technology Independent Multi-Output) a so-

lution to deal with high-power non-IEEE 80211 interferers in ISM

bands

To conclude this section it is worth mentioning here also to

MAClets and WMP as given such flexibility to implement and dis-

tribute new MAC protocols and other IEEE 80211 functionalities we

believe that if a use-friendly implementation of such a MAC proces-

sors framework is provided it would significantly contribute to the

development and testing of new MAC enhancements by the research

community

63 CellularWLAN interworking

Public hotspots that offer Internet access over a WLAN using IEEE

80211 technology are now nearly ubiquitous It is forecasted that the

cumulative installed base of WiFi hotspots worldwide will amount

to 551 million by 2018 excluding private hotspots (eg WiFi access

points deployed at home) [203] The sharp increase in the availabil-

ity of public WiFi was initially perceived by mobile cellular operators

as a threat due to the additional competition from wireline Internet

service providers or emerging crowdsourced WiFi networks such as

FON5 However as cellular operators are fighting to cope with the ex-

plosion of mobile data traffic created by the rising use of multimedia

content traffic over mobile devices [15] they are also starting to use

WLANs based on the IEEE 80211 technology to offload data from their

core and access networks In general mobile data offloading refers to

the use of complementary network technologies (in licensed or un-

licensed spectrum) for delivering data originally targeted to cellular

networks Intuitively the simplest type of offloading consists of ex-

ploiting connectivity to existing co-located WiFi networks and trans-

ferring data without any delay (Fig 12) Thus this offloading tech-

nique is known as on-the-spot offloading As a consequence of this

new trend the seamless integration of cellular (eg 3GLTE) and WiFi

technologies has attracted significant research interest in recent years

(see [204] for a survey) a few solutions have already been standard-

ised [205206] and roaming between cellular and WiFi is becoming

increasingly transparent to end users CellularWLAN interworking is

5 httpscorpfoncomen

fl

t

f

lso fostered by the support in the evolving 4G standards of heteroge-

eous network deployments (HetNets) in which the existing macro

ells are complemented with a number of small low-power base sta-

ions with the goal of increasing capacity in highly congested areas

207208] It is envisaged that small cells will be based on 4G stan-

ards (eg pico and femto cells) as well as IEEE 80211 technologies

nd multimode base stations that work simultaneously with LTE and

iFi are already entering the market

It is important to point out that LTE standards already support a

ariety of mechanisms that enable data offloading However most of

he existing solutions such as Local IP Access (LIPA) and Selected In-

ernet IP Traffic Offload (SIPTO) focus on data offloading at the core

f the network [205] For instance LIPA allows an IP-enabled mo-

ile device to transfer data to another device in the same pico or

emto cell without passing through the cellular access network while

IPTO enables the routing of selected IP data flows through different

ateways The only LTE offloading mechanism that supports seam-

ess interworking with IEEE 80211-based WLANs is IP Flow Mobil-

ty and Seamless Offload (IFOM) Specifically IFOM relies on Mobile

Pv6 technologies to allow a user terminal to simultaneously route

elected IP flows over different radio access technologies [209210]

n this way a user terminal can offload selected flows to a WLAN

ased on some operator-defined policy while keeping the LTE con-

ection running However to enable such an approach it is neces-

ary to support an entity in the cellular core network that can com-

unicate with the user terminal to exchange information about the

vailability and quality of neighbouring access networks as well as to

rovide the user terminal with predefined rules to manage the han-

over process This entity in current LTE standards is called the Ac-

ess Network Discovery and Selection Function (ANDSF) server [211]

ote that the problem of selecting the best communication technol-

gy in an heterogeneous wireless network has been extensively in-

estigated in the literature both with centralised as well as decen-

ralised approaches (see [212] for a survey) However the use of mul-

iple interfaces in parallel as well as per-flow offloading are relatively

ew concepts Thus the design of scalable and efficient network se-

ection strategies for the ANDSF framework is still an open issue For

nstance three offloading methods suitable for the ANDSF framework

re proposed in [213] which are based on coverage SNR and sys-

em load A reinforcement learning approach is designed in [214] that

llows multimode base stations to autonomously steer their traffic

ows across different access technologies depending on the traffic

ype the usersrsquo QoS requirements the network load and the inter-

erence levels A number of studies have also tried to quantify the

B Bellalta et al Computer Communications 75 (2016) 1ndash25 15

m

b

s

c

t

u

a

t

o

l

t

w

t

t

a

o

c

c

a

(

t

o

t

t

D

t

t

o

t

f

a

t

c

i

a

c

d

t

s

t

b

I

l

(

e

a

i

R

p

e

t

c

c

i

b

i

i

t

b

t

t

t

W

a

p

c

i

t

h

p

t

n

a

t

e

t

t

t

p

t

a

r

8

w

s

w

[

i

S

o

a

t

f

i

8

i

r

a

a

i

t

c

(

I

s

a

e

t

t

c

[

i

c

i

g

s

n

s

t

i

o

o

s

c

D

o

ethodology that is receiving more attention in the literature is

ased on distributed channel sensing solutions [127144145] These

olutions include the observation of spectrum use and possible

ooperative aggregation of spectrum sensing information to increase

he accuracy of detection and reduce the vulnerability of primary

sersrsquo transmissions The spectrum sensing capability is required in

ll the secondary devices with a minus114 dBm sensitivity Dynamic

ransmission power control must be provided and the upper limit

n emissions is 100 mW EIRP (20 dBm) for portable devices (further

imited in the case of adjacent channel use to reduce out-of-band in-

erference) However database (DB) coordination of TVWS combined

ith spectrum sensing is considered the most promising and effec-

ive technique compared with spectrum sensing alone The DB spec-

rum information has more chances to be effective (ie identifying

ll free channels opportunities) and reliable (ie not prone to attacks

r erroneous interpretation of channels status) In fact regulators de-

ided to push for the conservative solution of a remotely accessible

entralised spectrum DB that maintains the information on the avail-

bility of all TVWS channels at any given point in time and location

with a target accuracy around 50 m) under their responsibility For

his reason the spectrum databases have become a mandatory part

f many spectrum sharing systems and the dominant technical solu-

ion to support TV white spaces

The IEEE 80211af standard adheres to this vision and also has

he role of a common regulatory framework for different spectrum

B implementations defining a generalised coordination architec-

ure protocols and interfaces for spectrum queries and local spec-

rum information management The protocol explicitly considers out

f scope the communication protocol and technology adopted in

he background leaving the freedom to those players responsible

or the deployment to select any suitable Internet-based and local

ccess network technology Secondary devices that need to access

he TVWS spectrum should get information on which channels they

an effectively use and which parameters to adopt without impact-

ng primary usersrsquo transmissions (and receptions) by querying the

vailable spectrum DB The query must contain the transmission

haracteristics and accurate geographical position of the secondary

evice which can be determined by means of a geo-positioning sys-

em (GPS) regularly updated in the case of movement (or manually

et for static devices) The IEEE 80211af reference system architec-

ure for DB spectrum access is composed of multiple Geolocation-

ased Spectrum information Databases (GDBs) entities connected via

nternet to Registered Location Secure Servers (RLSS) RLSSs work as

ocal proxies of the GDB for a localised group of Basic Service Sets

BSS) RLSS are connected via a secure protocol architecture with the

quivalent of multiple Access Points (AP) in different BSS realising

trusted DB infrastructure APs locally coordinate the exchange of

nformation and channel access management between the GDB (via

LSS) and the secondary usersrsquo end stations (STA)

Mode I devices are those under the control of a device that em-

loys geo-location database access while Mode II devices are those

mploying geo-location database access by themselves The informa-

ion exchange provided between the GDB and a secondary user STA

an be provided in both an open-loop (eg adopted by FCC) and a

losed-loop (eg adopted by ETSI) implementation In an open loop

mplementation daily spectrum availability information is provided

y the GDB and no feedback on the spectrum information received

s provided by STAs thus the system approach for spectrum access

s much more conservative leading to low channelsrsquo utilisation po-

ential In a closed loop implementation the STA can provide feed-

ack to the GDB and there are more communication overheads due

o the high granularity of updates however the system is more effec-

ive and reliable in the exploitation of TVWS on behalf of STAs The

ypical information provided by the GDB includes (1) the updated

hite Space Maps (WSM) of frequencies allowed for secondary use

t the timespace of the querying STA and (2) the device-dependent

ower limitations for transmission (in general conservative and ac-

urate enough to avoid relevant interference effects on primary users

dentified in the area)

The basic IEEE 80211af mechanisms regulating the communica-

ion between STA and GDB can be found in [137138]

In parallel with IEEE 80211af standards such as the IEEE 19006

ave been created that consider interfaces and data structures sup-

orting spectrum sensing information exchange applicable to spec-

rum sensing (and particularly distributed spectrum sensing) sce-

arios In particular in October 2014 the IEEE 19006 WG initiated

project for a new standard called IEEE 19006b [146] concerning

he use of spectrum sensing information to support and optimise the

ffectiveness reliability and robustness of spectrum database solu-

ions The aim is to enhance the performance and capabilities of spec-

rum databases through the use of spectrum sensing information

Co-existence Support for co-existence mechanisms so that mul-

iple technologies can effectively utilise the TVWS spectrum is im-

ortant Self co-existence between network devices of a common

echnology (eg deployed by different operators in the same area)

nd co-existence among different technologies are relevant topic of

esearch for Cognitive Radio (CR) systems and specifically for IEEE

0211af Many solutions appeared on the research scene but no one

as so far finalised as the target solution for IEEE 80211af Specific

tandardisation has been started to regulate coexistence between

ireless standards of unlicensed devices including the IEEE 802191

147] The purpose of the IEEE 802191 standard is to enable the fam-

ly of IEEE 802 Wireless Standards to most effectively use TV White

pace by providing standard coexistence methods among dissimilar

r independently operated TVWS devices Early examples of gener-

lised coexistence mechanisms included Dynamic Frequency Selec-

ion (DFS) Transmission Power Control (TPC) listen before talk (eg

or contention based IEEE 80211 80215) time division multiplex-

ng (also among different techniques such as the IEEE 80216 80220

0222) and Message-based Spectrum Contention (that is beacon-

ng messages that carry coexistence information) Opportune met-

ics must be defined to assess the measurable coexistence achieved

mong different technologies as an example the hidden node prob-

bility for a target scenario or the estimate of percentage variation

n normalised network throughput and latency (before and during

he SU transmissions) On the other hand a centralised coexistence

ontrol mechanism could be effectively realised by a central manager

or coexistence-DB like the GDB) in critical scenarios To this end

EEE 19004 [148] (a standard for heterogeneous networks in dynamic

pectrum context part of IEEE Standards Coordinating Committee 41)

ims to standardise the overall system architecture and information

xchange between the network and mobile devices which will allow

hese elements to optimally choose from available radio resources

There are three possible classification of co-existence archi-

ectures and inter-network coordination channels for CR systems

entralised coordinated and autonomous co-existence mechanisms

149] In centralised co-existence schemes the co-existence is admin-

stered by a central entity (eg like in IEEE 802191) This solution

ould be applied in both homogeneous IEEE 80211af systems (eg

ncarnated by the centralised spectrum DB entity) and in hetero-

eneous inter-networks without requiring modifications to existing

tandards under the assumption that all the involved co-existing

etwork entities would adhere to the same centralised coordination

cheme and inter-network coordination protocol On the other hand

his assumption could be hard to satisfy in some practical scenar-

os The centralised schemes are effective in providing minimisation

f inter-network interference based on the availability and quality

f centralised co-existence information In coordinated co-existence

chemes the centralised entity could be present but not taking de-

isions in general The central entity could provide a co-existence

B with required information which can be used via in-band and

ut-band signalling for co-existence decisions taken by the cluster

16 B Bellalta et al Computer Communications 75 (2016) 1ndash25

c

e

n

c

o

s

i

d

a

S

8

n

a

b

o

b

i

t

a

8

b

[

t

i

n

b

5

t

a

c

b

s

a

c

n

s

p

i

c

l

r

c

f

a

v

t

a

e

c

t

Q

a

b

s

i

c

u

n

c

i

heads of many co-existing inter-networks for implementing a proper

common coordination protocol (eg out-band busy tone signalling)

These are often hybrid solutions realising a compromise between ef-

fectiveness and suitability In autonomous co-existence schemes all the

decisions and coordination are implemented in a distributed way

by the involved network entities In these schemes many policies

can be adopted to realise a sufficient level of co-existence under a

best effort approach Solutions that could be implemented include

the use of busy tones beaconing and other signalling protocols in

dedicated control channels dynamic distributed frequency selection

schemes listen before transmit token-based and dynamic reserva-

tion schemes All the above mentioned solutions have goods and

bads and a general illustration can be found in [149] Other proac-

tive co-existence techniques try to early detect and recovermitigate

co-existence issues with relaxed inter-network coordination eg re-

alised via spectrum sensing and interference avoidancesuppression

techniques

It must be clear that different aims exist regarding the co-

existence for primary TVWS users protection against secondary

users and for secondary devices mutual interference avoidance

Many co-existence problems have been analysed in the literature

for TVWS technologies in particular between the primary (licensed)

and secondary (unlicensed) devices The coexistence management in

TVWS is complex because primary users of DVB transmissions are in-

tended to be the pure receivers rather than the broadcasters In other

words potential protected users could be everywhere and the hid-

den terminal problem could arise for these systems The co-existence

problem is also complicated by many factors that cause asymme-

try and dynamicity in the TVWS such as the mobility variable den-

sity power asymmetry and heterogeneous MACPHY layers TVWS

co-existence problems originate between the IEEE 80211af technolo-

gies and the IEEE 80222 technologies for wireless regional area net-

works (WRAN) Results of analysis of TVWS usage in Europe show

that white spaces are typically present and fragmented They are typ-

ically more abundant in rural areas where larger contiguous blocks

of unused channels are available due to broadcast network planning

giving priorities linked to population density However the exploita-

tion of TVWS in urban areas is possible eg some recent research

was realised under the assumption of exploiting shadowing effects

created in the communication environment (eg by buildings obsta-

cles) in favour of frequency reuse [150]

Since different standards for opportunistic communication in the

TV White Spaces have now been published such as IEEE 80222 [151]

IEEE 802154m [152] Weightless [153] and of course IEEE 80211af

improved methods to guarantee the coexistence of different devices

operating on several protocols in the same bands must be deployed

We highlight here some work eg [150154ndash157] that focuses on

the coexistence between different technologies operating in the same

bands In particular [155] studies the performance degradation of

IEEE 80222 when an IEEE 80211af network is operated in the same

area The problem becomes even worse when the IEEE 80211af net-

work is located near 80222 user equipment causing both networks

to perceive a strong interference from each other One proposed solu-

tion is the Coexistence Beacon Protocol [158] studied for 80222 net-

works which foresees the exchange of a periodic beacon to identify

the neighbouring and possibly interfering networks

For what concerns the co-existence problems research has ad-

dressed other important related aspects of new WLANs on TV White

Spaces For instance the higher coverage range makes them attrac-

tive for applications in the smart grid and also for M2M on TVWS

[134150159] Here the increased range compared to standard tech-

nologies in the ISM bands can have beneficial effects for indoor

mobile devices [150] and M2M [159] Specifically indoor devices

communicating on TV White Spaces offer better propagation char-

acteristics and penetration through obstacles making it easier to

realise scalable home connectivity [135] although interference and

o-existence problems could be exacerbated and must be resolved

g see [139] However studies in the literature show how the sig-

al coming from an indoor transmitter on TV White Spaces remains

onstrained inside the house making it difficult to deploy indoor-to-

utdoor networks On the other hand this shadowing limit can be

een as an opportunity lowering the interference outside the build-

ng in which the TVWS network is locally operated [134150] Un-

er this approach HDTV streaming has recently been considered as

possible use case of WLAN networks operating in the TV White

pace [160]

Non-contiguous channel bonding A relevant novel feature of IEEE

0211af is the potential for contiguous and non-contiguous chan-

el bonding which permits aggregating basic channels (also non-

djacent ones) and leveraging the possible large frequency spread

etween multiple available channels Due to the rather static nature

f primary DTV transmissions where a busy channel is unlikely to

ecome free in the near future and state changes are coarse-grained

n general it is crucial to exploit the time-locality effect and exploit

he maximum physical channel availability that could be aggregated

t any given location With the methodology inherited from IEEE

0211ac IEEE 80211af is capable of bonding together two up to four

asic channels grouped in up to two different non-contiguous chunks

137] The spectrum bandwidth of a DVB-T basic channel can be ei-

her 6 7 or 8 MHz depending on the country in which the service

s operated As an example this creates a 144 to 168 OFDM chan-

elsrsquo bandwidth potential when up to four 6ndash7ndash8 MHz channels are

onded [137]

12 Open challenges

The deployment of WLANs in the TVWS at their maximum poten-

ial still requires the resolution of open problems Generally speaking

s mentioned in previous sections the challenges for such networks

an be divided into three main categories (1) spectrum sharing

etween opportunistic (secondary) and heterogeneous devices (2)

pectrum sensingmanagement and maximum exploitation of avail-

ble spectrum (potentially enabling the spectrum-on-demand con-

ept) and (3) co-existence and interference mitigation to the primary

etwork (primary user protection) and secondary networks [154]

The overall aim of spectrum sharing techniques is to maximise

eparation to avoid overlapping or contiguous channels operations

rovided that sufficient channels are available in the TVWS spectrum

n a given spacetime scenario [154] In general important research

ontributions have to be realised for the design of proper spectrum al-

ocation methodologies and techniques satisfying multi-factorial QoS

equirements at the system-level and the user-level Novel ideas in-

lude the spectrum sharing and spectrum sensing techniques being

urther divided into cooperative and non-cooperative

Non-cooperative solutions attempt to realise spectrum sharing

nd sensing on a local basis without direct cooperation between de-

ices Examples include the ldquolisten-before-transmitrdquo approach and

ransmission power control to limit the interference and spectrum

llocation policies that are based on local node feedback only Coop-

rative solutions rely on the existence of a common communication

hannel through which devices can tentatively agree on the spec-

rum allocation that provides the desired spectrum separation and

oS Major issues include the definition of a methodology to identify

common channel (or multiple channels mutually shared in space

y pairs of heterogeneous and distributed devices) and to efficiently

hare common information For the access to the spectrum DBs the

dentification of proper common access channels is still under dis-

ussion out-of-TVWS-band cellular communications can be widely

sed given their high coverage also indoor

Trying to avoid the use of complex and resource-hungry coordi-

ation schemes and dedicated control channels rendezvous proto-

ols play a significant role both for co-existence and for establish-

ng dynamic communication channels based on spectrum sharing

B Bellalta et al Computer Communications 75 (2016) 1ndash25 17

R

n

c

t

l

o

m

e

n

i

n

w

n

a

u

s

p

b

s

m

s

a

d

t

b

a

a

m

a

t

u

8

n

t

d

d

m

f

t

n

d

n

e

r

a

h

b

f

a

c

t

u

i

f

t

d

t

s

i

f

I

l

i

a

n

c

d

I

n

o

t

i

t

e

w

t

c

m

t

a

i

t

q

b

c

a

i

l

i

a

v

t

a

a

c

t

i

t

b

t

t

a

D

p

c

s

o

q

r

p

a

g

t

e

u

d

t

m

n

t

s

f

i

1

8

f

T

endezvous is considered a fundamental problem in generalised CR

etworks at the basis of many communication processes eg in-

luding neighbour discovery routing and broadcast Rendezvous pro-

ocols attempt to establish a new link for communication on a se-

ected and agreed frequency band (channel) identified within a set

f available resources (channels) thus creating the basis for com-

unication between two or more SUs This problem is often exac-

rbated by the high (and unknown) numbers of contending SUs high

umber of available channels and their ambiguous boundaries and

dentifiers and tough co-existence requirements satisfaction in cog-

itive radio systems [161] Preliminary approaches for rendezvous

ere based on the existence of a dedicated Common Control Chan-

el (CCC) agreed among all the SUs or the existence of a centralised

gent (decision maker) assigning channels to any pairs of requesting

sers This approach could be considered in the case of IEEE 80211af

olutions based on the spectrum DB implementation however the

roblem is complicated by the fact that the common channel must

e identified and made available for all the requesting SUs at the

ame time even when these are spread over multiple collision do-

ains in space On the other hand the scalability and reliability is-

ues which are caused by the congestion of CCCs and the vulner-

bility risks of centralised DBs motivate the research of alternative

istributed solutions called blind rendezvous protocols [162] Dis-

ributed solutions have been mainly based on beaconing andor slot-

ased channel hopping algorithms exploiting randomisation sensing

nd discoverysynchronisation messages spread in slotted time over

set of candidate available channels [161] These protocols aim to

inimise the convergence time to maximise the use of resources

nd to avoid collisions among SUs by identifying and agreeing on

he use of a given common channel among N non-overlapping and

nambiguously labelled channels The main complication in IEEE

0211af systems is given by the wide availability of many different

on-contiguous spectrum bands which cannot be uniformly quan-

ised in single channels and unambiguously labelled in space in or-

er to realise a common reference domain for all the IEEE 80211af

evices In fact the knowledge of the number of channels their com-

on labels the unique IDs of SUs and the number of SUs contending

or the channel assignment are the key factors determining the effec-

iveness and good properties of blind rendezvous protocols Recently

ew classes of rendezvous protocols have been proposed which are

istributed blind and oblivious Oblivious means that available chan-

els may be identified with different labels on behalf of the differ-

nt involved SUs A number of these distributed blind and oblivious

endezvous protocols have been proposed and have been referenced

nd analysed in [162] To conclude solutions for dynamic spectrum

andoff issues (eg in case PUs suddenly appear in a region) must

e resolved in advance to minimise the latency of re-establishing a

unctional communication channel eg by implementing a proactive

lternative rendezvous channel selection in background to ongoing

ommunication processes [163]

Another challenge is to provide cooperative techniques in a dis-

ributed vs centralised implementation by analysing their convergence

nder variable conditions overheads and tradeoffs In some cases as

n vehicular networks the exploitation of scenario characteristics and

actors such as the constrained mobility could help to realise effec-

ive dissemination of sensing information over an extended and pre-

ictable horizon under a cooperative approach [164ndash166] Coopera-

ion could be helpful in the realisation of the cooperative spectrum

ensing in combination with the Geolocation DB approach Another

nteresting direction of research is the investigation of mutual ef-

ects of coexisting cooperation-based vs non-cooperative techniques

n cooperation-based devices the exchange of information could al-

ow a quick convergence to a solution and avoidance of further

nterference [145] Certainly cooperative techniques have desirable

dvantages over non-cooperative ones However tight time synchro-

isation and high coordination overheads are required The common

hannel for cooperation purposes could become a bottleneck and re-

uce the potential advantages of dynamic spectrum allocation The

EEE 80219 foresees the presence of a shared common control chan-

el (CCC) to which devices need to tune in order to gather the state

f the network Solutions such as [141] propose the use of TVWS for

he CCC

A promising direction for research is the adoption of cluster-

ng schemes for secondary devices associated with a hybrid dis-

ributedcooperative vs centralisednon-cooperative approach in an

ffort to reduce overheads while maximising advantages In this

ay secondary nodes could implement cooperation for sensing func-

ions using properly identified common channels and delegate the

entralised decisions and spectrum-DB management to well instru-

ented cluster-leader nodes This concept is incarnated to some ex-

ent by RLSS in IEEE 80211af More recently there have been propos-

ls to build distributed databases that refresh their contents period-

cally by querying the remote spectrum database [167] Here Mas-

er devices (according to Ofcom terminology) periodically cache the

uery replies from the spectrum database in order to reply to and

roadcast the spectrum availability to neighbour Slave devices

Another challenging direction for research is the definition of ac-

urate models and efficient simulation tools enabling dynamic spectrum

nalysis for both rural and urban areas Recent developments of dig-

tal maps simulation tools and propagation models theoretically al-

ow improving the capability of predicting radio propagation effects

n complex scenarios with an acceptable computation time This en-

bling technology could be used in parallel to spectrum DBs to pro-

ide more accurate forecasting of dynamic frequency allocation in

imespace scenarios Another issue to be taken into account is the

ccuracy of the remote spectrum database regarding the channel

vailability estimation Several works have already shown the inac-

uracy of propagation models due to the complexity of parameters

o be considered and differences between modelled and real scenar-

os [132134168169] Better propagation models will make it possible

o more efficiently estimate the interference between devices and

uild Radio Environment Maps (REM) or White Space Maps (WSM)

hat will possibly lead to a more accurate sharing of the radio spec-

rum [144] To this end current research could focus on how to lever-

ge simulation models in order to build more accurate frequency

Bs

In general the adoption of remote spectrum DBs methodology

ushed by national regulators has made the spectrum sensing pro-

ess optional and conservatively realised (eg power control for

ensing-only devices is limited to below 17 dBm EIRP) Thus much

f the research has focused on techniques and solutions to successfully

uery the remote database also in challenging environments such as

ural areas and indoor However since the queries must provide the

osition of the mobile device the accuracy of positioning for mobile

nd handheld devices still needs more appropriate solutions Trian-

ulation can be used to estimate the position of mobile devices with

he accuracy needed by the current regulations outdoor (50 m) How-

ver a low GPS accuracy is possible indoors [170] Certified man-

al positioning can be provided by technicians for static and indoor

evices although limiting mobility Alternatives to costly infrastruc-

ures for positioning (eg based on short range beaconing devices)

ust be deployed in indoor scenarios

Finally regarding spectrum sensing and interference mitigation

ew challenges come from the heterogeneity and different charac-

eristics that networks operating in the TV White Spaces can have

uch as different transmitting power (4 W for static devices 100 mW

or mobile and portable devices and 40 mW for communication

n channels adjacent to occupied ones) different bandwidths (5

0 15 and 20 MHz are currently foreseen for IEEE 80211af IEEE

02154m can have narrower bandwidths or wider ones) and dif-

erent medium access schemes (protocols that use either CSMA or

DMA are required to co-exist in the TV White Spaces) With all this

18 B Bellalta et al Computer Communications 75 (2016) 1ndash25

d

a

T

i

a

c

t

l

b

o

t

a

d

p

a

r

p

n

a

m

t

p

p

d

d

o

t

i

D

w

p

i

a

a

d

b

fi

d

o

t

l

b

t

b

a

i

e

o

o

c

t

t

w

A

c

M

s

c

l

O

w

g

v

t

in place the problem of determining whether another secondary de-

vice is currently transmitting in the same channel is still an open

issue

Spectrum sensing is a very hot topic in the literature with many

proposals that leverage the cooperation between devices to raise the

accuracy of the detection process eg [144171] The current litera-

ture also offers room for proposals that suggest exploiting sensing to

build more accurate spectrum DBs [164172173] Lastly regarding the

interference mitigation to the primary network regulators provide a

conservative approach based on a remote spectrum database to avoid

high risk for the primary users However in particular for the indoor

scenario the concept of secondary usersrsquo transmission in TV Grey

Spaces has been proposed TV Grey Space identifies busy channels

that are formally used by primary users in the area of interest (eg

at the rooftop level of a building for DVB-T) but which could be con-

sidered usable indoors (eg at the basement level) without causing

effective interference to the primary users [134150174] This concept

is based on the generalised spectrum underlay paradigm [175] The

remaining research challenge is to provide TV Grey space access guar-

anteeing a sufficient amount of protection to the primary user while

providing a satisfactory QoS to opportunistic devices eg through

the use of customised and validated propagation models for indoor

TVWS networks and feedback mechanisms to constantly monitor the

interference generated to the primary network

6 Emerging new trends and technologies

In this last section we review three emerging trends that in our

opinion will have a large influence in the conception of future WLANs

as they change the way WLAN protocols and functionalities are de-

veloped implemented tested and integrated with other wireless

networks

61 Programmable Wireless LANs

Especially in the enterprise environment WLAN deployments

need to support a wide range of functionalities and services This is

intrinsically difficult because of the large number of APs that must be

managed which calls for scalable solutions Typical services include

channel assignment load balancing among APs authentication au-

thorisation and accounting (AAA) policy management support for

client mobility and interference coordination Another problem is the

fact that WLAN clients autonomously take several decisions such as

which APs to associate with when to hand-over etc Therefore sup-

porting roaming clients requires the management of a large number

of association states across several APs which is a challenge if sup-

port for real-time hand-over is desired Typically such management

schemes are centralised and most of them are proprietary such as

WLAN controller solutions from Aruba [176] or Cisco [177] although

the IEEE 80211u amendment has been released to allow mobile users

to seamless roam between WiFi networks with automatic authenti-

cation and handoff [178] For example Dyson [179] enables STAs to

send information such as radio channel conditions to a centralised

controller based on a custom API (eg based on Python) As the con-

troller has a centralised view of the network it can enforce a rich

set of policies to control the network also using historical informa-

tion A demo system has been implemented along with applications

such as airtime reservations for specific clients or optimised hand-

offs However Dyson requires STAs to be modified in order to use

those new services offered by the centralised controller TRANTOR

[180] is another example of a centralised management system that

requires changes of clients in order for the infrastructure to gather

information from them in terms of eg interference measurements

In addition control commands enable the infrastructure to exercise

control such as modifying the transmit power or influencing the as-

sociation procedure Clients still use standard CSMA MAC layer for

ata transmission In contrast to DenseAP [181] or DIRAC [182] which

re also based on a centralised controller but do not modify clients

RANTOR exercises much more control on clients by the use of a ded-

cated API which enables a significantly higher gain in coordination

nd thus capacity CENTAUR [183] is another example of a centralised

ontroller that centrally schedules hidden and exposed terminals in

he downlink while using standard CSMA MAC for uplink traffic and

egacy downlink traffic To achieve good performance it uses a fixed

ack-off packet staggering techniques and a hybrid data path that

nly schedules downlink transmissions towards hidden and exposed

erminals centrally All other traffic is sent using standard DCF with

standard back-off procedure In contrast to previous approaches it

oes not require any modifications to the clients but does require data

lane centralisation and it remains questionable how scalable such

solution is for large WLAN deployments with high PHY layer data

ates beyond 100 Mbps However new approaches for fast data plane

rocessing that are currently explored in the virtualisation commu-

ity which move all the packet processing into user space may be

n option to significantly increase speed Designing such manage-

ent systems raises several interesting research questions (a) Cen-

ralised infrastructures may incur high latency but are more simple to

rogram and maintain while designing a completely distributed ap-

roach that operates close to WLAN devices may result in significant

istributed coordination and consistency problems (b) what level of

istributed control is required in order to support the flexibility needs

f future WiFi based networks supporting a high level of mobility

Traditionally WLAN APs are built on proprietary operating sys-

ems that are tightly coupled with the hardware This design makes

t hard to create new applications on top of such networking devices

espite the fact that protocols and mechanisms are available that

ould greatly improve the utility of existing networks those new

rotocols are not deployed because the closed system design makes

t very difficult to extend their functionality An important aspect for

ll such new approaches is therefore how to provide open interfaces

nd open source to speed up innovation As an example Linux-based

evices are completely open source but in order to increase flexi-

ility much more work is needed in the area of open drivers and

rmware In fact the Atheros based ath9k driver has been the main

river of innovation in the open source community because it is the

nly driver that interworks with open firmware In general solutions

hat modify the OS driver focus mainly on programmable network

evel solutions for WLANs such as channel switching or handovers

etween APs While the ath9k driver is an excellent example in how

o speed up innovation it is still very cumbersome to support flexi-

le MAC engine reprogrammability with ath9k In contrast MAClets

nd Wireless MAC processors (WMP) [184] allow a much more flex-

ble reprogramming of MAC functionalities A MAC processor is an

ntity able to execute general MAC commands that specify the MAC

perations through a software-defined state machine The behaviour

f the MAC protocol can therefore be updated at run-time by simply

hanging the sequence in which those commands are executed (ie

he MAClets) As a proof-of-concept the authors of [184] implement

he proposed MAC processor solution in a commodity WLAN hard-

are card extending the basic DCF in three directions piggy-backing

CKs a pseudo TDMA and the use of multiple channels In [185] a

ontrol framework for this WMP system is also proposed to support

AClet code mobility ie for moving loading and activating MAC

oftware programs embedded into ordinary data packets (akin to the

apsule model of traditional active networks)

The difficulty in re-programming networking hardware has also

ed to the concept of Software Defined Networks (SDNs) based on the

penFlow protocol [186] The main idea of SDNs is to extend net-

orking devices with standardised APIs that allow third-party pro-

rammers to flexibly control the data path In addition SDNs pro-

ide higher level abstractions to network designers and programmers

hrough the use of a centralised control plane offered by a network

B Bellalta et al Computer Communications 75 (2016) 1ndash25 19

Fig 11 The CLOUDMAC architecture VAP stands for Virtual Access Point and WT

stands for Wireless Termination

c

a

c

t

m

W

n

t

t

c

p

t

u

h

a

i

e

s

o

c

a

a

f

T

r

s

a

s

p

o

M

C

i

T

t

fl

A

a

a

i

c

t

s

t

e

I

t

w

H

e

b

I

l

d

o

6

r

W

s

l

a

e

e

a

s

b

c

t

(

S

p

c

F

B

t

p

T

e

t

s

u

m

i

e

o

a

A

p

d

t

t

n

p

i

t

t

f

i

ontroller such as NOX [187] which allows reuse of components such

s topology discovery or network access control for different appli-

ations The SDN concept has recently been applied to WLAN archi-

ecture in order to enable fine grained control over mobility manage-

ent and data forwarding focusing on a programmable enterprise

LAN architecture An important part of the SDN architecture is the

etwork controller which provides a centralised view of (parts of)

he SDN enabled network and uses the OpenFlow protocol to install

he forwarding rules on SDN-enabled devices (routers switches ac-

ess points cellular base stations etc)

ODIN [188] is designed to support programmability in enter-

rise WLAN architecture by separating the association state from

he physical AP They implement Light Virtual Access Points (LVAP)

sing a Split-MAC approach where the infrastructure controls the

andover procedure among different WLAN APs By managing

ssociations through SDN controllers ODIN enables proactive mobil-

ty management and load balancing within the SDN enabled WLAN

nterprise network without the need for changes in the client WLAN

tack or IEEE 80211 MAC layer While ODIN requires agents to reside

n the APs that communicate with the Odin Master within the SDN

ontroller CLOUDMAC [189] is based purely on the concept of SDN

nd virtual APs Similar to ODIN CLOUDMAC implements a Split-MAC

pproach but in addition enables the processing of WLAN MAC layer

rames within a co-located Cloud using so-called Virtual APs (VAPs)

he physical APs in CLOUDMAC are lightweight WLAN APs that are

esponsible only for sending out IEEE 80211 based MAC layer ACKs to

tandard WLAN clients and tunnel WLAN MAC layer frames through

n SDN-enabled enterprise WLAN towards the VAPs As association

tates are kept in the VAPs fast mobility is supported using sim-

le OpenFlow forwarding rules Because of the additional processing

f IEEE 80211 WLAN MAC frames in the co-located Cloud CLOUD-

AC has higher latency than a standard WLAN deployment However

LOUDMAC offers a Webservice based API to third party applications

n order to program the enterprise WLAN and enable new services

he architecture of CLOUDMAC is depicted in Fig 11 It has been ex-

ended in [190] to support QoS management and in [191] to support

exible MAC management frame prioritisation based on IEEE 80211

system based on OpenFlow has been recently proposed in [192] to

llow the station to be associated with multiple APs simultaneously

nd to switch between APs with low overhead

An important aspect to consider in order to enable programmabil-

ty is backwards compatibility Several approaches (eg [180]) require

lients to be modified to utilise the features provided This is difficult

o do in practice because it requires changes in the operating system

oftware of all clients If not all clients can utilise those APIs it is ques-

ionable how usable the new architecture will be and how much ben-

fit in terms of aggregate performance such architecture will allow

n contrast the SDN based approaches are interesting in the sense

hat they do not require changes in the client WLAN stack and work

ith the standard IEEE 80211 MAC layer deployed within the clients

owever it remains questionable how scalable such solutions are For

xample an interesting open research topic is to evaluate the scala-

ility of approaches such as CLOUDMAC [189] In addition processing

EEE 80211 MAC frames within co-located private Cloud requires low

atency support from local Cloud solutions (such as OpenStack) in or-

er to reduce the IEEE 80211 MAC processing time which is an area

f active research

2 Prototyping and testing IEEE 80211 enhancements

Most of the new proposals for next-generation WLANs are cur-

ently only evaluated using mathematical analysis and simulation

hile both analysis and simulation are necessary to characterise and

tudy those enhancements in the initial design phase or to consider

arge-scale scenarios it is difficult to consider all practical aspects of

real-world scenario This can sometimes cause significant differ-

nces between what simulations and real experiments show How-

ver real experiments are challenging because of the high complexity

nd costs of building the new hardware and software for each specific

olution to test

To mitigate implementation complexity there are several flexi-

le hardware platforms where low-level MAC mechanisms can be

ompletely implemented in software Examples based on FPGAs are

he USRP (Universal Software Radio Peripheral) [193] and the WARP

Wireless Open-Access Research Platform) [194] As an alternative

ORA (Microsoft Research Software Radio) [195] works on general

urpose computers by taking some advantage from modern multiple

ore systems

A different approach is provided by OpenFWWF [196] Open-

WWF provides an open CSMACA firmware for specific models of

roadcom chipsets so the resulting firmware can be uploaded and

ested in real commercial hardware OpenFWWF implements a sim-

le State Machine (SM) for controlling the hardware in real time

he SM evolution is driven by a main loop that reacts to events by

xecuting specific handlers When a packet originally prepared by

he Linux kernel is ready in the NIC memory handler Packet_Ready

ets up the radio hardware according to the packet metadata sched-

les the transmission and jumps back to the main loop The Trans-

ission Engine (TXE) then takes care of accessing the channel ie

t decrements the back-off counter according to the DCF rules and

ventually starts the actual transmission This triggers the execution

f the TX_frame_now event that prepares the ACK time-out clock

nd finalises the MAC header If the ACK-frame is received or if the

CK time-out expires and the maximum number of attempts for this

acket is reached handler Update_Params resets the contention win-

ow to its minimum value or otherwise doubles it Finally it loads

he back-off counter with a fresh value

In the following we will review a list of selected papers that use

he OpenFWWF firmware and the WARP or USRP platforms to test

ew proposals for WLANs in real scenarios as examples A first im-

lementation of the groupcast mechanisms defined in IEEE 80211aa

s presented in [197] The authors modify the OpenFWWF firmware

o include those functionalities and evaluate them using a 30 STAs

estbed A second example of the use of OpenFWWF is the collision-

ree MAC protocol presented in [198] In [198] the authors design and

mplement in OpenFWWF a MAC protocol that is able to achieve a

20 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 12 LTE to WIFI offloading

a

n

c

t

[

d

a

W

v

t

t

o

b

f

S

g

l

i

I

s

I

b

n

s

m

a

p

d

c

N

o

v

t

t

n

l

i

a

t

a

collision-free operation by waiting a deterministic time after success-

ful transmissions before to start a new transmission This work shows

the experimental performance of a collision-free MAC (CF-MAC) pro-

tocol for WLANs using commercial hardware The results show that

the proposed CF-MAC protocol leads to a better distribution of the

available bandwidth among users a higher throughput and lower

losses than DCF

In [199] the authors use the WARP platform to test a variant of the

DCF ndash called IEEE 80211ec ndash that substitutes control packets such as

the RTS CTS and ACK with short detectable sequences Since those se-

quences are shorter than the control packets and can be detected cor-

rectly even at lower SNRs values a significant gain in performance is

achieved MIDAS (Multiple-Input Distributed Antenna Systems) [200]

implementation using the WARP platform shows the benefits of dis-

tributing the antennas over the area to cover instead of co-locating

them at the AP in terms of capacity when MU-MIMO is employed

The authors also propose a new MAC protocol to benefit from the

spatial reuse that the DAS allows taking as a basis the IEEE 80211ac

Experimental results show that their proposal is able to achieve up

to 200 gains versus the traditional approach where all antennas are

co-located at the AP

The USRP platform has recently been proposed to develop a first

implementation of IEEE 80211ah in order to obtain a preliminary ex-

perimental performance assessment of such a technology because

no commercial off-the-shelf hardware is yet available as the amend-

ment is still in progress [201] The USRP platform is also used in

[202] to evaluate TIMO (Technology Independent Multi-Output) a so-

lution to deal with high-power non-IEEE 80211 interferers in ISM

bands

To conclude this section it is worth mentioning here also to

MAClets and WMP as given such flexibility to implement and dis-

tribute new MAC protocols and other IEEE 80211 functionalities we

believe that if a use-friendly implementation of such a MAC proces-

sors framework is provided it would significantly contribute to the

development and testing of new MAC enhancements by the research

community

63 CellularWLAN interworking

Public hotspots that offer Internet access over a WLAN using IEEE

80211 technology are now nearly ubiquitous It is forecasted that the

cumulative installed base of WiFi hotspots worldwide will amount

to 551 million by 2018 excluding private hotspots (eg WiFi access

points deployed at home) [203] The sharp increase in the availabil-

ity of public WiFi was initially perceived by mobile cellular operators

as a threat due to the additional competition from wireline Internet

service providers or emerging crowdsourced WiFi networks such as

FON5 However as cellular operators are fighting to cope with the ex-

plosion of mobile data traffic created by the rising use of multimedia

content traffic over mobile devices [15] they are also starting to use

WLANs based on the IEEE 80211 technology to offload data from their

core and access networks In general mobile data offloading refers to

the use of complementary network technologies (in licensed or un-

licensed spectrum) for delivering data originally targeted to cellular

networks Intuitively the simplest type of offloading consists of ex-

ploiting connectivity to existing co-located WiFi networks and trans-

ferring data without any delay (Fig 12) Thus this offloading tech-

nique is known as on-the-spot offloading As a consequence of this

new trend the seamless integration of cellular (eg 3GLTE) and WiFi

technologies has attracted significant research interest in recent years

(see [204] for a survey) a few solutions have already been standard-

ised [205206] and roaming between cellular and WiFi is becoming

increasingly transparent to end users CellularWLAN interworking is

5 httpscorpfoncomen

fl

t

f

lso fostered by the support in the evolving 4G standards of heteroge-

eous network deployments (HetNets) in which the existing macro

ells are complemented with a number of small low-power base sta-

ions with the goal of increasing capacity in highly congested areas

207208] It is envisaged that small cells will be based on 4G stan-

ards (eg pico and femto cells) as well as IEEE 80211 technologies

nd multimode base stations that work simultaneously with LTE and

iFi are already entering the market

It is important to point out that LTE standards already support a

ariety of mechanisms that enable data offloading However most of

he existing solutions such as Local IP Access (LIPA) and Selected In-

ernet IP Traffic Offload (SIPTO) focus on data offloading at the core

f the network [205] For instance LIPA allows an IP-enabled mo-

ile device to transfer data to another device in the same pico or

emto cell without passing through the cellular access network while

IPTO enables the routing of selected IP data flows through different

ateways The only LTE offloading mechanism that supports seam-

ess interworking with IEEE 80211-based WLANs is IP Flow Mobil-

ty and Seamless Offload (IFOM) Specifically IFOM relies on Mobile

Pv6 technologies to allow a user terminal to simultaneously route

elected IP flows over different radio access technologies [209210]

n this way a user terminal can offload selected flows to a WLAN

ased on some operator-defined policy while keeping the LTE con-

ection running However to enable such an approach it is neces-

ary to support an entity in the cellular core network that can com-

unicate with the user terminal to exchange information about the

vailability and quality of neighbouring access networks as well as to

rovide the user terminal with predefined rules to manage the han-

over process This entity in current LTE standards is called the Ac-

ess Network Discovery and Selection Function (ANDSF) server [211]

ote that the problem of selecting the best communication technol-

gy in an heterogeneous wireless network has been extensively in-

estigated in the literature both with centralised as well as decen-

ralised approaches (see [212] for a survey) However the use of mul-

iple interfaces in parallel as well as per-flow offloading are relatively

ew concepts Thus the design of scalable and efficient network se-

ection strategies for the ANDSF framework is still an open issue For

nstance three offloading methods suitable for the ANDSF framework

re proposed in [213] which are based on coverage SNR and sys-

em load A reinforcement learning approach is designed in [214] that

llows multimode base stations to autonomously steer their traffic

ows across different access technologies depending on the traffic

ype the usersrsquo QoS requirements the network load and the inter-

erence levels A number of studies have also tried to quantify the

16 B Bellalta et al Computer Communications 75 (2016) 1ndash25

c

e

n

c

o

s

i

d

a

S

8

n

a

b

o

b

i

t

a

8

b

[

t

i

n

b

5

t

a

c

b

s

a

c

n

s

p

i

c

l

r

c

f

a

v

t

a

e

c

t

Q

a

b

s

i

c

u

n

c

i

heads of many co-existing inter-networks for implementing a proper

common coordination protocol (eg out-band busy tone signalling)

These are often hybrid solutions realising a compromise between ef-

fectiveness and suitability In autonomous co-existence schemes all the

decisions and coordination are implemented in a distributed way

by the involved network entities In these schemes many policies

can be adopted to realise a sufficient level of co-existence under a

best effort approach Solutions that could be implemented include

the use of busy tones beaconing and other signalling protocols in

dedicated control channels dynamic distributed frequency selection

schemes listen before transmit token-based and dynamic reserva-

tion schemes All the above mentioned solutions have goods and

bads and a general illustration can be found in [149] Other proac-

tive co-existence techniques try to early detect and recovermitigate

co-existence issues with relaxed inter-network coordination eg re-

alised via spectrum sensing and interference avoidancesuppression

techniques

It must be clear that different aims exist regarding the co-

existence for primary TVWS users protection against secondary

users and for secondary devices mutual interference avoidance

Many co-existence problems have been analysed in the literature

for TVWS technologies in particular between the primary (licensed)

and secondary (unlicensed) devices The coexistence management in

TVWS is complex because primary users of DVB transmissions are in-

tended to be the pure receivers rather than the broadcasters In other

words potential protected users could be everywhere and the hid-

den terminal problem could arise for these systems The co-existence

problem is also complicated by many factors that cause asymme-

try and dynamicity in the TVWS such as the mobility variable den-

sity power asymmetry and heterogeneous MACPHY layers TVWS

co-existence problems originate between the IEEE 80211af technolo-

gies and the IEEE 80222 technologies for wireless regional area net-

works (WRAN) Results of analysis of TVWS usage in Europe show

that white spaces are typically present and fragmented They are typ-

ically more abundant in rural areas where larger contiguous blocks

of unused channels are available due to broadcast network planning

giving priorities linked to population density However the exploita-

tion of TVWS in urban areas is possible eg some recent research

was realised under the assumption of exploiting shadowing effects

created in the communication environment (eg by buildings obsta-

cles) in favour of frequency reuse [150]

Since different standards for opportunistic communication in the

TV White Spaces have now been published such as IEEE 80222 [151]

IEEE 802154m [152] Weightless [153] and of course IEEE 80211af

improved methods to guarantee the coexistence of different devices

operating on several protocols in the same bands must be deployed

We highlight here some work eg [150154ndash157] that focuses on

the coexistence between different technologies operating in the same

bands In particular [155] studies the performance degradation of

IEEE 80222 when an IEEE 80211af network is operated in the same

area The problem becomes even worse when the IEEE 80211af net-

work is located near 80222 user equipment causing both networks

to perceive a strong interference from each other One proposed solu-

tion is the Coexistence Beacon Protocol [158] studied for 80222 net-

works which foresees the exchange of a periodic beacon to identify

the neighbouring and possibly interfering networks

For what concerns the co-existence problems research has ad-

dressed other important related aspects of new WLANs on TV White

Spaces For instance the higher coverage range makes them attrac-

tive for applications in the smart grid and also for M2M on TVWS

[134150159] Here the increased range compared to standard tech-

nologies in the ISM bands can have beneficial effects for indoor

mobile devices [150] and M2M [159] Specifically indoor devices

communicating on TV White Spaces offer better propagation char-

acteristics and penetration through obstacles making it easier to

realise scalable home connectivity [135] although interference and

o-existence problems could be exacerbated and must be resolved

g see [139] However studies in the literature show how the sig-

al coming from an indoor transmitter on TV White Spaces remains

onstrained inside the house making it difficult to deploy indoor-to-

utdoor networks On the other hand this shadowing limit can be

een as an opportunity lowering the interference outside the build-

ng in which the TVWS network is locally operated [134150] Un-

er this approach HDTV streaming has recently been considered as

possible use case of WLAN networks operating in the TV White

pace [160]

Non-contiguous channel bonding A relevant novel feature of IEEE

0211af is the potential for contiguous and non-contiguous chan-

el bonding which permits aggregating basic channels (also non-

djacent ones) and leveraging the possible large frequency spread

etween multiple available channels Due to the rather static nature

f primary DTV transmissions where a busy channel is unlikely to

ecome free in the near future and state changes are coarse-grained

n general it is crucial to exploit the time-locality effect and exploit

he maximum physical channel availability that could be aggregated

t any given location With the methodology inherited from IEEE

0211ac IEEE 80211af is capable of bonding together two up to four

asic channels grouped in up to two different non-contiguous chunks

137] The spectrum bandwidth of a DVB-T basic channel can be ei-

her 6 7 or 8 MHz depending on the country in which the service

s operated As an example this creates a 144 to 168 OFDM chan-

elsrsquo bandwidth potential when up to four 6ndash7ndash8 MHz channels are

onded [137]

12 Open challenges

The deployment of WLANs in the TVWS at their maximum poten-

ial still requires the resolution of open problems Generally speaking

s mentioned in previous sections the challenges for such networks

an be divided into three main categories (1) spectrum sharing

etween opportunistic (secondary) and heterogeneous devices (2)

pectrum sensingmanagement and maximum exploitation of avail-

ble spectrum (potentially enabling the spectrum-on-demand con-

ept) and (3) co-existence and interference mitigation to the primary

etwork (primary user protection) and secondary networks [154]

The overall aim of spectrum sharing techniques is to maximise

eparation to avoid overlapping or contiguous channels operations

rovided that sufficient channels are available in the TVWS spectrum

n a given spacetime scenario [154] In general important research

ontributions have to be realised for the design of proper spectrum al-

ocation methodologies and techniques satisfying multi-factorial QoS

equirements at the system-level and the user-level Novel ideas in-

lude the spectrum sharing and spectrum sensing techniques being

urther divided into cooperative and non-cooperative

Non-cooperative solutions attempt to realise spectrum sharing

nd sensing on a local basis without direct cooperation between de-

ices Examples include the ldquolisten-before-transmitrdquo approach and

ransmission power control to limit the interference and spectrum

llocation policies that are based on local node feedback only Coop-

rative solutions rely on the existence of a common communication

hannel through which devices can tentatively agree on the spec-

rum allocation that provides the desired spectrum separation and

oS Major issues include the definition of a methodology to identify

common channel (or multiple channels mutually shared in space

y pairs of heterogeneous and distributed devices) and to efficiently

hare common information For the access to the spectrum DBs the

dentification of proper common access channels is still under dis-

ussion out-of-TVWS-band cellular communications can be widely

sed given their high coverage also indoor

Trying to avoid the use of complex and resource-hungry coordi-

ation schemes and dedicated control channels rendezvous proto-

ols play a significant role both for co-existence and for establish-

ng dynamic communication channels based on spectrum sharing

B Bellalta et al Computer Communications 75 (2016) 1ndash25 17

R

n

c

t

l

o

m

e

n

i

n

w

n

a

u

s

p

b

s

m

s

a

d

t

b

a

a

m

a

t

u

8

n

t

d

d

m

f

t

n

d

n

e

r

a

h

b

f

a

c

t

u

i

f

t

d

t

s

i

f

I

l

i

a

n

c

d

I

n

o

t

i

t

e

w

t

c

m

t

a

i

t

q

b

c

a

i

l

i

a

v

t

a

a

c

t

i

t

b

t

t

a

D

p

c

s

o

q

r

p

a

g

t

e

u

d

t

m

n

t

s

f

i

1

8

f

T

endezvous is considered a fundamental problem in generalised CR

etworks at the basis of many communication processes eg in-

luding neighbour discovery routing and broadcast Rendezvous pro-

ocols attempt to establish a new link for communication on a se-

ected and agreed frequency band (channel) identified within a set

f available resources (channels) thus creating the basis for com-

unication between two or more SUs This problem is often exac-

rbated by the high (and unknown) numbers of contending SUs high

umber of available channels and their ambiguous boundaries and

dentifiers and tough co-existence requirements satisfaction in cog-

itive radio systems [161] Preliminary approaches for rendezvous

ere based on the existence of a dedicated Common Control Chan-

el (CCC) agreed among all the SUs or the existence of a centralised

gent (decision maker) assigning channels to any pairs of requesting

sers This approach could be considered in the case of IEEE 80211af

olutions based on the spectrum DB implementation however the

roblem is complicated by the fact that the common channel must

e identified and made available for all the requesting SUs at the

ame time even when these are spread over multiple collision do-

ains in space On the other hand the scalability and reliability is-

ues which are caused by the congestion of CCCs and the vulner-

bility risks of centralised DBs motivate the research of alternative

istributed solutions called blind rendezvous protocols [162] Dis-

ributed solutions have been mainly based on beaconing andor slot-

ased channel hopping algorithms exploiting randomisation sensing

nd discoverysynchronisation messages spread in slotted time over

set of candidate available channels [161] These protocols aim to

inimise the convergence time to maximise the use of resources

nd to avoid collisions among SUs by identifying and agreeing on

he use of a given common channel among N non-overlapping and

nambiguously labelled channels The main complication in IEEE

0211af systems is given by the wide availability of many different

on-contiguous spectrum bands which cannot be uniformly quan-

ised in single channels and unambiguously labelled in space in or-

er to realise a common reference domain for all the IEEE 80211af

evices In fact the knowledge of the number of channels their com-

on labels the unique IDs of SUs and the number of SUs contending

or the channel assignment are the key factors determining the effec-

iveness and good properties of blind rendezvous protocols Recently

ew classes of rendezvous protocols have been proposed which are

istributed blind and oblivious Oblivious means that available chan-

els may be identified with different labels on behalf of the differ-

nt involved SUs A number of these distributed blind and oblivious

endezvous protocols have been proposed and have been referenced

nd analysed in [162] To conclude solutions for dynamic spectrum

andoff issues (eg in case PUs suddenly appear in a region) must

e resolved in advance to minimise the latency of re-establishing a

unctional communication channel eg by implementing a proactive

lternative rendezvous channel selection in background to ongoing

ommunication processes [163]

Another challenge is to provide cooperative techniques in a dis-

ributed vs centralised implementation by analysing their convergence

nder variable conditions overheads and tradeoffs In some cases as

n vehicular networks the exploitation of scenario characteristics and

actors such as the constrained mobility could help to realise effec-

ive dissemination of sensing information over an extended and pre-

ictable horizon under a cooperative approach [164ndash166] Coopera-

ion could be helpful in the realisation of the cooperative spectrum

ensing in combination with the Geolocation DB approach Another

nteresting direction of research is the investigation of mutual ef-

ects of coexisting cooperation-based vs non-cooperative techniques

n cooperation-based devices the exchange of information could al-

ow a quick convergence to a solution and avoidance of further

nterference [145] Certainly cooperative techniques have desirable

dvantages over non-cooperative ones However tight time synchro-

isation and high coordination overheads are required The common

hannel for cooperation purposes could become a bottleneck and re-

uce the potential advantages of dynamic spectrum allocation The

EEE 80219 foresees the presence of a shared common control chan-

el (CCC) to which devices need to tune in order to gather the state

f the network Solutions such as [141] propose the use of TVWS for

he CCC

A promising direction for research is the adoption of cluster-

ng schemes for secondary devices associated with a hybrid dis-

ributedcooperative vs centralisednon-cooperative approach in an

ffort to reduce overheads while maximising advantages In this

ay secondary nodes could implement cooperation for sensing func-

ions using properly identified common channels and delegate the

entralised decisions and spectrum-DB management to well instru-

ented cluster-leader nodes This concept is incarnated to some ex-

ent by RLSS in IEEE 80211af More recently there have been propos-

ls to build distributed databases that refresh their contents period-

cally by querying the remote spectrum database [167] Here Mas-

er devices (according to Ofcom terminology) periodically cache the

uery replies from the spectrum database in order to reply to and

roadcast the spectrum availability to neighbour Slave devices

Another challenging direction for research is the definition of ac-

urate models and efficient simulation tools enabling dynamic spectrum

nalysis for both rural and urban areas Recent developments of dig-

tal maps simulation tools and propagation models theoretically al-

ow improving the capability of predicting radio propagation effects

n complex scenarios with an acceptable computation time This en-

bling technology could be used in parallel to spectrum DBs to pro-

ide more accurate forecasting of dynamic frequency allocation in

imespace scenarios Another issue to be taken into account is the

ccuracy of the remote spectrum database regarding the channel

vailability estimation Several works have already shown the inac-

uracy of propagation models due to the complexity of parameters

o be considered and differences between modelled and real scenar-

os [132134168169] Better propagation models will make it possible

o more efficiently estimate the interference between devices and

uild Radio Environment Maps (REM) or White Space Maps (WSM)

hat will possibly lead to a more accurate sharing of the radio spec-

rum [144] To this end current research could focus on how to lever-

ge simulation models in order to build more accurate frequency

Bs

In general the adoption of remote spectrum DBs methodology

ushed by national regulators has made the spectrum sensing pro-

ess optional and conservatively realised (eg power control for

ensing-only devices is limited to below 17 dBm EIRP) Thus much

f the research has focused on techniques and solutions to successfully

uery the remote database also in challenging environments such as

ural areas and indoor However since the queries must provide the

osition of the mobile device the accuracy of positioning for mobile

nd handheld devices still needs more appropriate solutions Trian-

ulation can be used to estimate the position of mobile devices with

he accuracy needed by the current regulations outdoor (50 m) How-

ver a low GPS accuracy is possible indoors [170] Certified man-

al positioning can be provided by technicians for static and indoor

evices although limiting mobility Alternatives to costly infrastruc-

ures for positioning (eg based on short range beaconing devices)

ust be deployed in indoor scenarios

Finally regarding spectrum sensing and interference mitigation

ew challenges come from the heterogeneity and different charac-

eristics that networks operating in the TV White Spaces can have

uch as different transmitting power (4 W for static devices 100 mW

or mobile and portable devices and 40 mW for communication

n channels adjacent to occupied ones) different bandwidths (5

0 15 and 20 MHz are currently foreseen for IEEE 80211af IEEE

02154m can have narrower bandwidths or wider ones) and dif-

erent medium access schemes (protocols that use either CSMA or

DMA are required to co-exist in the TV White Spaces) With all this

18 B Bellalta et al Computer Communications 75 (2016) 1ndash25

d

a

T

i

a

c

t

l

b

o

t

a

d

p

a

r

p

n

a

m

t

p

p

d

d

o

t

i

D

w

p

i

a

a

d

b

fi

d

o

t

l

b

t

b

a

i

e

o

o

c

t

t

w

A

c

M

s

c

l

O

w

g

v

t

in place the problem of determining whether another secondary de-

vice is currently transmitting in the same channel is still an open

issue

Spectrum sensing is a very hot topic in the literature with many

proposals that leverage the cooperation between devices to raise the

accuracy of the detection process eg [144171] The current litera-

ture also offers room for proposals that suggest exploiting sensing to

build more accurate spectrum DBs [164172173] Lastly regarding the

interference mitigation to the primary network regulators provide a

conservative approach based on a remote spectrum database to avoid

high risk for the primary users However in particular for the indoor

scenario the concept of secondary usersrsquo transmission in TV Grey

Spaces has been proposed TV Grey Space identifies busy channels

that are formally used by primary users in the area of interest (eg

at the rooftop level of a building for DVB-T) but which could be con-

sidered usable indoors (eg at the basement level) without causing

effective interference to the primary users [134150174] This concept

is based on the generalised spectrum underlay paradigm [175] The

remaining research challenge is to provide TV Grey space access guar-

anteeing a sufficient amount of protection to the primary user while

providing a satisfactory QoS to opportunistic devices eg through

the use of customised and validated propagation models for indoor

TVWS networks and feedback mechanisms to constantly monitor the

interference generated to the primary network

6 Emerging new trends and technologies

In this last section we review three emerging trends that in our

opinion will have a large influence in the conception of future WLANs

as they change the way WLAN protocols and functionalities are de-

veloped implemented tested and integrated with other wireless

networks

61 Programmable Wireless LANs

Especially in the enterprise environment WLAN deployments

need to support a wide range of functionalities and services This is

intrinsically difficult because of the large number of APs that must be

managed which calls for scalable solutions Typical services include

channel assignment load balancing among APs authentication au-

thorisation and accounting (AAA) policy management support for

client mobility and interference coordination Another problem is the

fact that WLAN clients autonomously take several decisions such as

which APs to associate with when to hand-over etc Therefore sup-

porting roaming clients requires the management of a large number

of association states across several APs which is a challenge if sup-

port for real-time hand-over is desired Typically such management

schemes are centralised and most of them are proprietary such as

WLAN controller solutions from Aruba [176] or Cisco [177] although

the IEEE 80211u amendment has been released to allow mobile users

to seamless roam between WiFi networks with automatic authenti-

cation and handoff [178] For example Dyson [179] enables STAs to

send information such as radio channel conditions to a centralised

controller based on a custom API (eg based on Python) As the con-

troller has a centralised view of the network it can enforce a rich

set of policies to control the network also using historical informa-

tion A demo system has been implemented along with applications

such as airtime reservations for specific clients or optimised hand-

offs However Dyson requires STAs to be modified in order to use

those new services offered by the centralised controller TRANTOR

[180] is another example of a centralised management system that

requires changes of clients in order for the infrastructure to gather

information from them in terms of eg interference measurements

In addition control commands enable the infrastructure to exercise

control such as modifying the transmit power or influencing the as-

sociation procedure Clients still use standard CSMA MAC layer for

ata transmission In contrast to DenseAP [181] or DIRAC [182] which

re also based on a centralised controller but do not modify clients

RANTOR exercises much more control on clients by the use of a ded-

cated API which enables a significantly higher gain in coordination

nd thus capacity CENTAUR [183] is another example of a centralised

ontroller that centrally schedules hidden and exposed terminals in

he downlink while using standard CSMA MAC for uplink traffic and

egacy downlink traffic To achieve good performance it uses a fixed

ack-off packet staggering techniques and a hybrid data path that

nly schedules downlink transmissions towards hidden and exposed

erminals centrally All other traffic is sent using standard DCF with

standard back-off procedure In contrast to previous approaches it

oes not require any modifications to the clients but does require data

lane centralisation and it remains questionable how scalable such

solution is for large WLAN deployments with high PHY layer data

ates beyond 100 Mbps However new approaches for fast data plane

rocessing that are currently explored in the virtualisation commu-

ity which move all the packet processing into user space may be

n option to significantly increase speed Designing such manage-

ent systems raises several interesting research questions (a) Cen-

ralised infrastructures may incur high latency but are more simple to

rogram and maintain while designing a completely distributed ap-

roach that operates close to WLAN devices may result in significant

istributed coordination and consistency problems (b) what level of

istributed control is required in order to support the flexibility needs

f future WiFi based networks supporting a high level of mobility

Traditionally WLAN APs are built on proprietary operating sys-

ems that are tightly coupled with the hardware This design makes

t hard to create new applications on top of such networking devices

espite the fact that protocols and mechanisms are available that

ould greatly improve the utility of existing networks those new

rotocols are not deployed because the closed system design makes

t very difficult to extend their functionality An important aspect for

ll such new approaches is therefore how to provide open interfaces

nd open source to speed up innovation As an example Linux-based

evices are completely open source but in order to increase flexi-

ility much more work is needed in the area of open drivers and

rmware In fact the Atheros based ath9k driver has been the main

river of innovation in the open source community because it is the

nly driver that interworks with open firmware In general solutions

hat modify the OS driver focus mainly on programmable network

evel solutions for WLANs such as channel switching or handovers

etween APs While the ath9k driver is an excellent example in how

o speed up innovation it is still very cumbersome to support flexi-

le MAC engine reprogrammability with ath9k In contrast MAClets

nd Wireless MAC processors (WMP) [184] allow a much more flex-

ble reprogramming of MAC functionalities A MAC processor is an

ntity able to execute general MAC commands that specify the MAC

perations through a software-defined state machine The behaviour

f the MAC protocol can therefore be updated at run-time by simply

hanging the sequence in which those commands are executed (ie

he MAClets) As a proof-of-concept the authors of [184] implement

he proposed MAC processor solution in a commodity WLAN hard-

are card extending the basic DCF in three directions piggy-backing

CKs a pseudo TDMA and the use of multiple channels In [185] a

ontrol framework for this WMP system is also proposed to support

AClet code mobility ie for moving loading and activating MAC

oftware programs embedded into ordinary data packets (akin to the

apsule model of traditional active networks)

The difficulty in re-programming networking hardware has also

ed to the concept of Software Defined Networks (SDNs) based on the

penFlow protocol [186] The main idea of SDNs is to extend net-

orking devices with standardised APIs that allow third-party pro-

rammers to flexibly control the data path In addition SDNs pro-

ide higher level abstractions to network designers and programmers

hrough the use of a centralised control plane offered by a network

B Bellalta et al Computer Communications 75 (2016) 1ndash25 19

Fig 11 The CLOUDMAC architecture VAP stands for Virtual Access Point and WT

stands for Wireless Termination

c

a

c

t

m

W

n

t

t

c

p

t

u

h

a

i

e

s

o

c

a

a

f

T

r

s

a

s

p

o

M

C

i

T

t

fl

A

a

a

i

c

t

s

t

e

I

t

w

H

e

b

I

l

d

o

6

r

W

s

l

a

e

e

a

s

b

c

t

(

S

p

c

F

B

t

p

T

e

t

s

u

m

i

e

o

a

A

p

d

t

t

n

p

i

t

t

f

i

ontroller such as NOX [187] which allows reuse of components such

s topology discovery or network access control for different appli-

ations The SDN concept has recently been applied to WLAN archi-

ecture in order to enable fine grained control over mobility manage-

ent and data forwarding focusing on a programmable enterprise

LAN architecture An important part of the SDN architecture is the

etwork controller which provides a centralised view of (parts of)

he SDN enabled network and uses the OpenFlow protocol to install

he forwarding rules on SDN-enabled devices (routers switches ac-

ess points cellular base stations etc)

ODIN [188] is designed to support programmability in enter-

rise WLAN architecture by separating the association state from

he physical AP They implement Light Virtual Access Points (LVAP)

sing a Split-MAC approach where the infrastructure controls the

andover procedure among different WLAN APs By managing

ssociations through SDN controllers ODIN enables proactive mobil-

ty management and load balancing within the SDN enabled WLAN

nterprise network without the need for changes in the client WLAN

tack or IEEE 80211 MAC layer While ODIN requires agents to reside

n the APs that communicate with the Odin Master within the SDN

ontroller CLOUDMAC [189] is based purely on the concept of SDN

nd virtual APs Similar to ODIN CLOUDMAC implements a Split-MAC

pproach but in addition enables the processing of WLAN MAC layer

rames within a co-located Cloud using so-called Virtual APs (VAPs)

he physical APs in CLOUDMAC are lightweight WLAN APs that are

esponsible only for sending out IEEE 80211 based MAC layer ACKs to

tandard WLAN clients and tunnel WLAN MAC layer frames through

n SDN-enabled enterprise WLAN towards the VAPs As association

tates are kept in the VAPs fast mobility is supported using sim-

le OpenFlow forwarding rules Because of the additional processing

f IEEE 80211 WLAN MAC frames in the co-located Cloud CLOUD-

AC has higher latency than a standard WLAN deployment However

LOUDMAC offers a Webservice based API to third party applications

n order to program the enterprise WLAN and enable new services

he architecture of CLOUDMAC is depicted in Fig 11 It has been ex-

ended in [190] to support QoS management and in [191] to support

exible MAC management frame prioritisation based on IEEE 80211

system based on OpenFlow has been recently proposed in [192] to

llow the station to be associated with multiple APs simultaneously

nd to switch between APs with low overhead

An important aspect to consider in order to enable programmabil-

ty is backwards compatibility Several approaches (eg [180]) require

lients to be modified to utilise the features provided This is difficult

o do in practice because it requires changes in the operating system

oftware of all clients If not all clients can utilise those APIs it is ques-

ionable how usable the new architecture will be and how much ben-

fit in terms of aggregate performance such architecture will allow

n contrast the SDN based approaches are interesting in the sense

hat they do not require changes in the client WLAN stack and work

ith the standard IEEE 80211 MAC layer deployed within the clients

owever it remains questionable how scalable such solutions are For

xample an interesting open research topic is to evaluate the scala-

ility of approaches such as CLOUDMAC [189] In addition processing

EEE 80211 MAC frames within co-located private Cloud requires low

atency support from local Cloud solutions (such as OpenStack) in or-

er to reduce the IEEE 80211 MAC processing time which is an area

f active research

2 Prototyping and testing IEEE 80211 enhancements

Most of the new proposals for next-generation WLANs are cur-

ently only evaluated using mathematical analysis and simulation

hile both analysis and simulation are necessary to characterise and

tudy those enhancements in the initial design phase or to consider

arge-scale scenarios it is difficult to consider all practical aspects of

real-world scenario This can sometimes cause significant differ-

nces between what simulations and real experiments show How-

ver real experiments are challenging because of the high complexity

nd costs of building the new hardware and software for each specific

olution to test

To mitigate implementation complexity there are several flexi-

le hardware platforms where low-level MAC mechanisms can be

ompletely implemented in software Examples based on FPGAs are

he USRP (Universal Software Radio Peripheral) [193] and the WARP

Wireless Open-Access Research Platform) [194] As an alternative

ORA (Microsoft Research Software Radio) [195] works on general

urpose computers by taking some advantage from modern multiple

ore systems

A different approach is provided by OpenFWWF [196] Open-

WWF provides an open CSMACA firmware for specific models of

roadcom chipsets so the resulting firmware can be uploaded and

ested in real commercial hardware OpenFWWF implements a sim-

le State Machine (SM) for controlling the hardware in real time

he SM evolution is driven by a main loop that reacts to events by

xecuting specific handlers When a packet originally prepared by

he Linux kernel is ready in the NIC memory handler Packet_Ready

ets up the radio hardware according to the packet metadata sched-

les the transmission and jumps back to the main loop The Trans-

ission Engine (TXE) then takes care of accessing the channel ie

t decrements the back-off counter according to the DCF rules and

ventually starts the actual transmission This triggers the execution

f the TX_frame_now event that prepares the ACK time-out clock

nd finalises the MAC header If the ACK-frame is received or if the

CK time-out expires and the maximum number of attempts for this

acket is reached handler Update_Params resets the contention win-

ow to its minimum value or otherwise doubles it Finally it loads

he back-off counter with a fresh value

In the following we will review a list of selected papers that use

he OpenFWWF firmware and the WARP or USRP platforms to test

ew proposals for WLANs in real scenarios as examples A first im-

lementation of the groupcast mechanisms defined in IEEE 80211aa

s presented in [197] The authors modify the OpenFWWF firmware

o include those functionalities and evaluate them using a 30 STAs

estbed A second example of the use of OpenFWWF is the collision-

ree MAC protocol presented in [198] In [198] the authors design and

mplement in OpenFWWF a MAC protocol that is able to achieve a

20 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 12 LTE to WIFI offloading

a

n

c

t

[

d

a

W

v

t

t

o

b

f

S

g

l

i

I

s

I

b

n

s

m

a

p

d

c

N

o

v

t

t

n

l

i

a

t

a

collision-free operation by waiting a deterministic time after success-

ful transmissions before to start a new transmission This work shows

the experimental performance of a collision-free MAC (CF-MAC) pro-

tocol for WLANs using commercial hardware The results show that

the proposed CF-MAC protocol leads to a better distribution of the

available bandwidth among users a higher throughput and lower

losses than DCF

In [199] the authors use the WARP platform to test a variant of the

DCF ndash called IEEE 80211ec ndash that substitutes control packets such as

the RTS CTS and ACK with short detectable sequences Since those se-

quences are shorter than the control packets and can be detected cor-

rectly even at lower SNRs values a significant gain in performance is

achieved MIDAS (Multiple-Input Distributed Antenna Systems) [200]

implementation using the WARP platform shows the benefits of dis-

tributing the antennas over the area to cover instead of co-locating

them at the AP in terms of capacity when MU-MIMO is employed

The authors also propose a new MAC protocol to benefit from the

spatial reuse that the DAS allows taking as a basis the IEEE 80211ac

Experimental results show that their proposal is able to achieve up

to 200 gains versus the traditional approach where all antennas are

co-located at the AP

The USRP platform has recently been proposed to develop a first

implementation of IEEE 80211ah in order to obtain a preliminary ex-

perimental performance assessment of such a technology because

no commercial off-the-shelf hardware is yet available as the amend-

ment is still in progress [201] The USRP platform is also used in

[202] to evaluate TIMO (Technology Independent Multi-Output) a so-

lution to deal with high-power non-IEEE 80211 interferers in ISM

bands

To conclude this section it is worth mentioning here also to

MAClets and WMP as given such flexibility to implement and dis-

tribute new MAC protocols and other IEEE 80211 functionalities we

believe that if a use-friendly implementation of such a MAC proces-

sors framework is provided it would significantly contribute to the

development and testing of new MAC enhancements by the research

community

63 CellularWLAN interworking

Public hotspots that offer Internet access over a WLAN using IEEE

80211 technology are now nearly ubiquitous It is forecasted that the

cumulative installed base of WiFi hotspots worldwide will amount

to 551 million by 2018 excluding private hotspots (eg WiFi access

points deployed at home) [203] The sharp increase in the availabil-

ity of public WiFi was initially perceived by mobile cellular operators

as a threat due to the additional competition from wireline Internet

service providers or emerging crowdsourced WiFi networks such as

FON5 However as cellular operators are fighting to cope with the ex-

plosion of mobile data traffic created by the rising use of multimedia

content traffic over mobile devices [15] they are also starting to use

WLANs based on the IEEE 80211 technology to offload data from their

core and access networks In general mobile data offloading refers to

the use of complementary network technologies (in licensed or un-

licensed spectrum) for delivering data originally targeted to cellular

networks Intuitively the simplest type of offloading consists of ex-

ploiting connectivity to existing co-located WiFi networks and trans-

ferring data without any delay (Fig 12) Thus this offloading tech-

nique is known as on-the-spot offloading As a consequence of this

new trend the seamless integration of cellular (eg 3GLTE) and WiFi

technologies has attracted significant research interest in recent years

(see [204] for a survey) a few solutions have already been standard-

ised [205206] and roaming between cellular and WiFi is becoming

increasingly transparent to end users CellularWLAN interworking is

5 httpscorpfoncomen

fl

t

f

lso fostered by the support in the evolving 4G standards of heteroge-

eous network deployments (HetNets) in which the existing macro

ells are complemented with a number of small low-power base sta-

ions with the goal of increasing capacity in highly congested areas

207208] It is envisaged that small cells will be based on 4G stan-

ards (eg pico and femto cells) as well as IEEE 80211 technologies

nd multimode base stations that work simultaneously with LTE and

iFi are already entering the market

It is important to point out that LTE standards already support a

ariety of mechanisms that enable data offloading However most of

he existing solutions such as Local IP Access (LIPA) and Selected In-

ernet IP Traffic Offload (SIPTO) focus on data offloading at the core

f the network [205] For instance LIPA allows an IP-enabled mo-

ile device to transfer data to another device in the same pico or

emto cell without passing through the cellular access network while

IPTO enables the routing of selected IP data flows through different

ateways The only LTE offloading mechanism that supports seam-

ess interworking with IEEE 80211-based WLANs is IP Flow Mobil-

ty and Seamless Offload (IFOM) Specifically IFOM relies on Mobile

Pv6 technologies to allow a user terminal to simultaneously route

elected IP flows over different radio access technologies [209210]

n this way a user terminal can offload selected flows to a WLAN

ased on some operator-defined policy while keeping the LTE con-

ection running However to enable such an approach it is neces-

ary to support an entity in the cellular core network that can com-

unicate with the user terminal to exchange information about the

vailability and quality of neighbouring access networks as well as to

rovide the user terminal with predefined rules to manage the han-

over process This entity in current LTE standards is called the Ac-

ess Network Discovery and Selection Function (ANDSF) server [211]

ote that the problem of selecting the best communication technol-

gy in an heterogeneous wireless network has been extensively in-

estigated in the literature both with centralised as well as decen-

ralised approaches (see [212] for a survey) However the use of mul-

iple interfaces in parallel as well as per-flow offloading are relatively

ew concepts Thus the design of scalable and efficient network se-

ection strategies for the ANDSF framework is still an open issue For

nstance three offloading methods suitable for the ANDSF framework

re proposed in [213] which are based on coverage SNR and sys-

em load A reinforcement learning approach is designed in [214] that

llows multimode base stations to autonomously steer their traffic

ows across different access technologies depending on the traffic

ype the usersrsquo QoS requirements the network load and the inter-

erence levels A number of studies have also tried to quantify the

B Bellalta et al Computer Communications 75 (2016) 1ndash25 17

R

n

c

t

l

o

m

e

n

i

n

w

n

a

u

s

p

b

s

m

s

a

d

t

b

a

a

m

a

t

u

8

n

t

d

d

m

f

t

n

d

n

e

r

a

h

b

f

a

c

t

u

i

f

t

d

t

s

i

f

I

l

i

a

n

c

d

I

n

o

t

i

t

e

w

t

c

m

t

a

i

t

q

b

c

a

i

l

i

a

v

t

a

a

c

t

i

t

b

t

t

a

D

p

c

s

o

q

r

p

a

g

t

e

u

d

t

m

n

t

s

f

i

1

8

f

T

endezvous is considered a fundamental problem in generalised CR

etworks at the basis of many communication processes eg in-

luding neighbour discovery routing and broadcast Rendezvous pro-

ocols attempt to establish a new link for communication on a se-

ected and agreed frequency band (channel) identified within a set

f available resources (channels) thus creating the basis for com-

unication between two or more SUs This problem is often exac-

rbated by the high (and unknown) numbers of contending SUs high

umber of available channels and their ambiguous boundaries and

dentifiers and tough co-existence requirements satisfaction in cog-

itive radio systems [161] Preliminary approaches for rendezvous

ere based on the existence of a dedicated Common Control Chan-

el (CCC) agreed among all the SUs or the existence of a centralised

gent (decision maker) assigning channels to any pairs of requesting

sers This approach could be considered in the case of IEEE 80211af

olutions based on the spectrum DB implementation however the

roblem is complicated by the fact that the common channel must

e identified and made available for all the requesting SUs at the

ame time even when these are spread over multiple collision do-

ains in space On the other hand the scalability and reliability is-

ues which are caused by the congestion of CCCs and the vulner-

bility risks of centralised DBs motivate the research of alternative

istributed solutions called blind rendezvous protocols [162] Dis-

ributed solutions have been mainly based on beaconing andor slot-

ased channel hopping algorithms exploiting randomisation sensing

nd discoverysynchronisation messages spread in slotted time over

set of candidate available channels [161] These protocols aim to

inimise the convergence time to maximise the use of resources

nd to avoid collisions among SUs by identifying and agreeing on

he use of a given common channel among N non-overlapping and

nambiguously labelled channels The main complication in IEEE

0211af systems is given by the wide availability of many different

on-contiguous spectrum bands which cannot be uniformly quan-

ised in single channels and unambiguously labelled in space in or-

er to realise a common reference domain for all the IEEE 80211af

evices In fact the knowledge of the number of channels their com-

on labels the unique IDs of SUs and the number of SUs contending

or the channel assignment are the key factors determining the effec-

iveness and good properties of blind rendezvous protocols Recently

ew classes of rendezvous protocols have been proposed which are

istributed blind and oblivious Oblivious means that available chan-

els may be identified with different labels on behalf of the differ-

nt involved SUs A number of these distributed blind and oblivious

endezvous protocols have been proposed and have been referenced

nd analysed in [162] To conclude solutions for dynamic spectrum

andoff issues (eg in case PUs suddenly appear in a region) must

e resolved in advance to minimise the latency of re-establishing a

unctional communication channel eg by implementing a proactive

lternative rendezvous channel selection in background to ongoing

ommunication processes [163]

Another challenge is to provide cooperative techniques in a dis-

ributed vs centralised implementation by analysing their convergence

nder variable conditions overheads and tradeoffs In some cases as

n vehicular networks the exploitation of scenario characteristics and

actors such as the constrained mobility could help to realise effec-

ive dissemination of sensing information over an extended and pre-

ictable horizon under a cooperative approach [164ndash166] Coopera-

ion could be helpful in the realisation of the cooperative spectrum

ensing in combination with the Geolocation DB approach Another

nteresting direction of research is the investigation of mutual ef-

ects of coexisting cooperation-based vs non-cooperative techniques

n cooperation-based devices the exchange of information could al-

ow a quick convergence to a solution and avoidance of further

nterference [145] Certainly cooperative techniques have desirable

dvantages over non-cooperative ones However tight time synchro-

isation and high coordination overheads are required The common

hannel for cooperation purposes could become a bottleneck and re-

uce the potential advantages of dynamic spectrum allocation The

EEE 80219 foresees the presence of a shared common control chan-

el (CCC) to which devices need to tune in order to gather the state

f the network Solutions such as [141] propose the use of TVWS for

he CCC

A promising direction for research is the adoption of cluster-

ng schemes for secondary devices associated with a hybrid dis-

ributedcooperative vs centralisednon-cooperative approach in an

ffort to reduce overheads while maximising advantages In this

ay secondary nodes could implement cooperation for sensing func-

ions using properly identified common channels and delegate the

entralised decisions and spectrum-DB management to well instru-

ented cluster-leader nodes This concept is incarnated to some ex-

ent by RLSS in IEEE 80211af More recently there have been propos-

ls to build distributed databases that refresh their contents period-

cally by querying the remote spectrum database [167] Here Mas-

er devices (according to Ofcom terminology) periodically cache the

uery replies from the spectrum database in order to reply to and

roadcast the spectrum availability to neighbour Slave devices

Another challenging direction for research is the definition of ac-

urate models and efficient simulation tools enabling dynamic spectrum

nalysis for both rural and urban areas Recent developments of dig-

tal maps simulation tools and propagation models theoretically al-

ow improving the capability of predicting radio propagation effects

n complex scenarios with an acceptable computation time This en-

bling technology could be used in parallel to spectrum DBs to pro-

ide more accurate forecasting of dynamic frequency allocation in

imespace scenarios Another issue to be taken into account is the

ccuracy of the remote spectrum database regarding the channel

vailability estimation Several works have already shown the inac-

uracy of propagation models due to the complexity of parameters

o be considered and differences between modelled and real scenar-

os [132134168169] Better propagation models will make it possible

o more efficiently estimate the interference between devices and

uild Radio Environment Maps (REM) or White Space Maps (WSM)

hat will possibly lead to a more accurate sharing of the radio spec-

rum [144] To this end current research could focus on how to lever-

ge simulation models in order to build more accurate frequency

Bs

In general the adoption of remote spectrum DBs methodology

ushed by national regulators has made the spectrum sensing pro-

ess optional and conservatively realised (eg power control for

ensing-only devices is limited to below 17 dBm EIRP) Thus much

f the research has focused on techniques and solutions to successfully

uery the remote database also in challenging environments such as

ural areas and indoor However since the queries must provide the

osition of the mobile device the accuracy of positioning for mobile

nd handheld devices still needs more appropriate solutions Trian-

ulation can be used to estimate the position of mobile devices with

he accuracy needed by the current regulations outdoor (50 m) How-

ver a low GPS accuracy is possible indoors [170] Certified man-

al positioning can be provided by technicians for static and indoor

evices although limiting mobility Alternatives to costly infrastruc-

ures for positioning (eg based on short range beaconing devices)

ust be deployed in indoor scenarios

Finally regarding spectrum sensing and interference mitigation

ew challenges come from the heterogeneity and different charac-

eristics that networks operating in the TV White Spaces can have

uch as different transmitting power (4 W for static devices 100 mW

or mobile and portable devices and 40 mW for communication

n channels adjacent to occupied ones) different bandwidths (5

0 15 and 20 MHz are currently foreseen for IEEE 80211af IEEE

02154m can have narrower bandwidths or wider ones) and dif-

erent medium access schemes (protocols that use either CSMA or

DMA are required to co-exist in the TV White Spaces) With all this

18 B Bellalta et al Computer Communications 75 (2016) 1ndash25

d

a

T

i

a

c

t

l

b

o

t

a

d

p

a

r

p

n

a

m

t

p

p

d

d

o

t

i

D

w

p

i

a

a

d

b

fi

d

o

t

l

b

t

b

a

i

e

o

o

c

t

t

w

A

c

M

s

c

l

O

w

g

v

t

in place the problem of determining whether another secondary de-

vice is currently transmitting in the same channel is still an open

issue

Spectrum sensing is a very hot topic in the literature with many

proposals that leverage the cooperation between devices to raise the

accuracy of the detection process eg [144171] The current litera-

ture also offers room for proposals that suggest exploiting sensing to

build more accurate spectrum DBs [164172173] Lastly regarding the

interference mitigation to the primary network regulators provide a

conservative approach based on a remote spectrum database to avoid

high risk for the primary users However in particular for the indoor

scenario the concept of secondary usersrsquo transmission in TV Grey

Spaces has been proposed TV Grey Space identifies busy channels

that are formally used by primary users in the area of interest (eg

at the rooftop level of a building for DVB-T) but which could be con-

sidered usable indoors (eg at the basement level) without causing

effective interference to the primary users [134150174] This concept

is based on the generalised spectrum underlay paradigm [175] The

remaining research challenge is to provide TV Grey space access guar-

anteeing a sufficient amount of protection to the primary user while

providing a satisfactory QoS to opportunistic devices eg through

the use of customised and validated propagation models for indoor

TVWS networks and feedback mechanisms to constantly monitor the

interference generated to the primary network

6 Emerging new trends and technologies

In this last section we review three emerging trends that in our

opinion will have a large influence in the conception of future WLANs

as they change the way WLAN protocols and functionalities are de-

veloped implemented tested and integrated with other wireless

networks

61 Programmable Wireless LANs

Especially in the enterprise environment WLAN deployments

need to support a wide range of functionalities and services This is

intrinsically difficult because of the large number of APs that must be

managed which calls for scalable solutions Typical services include

channel assignment load balancing among APs authentication au-

thorisation and accounting (AAA) policy management support for

client mobility and interference coordination Another problem is the

fact that WLAN clients autonomously take several decisions such as

which APs to associate with when to hand-over etc Therefore sup-

porting roaming clients requires the management of a large number

of association states across several APs which is a challenge if sup-

port for real-time hand-over is desired Typically such management

schemes are centralised and most of them are proprietary such as

WLAN controller solutions from Aruba [176] or Cisco [177] although

the IEEE 80211u amendment has been released to allow mobile users

to seamless roam between WiFi networks with automatic authenti-

cation and handoff [178] For example Dyson [179] enables STAs to

send information such as radio channel conditions to a centralised

controller based on a custom API (eg based on Python) As the con-

troller has a centralised view of the network it can enforce a rich

set of policies to control the network also using historical informa-

tion A demo system has been implemented along with applications

such as airtime reservations for specific clients or optimised hand-

offs However Dyson requires STAs to be modified in order to use

those new services offered by the centralised controller TRANTOR

[180] is another example of a centralised management system that

requires changes of clients in order for the infrastructure to gather

information from them in terms of eg interference measurements

In addition control commands enable the infrastructure to exercise

control such as modifying the transmit power or influencing the as-

sociation procedure Clients still use standard CSMA MAC layer for

ata transmission In contrast to DenseAP [181] or DIRAC [182] which

re also based on a centralised controller but do not modify clients

RANTOR exercises much more control on clients by the use of a ded-

cated API which enables a significantly higher gain in coordination

nd thus capacity CENTAUR [183] is another example of a centralised

ontroller that centrally schedules hidden and exposed terminals in

he downlink while using standard CSMA MAC for uplink traffic and

egacy downlink traffic To achieve good performance it uses a fixed

ack-off packet staggering techniques and a hybrid data path that

nly schedules downlink transmissions towards hidden and exposed

erminals centrally All other traffic is sent using standard DCF with

standard back-off procedure In contrast to previous approaches it

oes not require any modifications to the clients but does require data

lane centralisation and it remains questionable how scalable such

solution is for large WLAN deployments with high PHY layer data

ates beyond 100 Mbps However new approaches for fast data plane

rocessing that are currently explored in the virtualisation commu-

ity which move all the packet processing into user space may be

n option to significantly increase speed Designing such manage-

ent systems raises several interesting research questions (a) Cen-

ralised infrastructures may incur high latency but are more simple to

rogram and maintain while designing a completely distributed ap-

roach that operates close to WLAN devices may result in significant

istributed coordination and consistency problems (b) what level of

istributed control is required in order to support the flexibility needs

f future WiFi based networks supporting a high level of mobility

Traditionally WLAN APs are built on proprietary operating sys-

ems that are tightly coupled with the hardware This design makes

t hard to create new applications on top of such networking devices

espite the fact that protocols and mechanisms are available that

ould greatly improve the utility of existing networks those new

rotocols are not deployed because the closed system design makes

t very difficult to extend their functionality An important aspect for

ll such new approaches is therefore how to provide open interfaces

nd open source to speed up innovation As an example Linux-based

evices are completely open source but in order to increase flexi-

ility much more work is needed in the area of open drivers and

rmware In fact the Atheros based ath9k driver has been the main

river of innovation in the open source community because it is the

nly driver that interworks with open firmware In general solutions

hat modify the OS driver focus mainly on programmable network

evel solutions for WLANs such as channel switching or handovers

etween APs While the ath9k driver is an excellent example in how

o speed up innovation it is still very cumbersome to support flexi-

le MAC engine reprogrammability with ath9k In contrast MAClets

nd Wireless MAC processors (WMP) [184] allow a much more flex-

ble reprogramming of MAC functionalities A MAC processor is an

ntity able to execute general MAC commands that specify the MAC

perations through a software-defined state machine The behaviour

f the MAC protocol can therefore be updated at run-time by simply

hanging the sequence in which those commands are executed (ie

he MAClets) As a proof-of-concept the authors of [184] implement

he proposed MAC processor solution in a commodity WLAN hard-

are card extending the basic DCF in three directions piggy-backing

CKs a pseudo TDMA and the use of multiple channels In [185] a

ontrol framework for this WMP system is also proposed to support

AClet code mobility ie for moving loading and activating MAC

oftware programs embedded into ordinary data packets (akin to the

apsule model of traditional active networks)

The difficulty in re-programming networking hardware has also

ed to the concept of Software Defined Networks (SDNs) based on the

penFlow protocol [186] The main idea of SDNs is to extend net-

orking devices with standardised APIs that allow third-party pro-

rammers to flexibly control the data path In addition SDNs pro-

ide higher level abstractions to network designers and programmers

hrough the use of a centralised control plane offered by a network

B Bellalta et al Computer Communications 75 (2016) 1ndash25 19

Fig 11 The CLOUDMAC architecture VAP stands for Virtual Access Point and WT

stands for Wireless Termination

c

a

c

t

m

W

n

t

t

c

p

t

u

h

a

i

e

s

o

c

a

a

f

T

r

s

a

s

p

o

M

C

i

T

t

fl

A

a

a

i

c

t

s

t

e

I

t

w

H

e

b

I

l

d

o

6

r

W

s

l

a

e

e

a

s

b

c

t

(

S

p

c

F

B

t

p

T

e

t

s

u

m

i

e

o

a

A

p

d

t

t

n

p

i

t

t

f

i

ontroller such as NOX [187] which allows reuse of components such

s topology discovery or network access control for different appli-

ations The SDN concept has recently been applied to WLAN archi-

ecture in order to enable fine grained control over mobility manage-

ent and data forwarding focusing on a programmable enterprise

LAN architecture An important part of the SDN architecture is the

etwork controller which provides a centralised view of (parts of)

he SDN enabled network and uses the OpenFlow protocol to install

he forwarding rules on SDN-enabled devices (routers switches ac-

ess points cellular base stations etc)

ODIN [188] is designed to support programmability in enter-

rise WLAN architecture by separating the association state from

he physical AP They implement Light Virtual Access Points (LVAP)

sing a Split-MAC approach where the infrastructure controls the

andover procedure among different WLAN APs By managing

ssociations through SDN controllers ODIN enables proactive mobil-

ty management and load balancing within the SDN enabled WLAN

nterprise network without the need for changes in the client WLAN

tack or IEEE 80211 MAC layer While ODIN requires agents to reside

n the APs that communicate with the Odin Master within the SDN

ontroller CLOUDMAC [189] is based purely on the concept of SDN

nd virtual APs Similar to ODIN CLOUDMAC implements a Split-MAC

pproach but in addition enables the processing of WLAN MAC layer

rames within a co-located Cloud using so-called Virtual APs (VAPs)

he physical APs in CLOUDMAC are lightweight WLAN APs that are

esponsible only for sending out IEEE 80211 based MAC layer ACKs to

tandard WLAN clients and tunnel WLAN MAC layer frames through

n SDN-enabled enterprise WLAN towards the VAPs As association

tates are kept in the VAPs fast mobility is supported using sim-

le OpenFlow forwarding rules Because of the additional processing

f IEEE 80211 WLAN MAC frames in the co-located Cloud CLOUD-

AC has higher latency than a standard WLAN deployment However

LOUDMAC offers a Webservice based API to third party applications

n order to program the enterprise WLAN and enable new services

he architecture of CLOUDMAC is depicted in Fig 11 It has been ex-

ended in [190] to support QoS management and in [191] to support

exible MAC management frame prioritisation based on IEEE 80211

system based on OpenFlow has been recently proposed in [192] to

llow the station to be associated with multiple APs simultaneously

nd to switch between APs with low overhead

An important aspect to consider in order to enable programmabil-

ty is backwards compatibility Several approaches (eg [180]) require

lients to be modified to utilise the features provided This is difficult

o do in practice because it requires changes in the operating system

oftware of all clients If not all clients can utilise those APIs it is ques-

ionable how usable the new architecture will be and how much ben-

fit in terms of aggregate performance such architecture will allow

n contrast the SDN based approaches are interesting in the sense

hat they do not require changes in the client WLAN stack and work

ith the standard IEEE 80211 MAC layer deployed within the clients

owever it remains questionable how scalable such solutions are For

xample an interesting open research topic is to evaluate the scala-

ility of approaches such as CLOUDMAC [189] In addition processing

EEE 80211 MAC frames within co-located private Cloud requires low

atency support from local Cloud solutions (such as OpenStack) in or-

er to reduce the IEEE 80211 MAC processing time which is an area

f active research

2 Prototyping and testing IEEE 80211 enhancements

Most of the new proposals for next-generation WLANs are cur-

ently only evaluated using mathematical analysis and simulation

hile both analysis and simulation are necessary to characterise and

tudy those enhancements in the initial design phase or to consider

arge-scale scenarios it is difficult to consider all practical aspects of

real-world scenario This can sometimes cause significant differ-

nces between what simulations and real experiments show How-

ver real experiments are challenging because of the high complexity

nd costs of building the new hardware and software for each specific

olution to test

To mitigate implementation complexity there are several flexi-

le hardware platforms where low-level MAC mechanisms can be

ompletely implemented in software Examples based on FPGAs are

he USRP (Universal Software Radio Peripheral) [193] and the WARP

Wireless Open-Access Research Platform) [194] As an alternative

ORA (Microsoft Research Software Radio) [195] works on general

urpose computers by taking some advantage from modern multiple

ore systems

A different approach is provided by OpenFWWF [196] Open-

WWF provides an open CSMACA firmware for specific models of

roadcom chipsets so the resulting firmware can be uploaded and

ested in real commercial hardware OpenFWWF implements a sim-

le State Machine (SM) for controlling the hardware in real time

he SM evolution is driven by a main loop that reacts to events by

xecuting specific handlers When a packet originally prepared by

he Linux kernel is ready in the NIC memory handler Packet_Ready

ets up the radio hardware according to the packet metadata sched-

les the transmission and jumps back to the main loop The Trans-

ission Engine (TXE) then takes care of accessing the channel ie

t decrements the back-off counter according to the DCF rules and

ventually starts the actual transmission This triggers the execution

f the TX_frame_now event that prepares the ACK time-out clock

nd finalises the MAC header If the ACK-frame is received or if the

CK time-out expires and the maximum number of attempts for this

acket is reached handler Update_Params resets the contention win-

ow to its minimum value or otherwise doubles it Finally it loads

he back-off counter with a fresh value

In the following we will review a list of selected papers that use

he OpenFWWF firmware and the WARP or USRP platforms to test

ew proposals for WLANs in real scenarios as examples A first im-

lementation of the groupcast mechanisms defined in IEEE 80211aa

s presented in [197] The authors modify the OpenFWWF firmware

o include those functionalities and evaluate them using a 30 STAs

estbed A second example of the use of OpenFWWF is the collision-

ree MAC protocol presented in [198] In [198] the authors design and

mplement in OpenFWWF a MAC protocol that is able to achieve a

20 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 12 LTE to WIFI offloading

a

n

c

t

[

d

a

W

v

t

t

o

b

f

S

g

l

i

I

s

I

b

n

s

m

a

p

d

c

N

o

v

t

t

n

l

i

a

t

a

collision-free operation by waiting a deterministic time after success-

ful transmissions before to start a new transmission This work shows

the experimental performance of a collision-free MAC (CF-MAC) pro-

tocol for WLANs using commercial hardware The results show that

the proposed CF-MAC protocol leads to a better distribution of the

available bandwidth among users a higher throughput and lower

losses than DCF

In [199] the authors use the WARP platform to test a variant of the

DCF ndash called IEEE 80211ec ndash that substitutes control packets such as

the RTS CTS and ACK with short detectable sequences Since those se-

quences are shorter than the control packets and can be detected cor-

rectly even at lower SNRs values a significant gain in performance is

achieved MIDAS (Multiple-Input Distributed Antenna Systems) [200]

implementation using the WARP platform shows the benefits of dis-

tributing the antennas over the area to cover instead of co-locating

them at the AP in terms of capacity when MU-MIMO is employed

The authors also propose a new MAC protocol to benefit from the

spatial reuse that the DAS allows taking as a basis the IEEE 80211ac

Experimental results show that their proposal is able to achieve up

to 200 gains versus the traditional approach where all antennas are

co-located at the AP

The USRP platform has recently been proposed to develop a first

implementation of IEEE 80211ah in order to obtain a preliminary ex-

perimental performance assessment of such a technology because

no commercial off-the-shelf hardware is yet available as the amend-

ment is still in progress [201] The USRP platform is also used in

[202] to evaluate TIMO (Technology Independent Multi-Output) a so-

lution to deal with high-power non-IEEE 80211 interferers in ISM

bands

To conclude this section it is worth mentioning here also to

MAClets and WMP as given such flexibility to implement and dis-

tribute new MAC protocols and other IEEE 80211 functionalities we

believe that if a use-friendly implementation of such a MAC proces-

sors framework is provided it would significantly contribute to the

development and testing of new MAC enhancements by the research

community

63 CellularWLAN interworking

Public hotspots that offer Internet access over a WLAN using IEEE

80211 technology are now nearly ubiquitous It is forecasted that the

cumulative installed base of WiFi hotspots worldwide will amount

to 551 million by 2018 excluding private hotspots (eg WiFi access

points deployed at home) [203] The sharp increase in the availabil-

ity of public WiFi was initially perceived by mobile cellular operators

as a threat due to the additional competition from wireline Internet

service providers or emerging crowdsourced WiFi networks such as

FON5 However as cellular operators are fighting to cope with the ex-

plosion of mobile data traffic created by the rising use of multimedia

content traffic over mobile devices [15] they are also starting to use

WLANs based on the IEEE 80211 technology to offload data from their

core and access networks In general mobile data offloading refers to

the use of complementary network technologies (in licensed or un-

licensed spectrum) for delivering data originally targeted to cellular

networks Intuitively the simplest type of offloading consists of ex-

ploiting connectivity to existing co-located WiFi networks and trans-

ferring data without any delay (Fig 12) Thus this offloading tech-

nique is known as on-the-spot offloading As a consequence of this

new trend the seamless integration of cellular (eg 3GLTE) and WiFi

technologies has attracted significant research interest in recent years

(see [204] for a survey) a few solutions have already been standard-

ised [205206] and roaming between cellular and WiFi is becoming

increasingly transparent to end users CellularWLAN interworking is

5 httpscorpfoncomen

fl

t

f

lso fostered by the support in the evolving 4G standards of heteroge-

eous network deployments (HetNets) in which the existing macro

ells are complemented with a number of small low-power base sta-

ions with the goal of increasing capacity in highly congested areas

207208] It is envisaged that small cells will be based on 4G stan-

ards (eg pico and femto cells) as well as IEEE 80211 technologies

nd multimode base stations that work simultaneously with LTE and

iFi are already entering the market

It is important to point out that LTE standards already support a

ariety of mechanisms that enable data offloading However most of

he existing solutions such as Local IP Access (LIPA) and Selected In-

ernet IP Traffic Offload (SIPTO) focus on data offloading at the core

f the network [205] For instance LIPA allows an IP-enabled mo-

ile device to transfer data to another device in the same pico or

emto cell without passing through the cellular access network while

IPTO enables the routing of selected IP data flows through different

ateways The only LTE offloading mechanism that supports seam-

ess interworking with IEEE 80211-based WLANs is IP Flow Mobil-

ty and Seamless Offload (IFOM) Specifically IFOM relies on Mobile

Pv6 technologies to allow a user terminal to simultaneously route

elected IP flows over different radio access technologies [209210]

n this way a user terminal can offload selected flows to a WLAN

ased on some operator-defined policy while keeping the LTE con-

ection running However to enable such an approach it is neces-

ary to support an entity in the cellular core network that can com-

unicate with the user terminal to exchange information about the

vailability and quality of neighbouring access networks as well as to

rovide the user terminal with predefined rules to manage the han-

over process This entity in current LTE standards is called the Ac-

ess Network Discovery and Selection Function (ANDSF) server [211]

ote that the problem of selecting the best communication technol-

gy in an heterogeneous wireless network has been extensively in-

estigated in the literature both with centralised as well as decen-

ralised approaches (see [212] for a survey) However the use of mul-

iple interfaces in parallel as well as per-flow offloading are relatively

ew concepts Thus the design of scalable and efficient network se-

ection strategies for the ANDSF framework is still an open issue For

nstance three offloading methods suitable for the ANDSF framework

re proposed in [213] which are based on coverage SNR and sys-

em load A reinforcement learning approach is designed in [214] that

llows multimode base stations to autonomously steer their traffic

ows across different access technologies depending on the traffic

ype the usersrsquo QoS requirements the network load and the inter-

erence levels A number of studies have also tried to quantify the

18 B Bellalta et al Computer Communications 75 (2016) 1ndash25

d

a

T

i

a

c

t

l

b

o

t

a

d

p

a

r

p

n

a

m

t

p

p

d

d

o

t

i

D

w

p

i

a

a

d

b

fi

d

o

t

l

b

t

b

a

i

e

o

o

c

t

t

w

A

c

M

s

c

l

O

w

g

v

t

in place the problem of determining whether another secondary de-

vice is currently transmitting in the same channel is still an open

issue

Spectrum sensing is a very hot topic in the literature with many

proposals that leverage the cooperation between devices to raise the

accuracy of the detection process eg [144171] The current litera-

ture also offers room for proposals that suggest exploiting sensing to

build more accurate spectrum DBs [164172173] Lastly regarding the

interference mitigation to the primary network regulators provide a

conservative approach based on a remote spectrum database to avoid

high risk for the primary users However in particular for the indoor

scenario the concept of secondary usersrsquo transmission in TV Grey

Spaces has been proposed TV Grey Space identifies busy channels

that are formally used by primary users in the area of interest (eg

at the rooftop level of a building for DVB-T) but which could be con-

sidered usable indoors (eg at the basement level) without causing

effective interference to the primary users [134150174] This concept

is based on the generalised spectrum underlay paradigm [175] The

remaining research challenge is to provide TV Grey space access guar-

anteeing a sufficient amount of protection to the primary user while

providing a satisfactory QoS to opportunistic devices eg through

the use of customised and validated propagation models for indoor

TVWS networks and feedback mechanisms to constantly monitor the

interference generated to the primary network

6 Emerging new trends and technologies

In this last section we review three emerging trends that in our

opinion will have a large influence in the conception of future WLANs

as they change the way WLAN protocols and functionalities are de-

veloped implemented tested and integrated with other wireless

networks

61 Programmable Wireless LANs

Especially in the enterprise environment WLAN deployments

need to support a wide range of functionalities and services This is

intrinsically difficult because of the large number of APs that must be

managed which calls for scalable solutions Typical services include

channel assignment load balancing among APs authentication au-

thorisation and accounting (AAA) policy management support for

client mobility and interference coordination Another problem is the

fact that WLAN clients autonomously take several decisions such as

which APs to associate with when to hand-over etc Therefore sup-

porting roaming clients requires the management of a large number

of association states across several APs which is a challenge if sup-

port for real-time hand-over is desired Typically such management

schemes are centralised and most of them are proprietary such as

WLAN controller solutions from Aruba [176] or Cisco [177] although

the IEEE 80211u amendment has been released to allow mobile users

to seamless roam between WiFi networks with automatic authenti-

cation and handoff [178] For example Dyson [179] enables STAs to

send information such as radio channel conditions to a centralised

controller based on a custom API (eg based on Python) As the con-

troller has a centralised view of the network it can enforce a rich

set of policies to control the network also using historical informa-

tion A demo system has been implemented along with applications

such as airtime reservations for specific clients or optimised hand-

offs However Dyson requires STAs to be modified in order to use

those new services offered by the centralised controller TRANTOR

[180] is another example of a centralised management system that

requires changes of clients in order for the infrastructure to gather

information from them in terms of eg interference measurements

In addition control commands enable the infrastructure to exercise

control such as modifying the transmit power or influencing the as-

sociation procedure Clients still use standard CSMA MAC layer for

ata transmission In contrast to DenseAP [181] or DIRAC [182] which

re also based on a centralised controller but do not modify clients

RANTOR exercises much more control on clients by the use of a ded-

cated API which enables a significantly higher gain in coordination

nd thus capacity CENTAUR [183] is another example of a centralised

ontroller that centrally schedules hidden and exposed terminals in

he downlink while using standard CSMA MAC for uplink traffic and

egacy downlink traffic To achieve good performance it uses a fixed

ack-off packet staggering techniques and a hybrid data path that

nly schedules downlink transmissions towards hidden and exposed

erminals centrally All other traffic is sent using standard DCF with

standard back-off procedure In contrast to previous approaches it

oes not require any modifications to the clients but does require data

lane centralisation and it remains questionable how scalable such

solution is for large WLAN deployments with high PHY layer data

ates beyond 100 Mbps However new approaches for fast data plane

rocessing that are currently explored in the virtualisation commu-

ity which move all the packet processing into user space may be

n option to significantly increase speed Designing such manage-

ent systems raises several interesting research questions (a) Cen-

ralised infrastructures may incur high latency but are more simple to

rogram and maintain while designing a completely distributed ap-

roach that operates close to WLAN devices may result in significant

istributed coordination and consistency problems (b) what level of

istributed control is required in order to support the flexibility needs

f future WiFi based networks supporting a high level of mobility

Traditionally WLAN APs are built on proprietary operating sys-

ems that are tightly coupled with the hardware This design makes

t hard to create new applications on top of such networking devices

espite the fact that protocols and mechanisms are available that

ould greatly improve the utility of existing networks those new

rotocols are not deployed because the closed system design makes

t very difficult to extend their functionality An important aspect for

ll such new approaches is therefore how to provide open interfaces

nd open source to speed up innovation As an example Linux-based

evices are completely open source but in order to increase flexi-

ility much more work is needed in the area of open drivers and

rmware In fact the Atheros based ath9k driver has been the main

river of innovation in the open source community because it is the

nly driver that interworks with open firmware In general solutions

hat modify the OS driver focus mainly on programmable network

evel solutions for WLANs such as channel switching or handovers

etween APs While the ath9k driver is an excellent example in how

o speed up innovation it is still very cumbersome to support flexi-

le MAC engine reprogrammability with ath9k In contrast MAClets

nd Wireless MAC processors (WMP) [184] allow a much more flex-

ble reprogramming of MAC functionalities A MAC processor is an

ntity able to execute general MAC commands that specify the MAC

perations through a software-defined state machine The behaviour

f the MAC protocol can therefore be updated at run-time by simply

hanging the sequence in which those commands are executed (ie

he MAClets) As a proof-of-concept the authors of [184] implement

he proposed MAC processor solution in a commodity WLAN hard-

are card extending the basic DCF in three directions piggy-backing

CKs a pseudo TDMA and the use of multiple channels In [185] a

ontrol framework for this WMP system is also proposed to support

AClet code mobility ie for moving loading and activating MAC

oftware programs embedded into ordinary data packets (akin to the

apsule model of traditional active networks)

The difficulty in re-programming networking hardware has also

ed to the concept of Software Defined Networks (SDNs) based on the

penFlow protocol [186] The main idea of SDNs is to extend net-

orking devices with standardised APIs that allow third-party pro-

rammers to flexibly control the data path In addition SDNs pro-

ide higher level abstractions to network designers and programmers

hrough the use of a centralised control plane offered by a network

B Bellalta et al Computer Communications 75 (2016) 1ndash25 19

Fig 11 The CLOUDMAC architecture VAP stands for Virtual Access Point and WT

stands for Wireless Termination

c

a

c

t

m

W

n

t

t

c

p

t

u

h

a

i

e

s

o

c

a

a

f

T

r

s

a

s

p

o

M

C

i

T

t

fl

A

a

a

i

c

t

s

t

e

I

t

w

H

e

b

I

l

d

o

6

r

W

s

l

a

e

e

a

s

b

c

t

(

S

p

c

F

B

t

p

T

e

t

s

u

m

i

e

o

a

A

p

d

t

t

n

p

i

t

t

f

i

ontroller such as NOX [187] which allows reuse of components such

s topology discovery or network access control for different appli-

ations The SDN concept has recently been applied to WLAN archi-

ecture in order to enable fine grained control over mobility manage-

ent and data forwarding focusing on a programmable enterprise

LAN architecture An important part of the SDN architecture is the

etwork controller which provides a centralised view of (parts of)

he SDN enabled network and uses the OpenFlow protocol to install

he forwarding rules on SDN-enabled devices (routers switches ac-

ess points cellular base stations etc)

ODIN [188] is designed to support programmability in enter-

rise WLAN architecture by separating the association state from

he physical AP They implement Light Virtual Access Points (LVAP)

sing a Split-MAC approach where the infrastructure controls the

andover procedure among different WLAN APs By managing

ssociations through SDN controllers ODIN enables proactive mobil-

ty management and load balancing within the SDN enabled WLAN

nterprise network without the need for changes in the client WLAN

tack or IEEE 80211 MAC layer While ODIN requires agents to reside

n the APs that communicate with the Odin Master within the SDN

ontroller CLOUDMAC [189] is based purely on the concept of SDN

nd virtual APs Similar to ODIN CLOUDMAC implements a Split-MAC

pproach but in addition enables the processing of WLAN MAC layer

rames within a co-located Cloud using so-called Virtual APs (VAPs)

he physical APs in CLOUDMAC are lightweight WLAN APs that are

esponsible only for sending out IEEE 80211 based MAC layer ACKs to

tandard WLAN clients and tunnel WLAN MAC layer frames through

n SDN-enabled enterprise WLAN towards the VAPs As association

tates are kept in the VAPs fast mobility is supported using sim-

le OpenFlow forwarding rules Because of the additional processing

f IEEE 80211 WLAN MAC frames in the co-located Cloud CLOUD-

AC has higher latency than a standard WLAN deployment However

LOUDMAC offers a Webservice based API to third party applications

n order to program the enterprise WLAN and enable new services

he architecture of CLOUDMAC is depicted in Fig 11 It has been ex-

ended in [190] to support QoS management and in [191] to support

exible MAC management frame prioritisation based on IEEE 80211

system based on OpenFlow has been recently proposed in [192] to

llow the station to be associated with multiple APs simultaneously

nd to switch between APs with low overhead

An important aspect to consider in order to enable programmabil-

ty is backwards compatibility Several approaches (eg [180]) require

lients to be modified to utilise the features provided This is difficult

o do in practice because it requires changes in the operating system

oftware of all clients If not all clients can utilise those APIs it is ques-

ionable how usable the new architecture will be and how much ben-

fit in terms of aggregate performance such architecture will allow

n contrast the SDN based approaches are interesting in the sense

hat they do not require changes in the client WLAN stack and work

ith the standard IEEE 80211 MAC layer deployed within the clients

owever it remains questionable how scalable such solutions are For

xample an interesting open research topic is to evaluate the scala-

ility of approaches such as CLOUDMAC [189] In addition processing

EEE 80211 MAC frames within co-located private Cloud requires low

atency support from local Cloud solutions (such as OpenStack) in or-

er to reduce the IEEE 80211 MAC processing time which is an area

f active research

2 Prototyping and testing IEEE 80211 enhancements

Most of the new proposals for next-generation WLANs are cur-

ently only evaluated using mathematical analysis and simulation

hile both analysis and simulation are necessary to characterise and

tudy those enhancements in the initial design phase or to consider

arge-scale scenarios it is difficult to consider all practical aspects of

real-world scenario This can sometimes cause significant differ-

nces between what simulations and real experiments show How-

ver real experiments are challenging because of the high complexity

nd costs of building the new hardware and software for each specific

olution to test

To mitigate implementation complexity there are several flexi-

le hardware platforms where low-level MAC mechanisms can be

ompletely implemented in software Examples based on FPGAs are

he USRP (Universal Software Radio Peripheral) [193] and the WARP

Wireless Open-Access Research Platform) [194] As an alternative

ORA (Microsoft Research Software Radio) [195] works on general

urpose computers by taking some advantage from modern multiple

ore systems

A different approach is provided by OpenFWWF [196] Open-

WWF provides an open CSMACA firmware for specific models of

roadcom chipsets so the resulting firmware can be uploaded and

ested in real commercial hardware OpenFWWF implements a sim-

le State Machine (SM) for controlling the hardware in real time

he SM evolution is driven by a main loop that reacts to events by

xecuting specific handlers When a packet originally prepared by

he Linux kernel is ready in the NIC memory handler Packet_Ready

ets up the radio hardware according to the packet metadata sched-

les the transmission and jumps back to the main loop The Trans-

ission Engine (TXE) then takes care of accessing the channel ie

t decrements the back-off counter according to the DCF rules and

ventually starts the actual transmission This triggers the execution

f the TX_frame_now event that prepares the ACK time-out clock

nd finalises the MAC header If the ACK-frame is received or if the

CK time-out expires and the maximum number of attempts for this

acket is reached handler Update_Params resets the contention win-

ow to its minimum value or otherwise doubles it Finally it loads

he back-off counter with a fresh value

In the following we will review a list of selected papers that use

he OpenFWWF firmware and the WARP or USRP platforms to test

ew proposals for WLANs in real scenarios as examples A first im-

lementation of the groupcast mechanisms defined in IEEE 80211aa

s presented in [197] The authors modify the OpenFWWF firmware

o include those functionalities and evaluate them using a 30 STAs

estbed A second example of the use of OpenFWWF is the collision-

ree MAC protocol presented in [198] In [198] the authors design and

mplement in OpenFWWF a MAC protocol that is able to achieve a

20 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 12 LTE to WIFI offloading

a

n

c

t

[

d

a

W

v

t

t

o

b

f

S

g

l

i

I

s

I

b

n

s

m

a

p

d

c

N

o

v

t

t

n

l

i

a

t

a

collision-free operation by waiting a deterministic time after success-

ful transmissions before to start a new transmission This work shows

the experimental performance of a collision-free MAC (CF-MAC) pro-

tocol for WLANs using commercial hardware The results show that

the proposed CF-MAC protocol leads to a better distribution of the

available bandwidth among users a higher throughput and lower

losses than DCF

In [199] the authors use the WARP platform to test a variant of the

DCF ndash called IEEE 80211ec ndash that substitutes control packets such as

the RTS CTS and ACK with short detectable sequences Since those se-

quences are shorter than the control packets and can be detected cor-

rectly even at lower SNRs values a significant gain in performance is

achieved MIDAS (Multiple-Input Distributed Antenna Systems) [200]

implementation using the WARP platform shows the benefits of dis-

tributing the antennas over the area to cover instead of co-locating

them at the AP in terms of capacity when MU-MIMO is employed

The authors also propose a new MAC protocol to benefit from the

spatial reuse that the DAS allows taking as a basis the IEEE 80211ac

Experimental results show that their proposal is able to achieve up

to 200 gains versus the traditional approach where all antennas are

co-located at the AP

The USRP platform has recently been proposed to develop a first

implementation of IEEE 80211ah in order to obtain a preliminary ex-

perimental performance assessment of such a technology because

no commercial off-the-shelf hardware is yet available as the amend-

ment is still in progress [201] The USRP platform is also used in

[202] to evaluate TIMO (Technology Independent Multi-Output) a so-

lution to deal with high-power non-IEEE 80211 interferers in ISM

bands

To conclude this section it is worth mentioning here also to

MAClets and WMP as given such flexibility to implement and dis-

tribute new MAC protocols and other IEEE 80211 functionalities we

believe that if a use-friendly implementation of such a MAC proces-

sors framework is provided it would significantly contribute to the

development and testing of new MAC enhancements by the research

community

63 CellularWLAN interworking

Public hotspots that offer Internet access over a WLAN using IEEE

80211 technology are now nearly ubiquitous It is forecasted that the

cumulative installed base of WiFi hotspots worldwide will amount

to 551 million by 2018 excluding private hotspots (eg WiFi access

points deployed at home) [203] The sharp increase in the availabil-

ity of public WiFi was initially perceived by mobile cellular operators

as a threat due to the additional competition from wireline Internet

service providers or emerging crowdsourced WiFi networks such as

FON5 However as cellular operators are fighting to cope with the ex-

plosion of mobile data traffic created by the rising use of multimedia

content traffic over mobile devices [15] they are also starting to use

WLANs based on the IEEE 80211 technology to offload data from their

core and access networks In general mobile data offloading refers to

the use of complementary network technologies (in licensed or un-

licensed spectrum) for delivering data originally targeted to cellular

networks Intuitively the simplest type of offloading consists of ex-

ploiting connectivity to existing co-located WiFi networks and trans-

ferring data without any delay (Fig 12) Thus this offloading tech-

nique is known as on-the-spot offloading As a consequence of this

new trend the seamless integration of cellular (eg 3GLTE) and WiFi

technologies has attracted significant research interest in recent years

(see [204] for a survey) a few solutions have already been standard-

ised [205206] and roaming between cellular and WiFi is becoming

increasingly transparent to end users CellularWLAN interworking is

5 httpscorpfoncomen

fl

t

f

lso fostered by the support in the evolving 4G standards of heteroge-

eous network deployments (HetNets) in which the existing macro

ells are complemented with a number of small low-power base sta-

ions with the goal of increasing capacity in highly congested areas

207208] It is envisaged that small cells will be based on 4G stan-

ards (eg pico and femto cells) as well as IEEE 80211 technologies

nd multimode base stations that work simultaneously with LTE and

iFi are already entering the market

It is important to point out that LTE standards already support a

ariety of mechanisms that enable data offloading However most of

he existing solutions such as Local IP Access (LIPA) and Selected In-

ernet IP Traffic Offload (SIPTO) focus on data offloading at the core

f the network [205] For instance LIPA allows an IP-enabled mo-

ile device to transfer data to another device in the same pico or

emto cell without passing through the cellular access network while

IPTO enables the routing of selected IP data flows through different

ateways The only LTE offloading mechanism that supports seam-

ess interworking with IEEE 80211-based WLANs is IP Flow Mobil-

ty and Seamless Offload (IFOM) Specifically IFOM relies on Mobile

Pv6 technologies to allow a user terminal to simultaneously route

elected IP flows over different radio access technologies [209210]

n this way a user terminal can offload selected flows to a WLAN

ased on some operator-defined policy while keeping the LTE con-

ection running However to enable such an approach it is neces-

ary to support an entity in the cellular core network that can com-

unicate with the user terminal to exchange information about the

vailability and quality of neighbouring access networks as well as to

rovide the user terminal with predefined rules to manage the han-

over process This entity in current LTE standards is called the Ac-

ess Network Discovery and Selection Function (ANDSF) server [211]

ote that the problem of selecting the best communication technol-

gy in an heterogeneous wireless network has been extensively in-

estigated in the literature both with centralised as well as decen-

ralised approaches (see [212] for a survey) However the use of mul-

iple interfaces in parallel as well as per-flow offloading are relatively

ew concepts Thus the design of scalable and efficient network se-

ection strategies for the ANDSF framework is still an open issue For

nstance three offloading methods suitable for the ANDSF framework

re proposed in [213] which are based on coverage SNR and sys-

em load A reinforcement learning approach is designed in [214] that

llows multimode base stations to autonomously steer their traffic

ows across different access technologies depending on the traffic

ype the usersrsquo QoS requirements the network load and the inter-

erence levels A number of studies have also tried to quantify the

B Bellalta et al Computer Communications 75 (2016) 1ndash25 19

Fig 11 The CLOUDMAC architecture VAP stands for Virtual Access Point and WT

stands for Wireless Termination

c

a

c

t

m

W

n

t

t

c

p

t

u

h

a

i

e

s

o

c

a

a

f

T

r

s

a

s

p

o

M

C

i

T

t

fl

A

a

a

i

c

t

s

t

e

I

t

w

H

e

b

I

l

d

o

6

r

W

s

l

a

e

e

a

s

b

c

t

(

S

p

c

F

B

t

p

T

e

t

s

u

m

i

e

o

a

A

p

d

t

t

n

p

i

t

t

f

i

ontroller such as NOX [187] which allows reuse of components such

s topology discovery or network access control for different appli-

ations The SDN concept has recently been applied to WLAN archi-

ecture in order to enable fine grained control over mobility manage-

ent and data forwarding focusing on a programmable enterprise

LAN architecture An important part of the SDN architecture is the

etwork controller which provides a centralised view of (parts of)

he SDN enabled network and uses the OpenFlow protocol to install

he forwarding rules on SDN-enabled devices (routers switches ac-

ess points cellular base stations etc)

ODIN [188] is designed to support programmability in enter-

rise WLAN architecture by separating the association state from

he physical AP They implement Light Virtual Access Points (LVAP)

sing a Split-MAC approach where the infrastructure controls the

andover procedure among different WLAN APs By managing

ssociations through SDN controllers ODIN enables proactive mobil-

ty management and load balancing within the SDN enabled WLAN

nterprise network without the need for changes in the client WLAN

tack or IEEE 80211 MAC layer While ODIN requires agents to reside

n the APs that communicate with the Odin Master within the SDN

ontroller CLOUDMAC [189] is based purely on the concept of SDN

nd virtual APs Similar to ODIN CLOUDMAC implements a Split-MAC

pproach but in addition enables the processing of WLAN MAC layer

rames within a co-located Cloud using so-called Virtual APs (VAPs)

he physical APs in CLOUDMAC are lightweight WLAN APs that are

esponsible only for sending out IEEE 80211 based MAC layer ACKs to

tandard WLAN clients and tunnel WLAN MAC layer frames through

n SDN-enabled enterprise WLAN towards the VAPs As association

tates are kept in the VAPs fast mobility is supported using sim-

le OpenFlow forwarding rules Because of the additional processing

f IEEE 80211 WLAN MAC frames in the co-located Cloud CLOUD-

AC has higher latency than a standard WLAN deployment However

LOUDMAC offers a Webservice based API to third party applications

n order to program the enterprise WLAN and enable new services

he architecture of CLOUDMAC is depicted in Fig 11 It has been ex-

ended in [190] to support QoS management and in [191] to support

exible MAC management frame prioritisation based on IEEE 80211

system based on OpenFlow has been recently proposed in [192] to

llow the station to be associated with multiple APs simultaneously

nd to switch between APs with low overhead

An important aspect to consider in order to enable programmabil-

ty is backwards compatibility Several approaches (eg [180]) require

lients to be modified to utilise the features provided This is difficult

o do in practice because it requires changes in the operating system

oftware of all clients If not all clients can utilise those APIs it is ques-

ionable how usable the new architecture will be and how much ben-

fit in terms of aggregate performance such architecture will allow

n contrast the SDN based approaches are interesting in the sense

hat they do not require changes in the client WLAN stack and work

ith the standard IEEE 80211 MAC layer deployed within the clients

owever it remains questionable how scalable such solutions are For

xample an interesting open research topic is to evaluate the scala-

ility of approaches such as CLOUDMAC [189] In addition processing

EEE 80211 MAC frames within co-located private Cloud requires low

atency support from local Cloud solutions (such as OpenStack) in or-

er to reduce the IEEE 80211 MAC processing time which is an area

f active research

2 Prototyping and testing IEEE 80211 enhancements

Most of the new proposals for next-generation WLANs are cur-

ently only evaluated using mathematical analysis and simulation

hile both analysis and simulation are necessary to characterise and

tudy those enhancements in the initial design phase or to consider

arge-scale scenarios it is difficult to consider all practical aspects of

real-world scenario This can sometimes cause significant differ-

nces between what simulations and real experiments show How-

ver real experiments are challenging because of the high complexity

nd costs of building the new hardware and software for each specific

olution to test

To mitigate implementation complexity there are several flexi-

le hardware platforms where low-level MAC mechanisms can be

ompletely implemented in software Examples based on FPGAs are

he USRP (Universal Software Radio Peripheral) [193] and the WARP

Wireless Open-Access Research Platform) [194] As an alternative

ORA (Microsoft Research Software Radio) [195] works on general

urpose computers by taking some advantage from modern multiple

ore systems

A different approach is provided by OpenFWWF [196] Open-

WWF provides an open CSMACA firmware for specific models of

roadcom chipsets so the resulting firmware can be uploaded and

ested in real commercial hardware OpenFWWF implements a sim-

le State Machine (SM) for controlling the hardware in real time

he SM evolution is driven by a main loop that reacts to events by

xecuting specific handlers When a packet originally prepared by

he Linux kernel is ready in the NIC memory handler Packet_Ready

ets up the radio hardware according to the packet metadata sched-

les the transmission and jumps back to the main loop The Trans-

ission Engine (TXE) then takes care of accessing the channel ie

t decrements the back-off counter according to the DCF rules and

ventually starts the actual transmission This triggers the execution

f the TX_frame_now event that prepares the ACK time-out clock

nd finalises the MAC header If the ACK-frame is received or if the

CK time-out expires and the maximum number of attempts for this

acket is reached handler Update_Params resets the contention win-

ow to its minimum value or otherwise doubles it Finally it loads

he back-off counter with a fresh value

In the following we will review a list of selected papers that use

he OpenFWWF firmware and the WARP or USRP platforms to test

ew proposals for WLANs in real scenarios as examples A first im-

lementation of the groupcast mechanisms defined in IEEE 80211aa

s presented in [197] The authors modify the OpenFWWF firmware

o include those functionalities and evaluate them using a 30 STAs

estbed A second example of the use of OpenFWWF is the collision-

ree MAC protocol presented in [198] In [198] the authors design and

mplement in OpenFWWF a MAC protocol that is able to achieve a

20 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 12 LTE to WIFI offloading

a

n

c

t

[

d

a

W

v

t

t

o

b

f

S

g

l

i

I

s

I

b

n

s

m

a

p

d

c

N

o

v

t

t

n

l

i

a

t

a

collision-free operation by waiting a deterministic time after success-

ful transmissions before to start a new transmission This work shows

the experimental performance of a collision-free MAC (CF-MAC) pro-

tocol for WLANs using commercial hardware The results show that

the proposed CF-MAC protocol leads to a better distribution of the

available bandwidth among users a higher throughput and lower

losses than DCF

In [199] the authors use the WARP platform to test a variant of the

DCF ndash called IEEE 80211ec ndash that substitutes control packets such as

the RTS CTS and ACK with short detectable sequences Since those se-

quences are shorter than the control packets and can be detected cor-

rectly even at lower SNRs values a significant gain in performance is

achieved MIDAS (Multiple-Input Distributed Antenna Systems) [200]

implementation using the WARP platform shows the benefits of dis-

tributing the antennas over the area to cover instead of co-locating

them at the AP in terms of capacity when MU-MIMO is employed

The authors also propose a new MAC protocol to benefit from the

spatial reuse that the DAS allows taking as a basis the IEEE 80211ac

Experimental results show that their proposal is able to achieve up

to 200 gains versus the traditional approach where all antennas are

co-located at the AP

The USRP platform has recently been proposed to develop a first

implementation of IEEE 80211ah in order to obtain a preliminary ex-

perimental performance assessment of such a technology because

no commercial off-the-shelf hardware is yet available as the amend-

ment is still in progress [201] The USRP platform is also used in

[202] to evaluate TIMO (Technology Independent Multi-Output) a so-

lution to deal with high-power non-IEEE 80211 interferers in ISM

bands

To conclude this section it is worth mentioning here also to

MAClets and WMP as given such flexibility to implement and dis-

tribute new MAC protocols and other IEEE 80211 functionalities we

believe that if a use-friendly implementation of such a MAC proces-

sors framework is provided it would significantly contribute to the

development and testing of new MAC enhancements by the research

community

63 CellularWLAN interworking

Public hotspots that offer Internet access over a WLAN using IEEE

80211 technology are now nearly ubiquitous It is forecasted that the

cumulative installed base of WiFi hotspots worldwide will amount

to 551 million by 2018 excluding private hotspots (eg WiFi access

points deployed at home) [203] The sharp increase in the availabil-

ity of public WiFi was initially perceived by mobile cellular operators

as a threat due to the additional competition from wireline Internet

service providers or emerging crowdsourced WiFi networks such as

FON5 However as cellular operators are fighting to cope with the ex-

plosion of mobile data traffic created by the rising use of multimedia

content traffic over mobile devices [15] they are also starting to use

WLANs based on the IEEE 80211 technology to offload data from their

core and access networks In general mobile data offloading refers to

the use of complementary network technologies (in licensed or un-

licensed spectrum) for delivering data originally targeted to cellular

networks Intuitively the simplest type of offloading consists of ex-

ploiting connectivity to existing co-located WiFi networks and trans-

ferring data without any delay (Fig 12) Thus this offloading tech-

nique is known as on-the-spot offloading As a consequence of this

new trend the seamless integration of cellular (eg 3GLTE) and WiFi

technologies has attracted significant research interest in recent years

(see [204] for a survey) a few solutions have already been standard-

ised [205206] and roaming between cellular and WiFi is becoming

increasingly transparent to end users CellularWLAN interworking is

5 httpscorpfoncomen

fl

t

f

lso fostered by the support in the evolving 4G standards of heteroge-

eous network deployments (HetNets) in which the existing macro

ells are complemented with a number of small low-power base sta-

ions with the goal of increasing capacity in highly congested areas

207208] It is envisaged that small cells will be based on 4G stan-

ards (eg pico and femto cells) as well as IEEE 80211 technologies

nd multimode base stations that work simultaneously with LTE and

iFi are already entering the market

It is important to point out that LTE standards already support a

ariety of mechanisms that enable data offloading However most of

he existing solutions such as Local IP Access (LIPA) and Selected In-

ernet IP Traffic Offload (SIPTO) focus on data offloading at the core

f the network [205] For instance LIPA allows an IP-enabled mo-

ile device to transfer data to another device in the same pico or

emto cell without passing through the cellular access network while

IPTO enables the routing of selected IP data flows through different

ateways The only LTE offloading mechanism that supports seam-

ess interworking with IEEE 80211-based WLANs is IP Flow Mobil-

ty and Seamless Offload (IFOM) Specifically IFOM relies on Mobile

Pv6 technologies to allow a user terminal to simultaneously route

elected IP flows over different radio access technologies [209210]

n this way a user terminal can offload selected flows to a WLAN

ased on some operator-defined policy while keeping the LTE con-

ection running However to enable such an approach it is neces-

ary to support an entity in the cellular core network that can com-

unicate with the user terminal to exchange information about the

vailability and quality of neighbouring access networks as well as to

rovide the user terminal with predefined rules to manage the han-

over process This entity in current LTE standards is called the Ac-

ess Network Discovery and Selection Function (ANDSF) server [211]

ote that the problem of selecting the best communication technol-

gy in an heterogeneous wireless network has been extensively in-

estigated in the literature both with centralised as well as decen-

ralised approaches (see [212] for a survey) However the use of mul-

iple interfaces in parallel as well as per-flow offloading are relatively

ew concepts Thus the design of scalable and efficient network se-

ection strategies for the ANDSF framework is still an open issue For

nstance three offloading methods suitable for the ANDSF framework

re proposed in [213] which are based on coverage SNR and sys-

em load A reinforcement learning approach is designed in [214] that

llows multimode base stations to autonomously steer their traffic

ows across different access technologies depending on the traffic

ype the usersrsquo QoS requirements the network load and the inter-

erence levels A number of studies have also tried to quantify the

20 B Bellalta et al Computer Communications 75 (2016) 1ndash25

Fig 12 LTE to WIFI offloading

a

n

c

t

[

d

a

W

v

t

t

o

b

f

S

g

l

i

I

s

I

b

n

s

m

a

p

d

c

N

o

v

t

t

n

l

i

a

t

a

collision-free operation by waiting a deterministic time after success-

ful transmissions before to start a new transmission This work shows

the experimental performance of a collision-free MAC (CF-MAC) pro-

tocol for WLANs using commercial hardware The results show that

the proposed CF-MAC protocol leads to a better distribution of the

available bandwidth among users a higher throughput and lower

losses than DCF

In [199] the authors use the WARP platform to test a variant of the

DCF ndash called IEEE 80211ec ndash that substitutes control packets such as

the RTS CTS and ACK with short detectable sequences Since those se-

quences are shorter than the control packets and can be detected cor-

rectly even at lower SNRs values a significant gain in performance is

achieved MIDAS (Multiple-Input Distributed Antenna Systems) [200]

implementation using the WARP platform shows the benefits of dis-

tributing the antennas over the area to cover instead of co-locating

them at the AP in terms of capacity when MU-MIMO is employed

The authors also propose a new MAC protocol to benefit from the

spatial reuse that the DAS allows taking as a basis the IEEE 80211ac

Experimental results show that their proposal is able to achieve up

to 200 gains versus the traditional approach where all antennas are

co-located at the AP

The USRP platform has recently been proposed to develop a first

implementation of IEEE 80211ah in order to obtain a preliminary ex-

perimental performance assessment of such a technology because

no commercial off-the-shelf hardware is yet available as the amend-

ment is still in progress [201] The USRP platform is also used in

[202] to evaluate TIMO (Technology Independent Multi-Output) a so-

lution to deal with high-power non-IEEE 80211 interferers in ISM

bands

To conclude this section it is worth mentioning here also to

MAClets and WMP as given such flexibility to implement and dis-

tribute new MAC protocols and other IEEE 80211 functionalities we

believe that if a use-friendly implementation of such a MAC proces-

sors framework is provided it would significantly contribute to the

development and testing of new MAC enhancements by the research

community

63 CellularWLAN interworking

Public hotspots that offer Internet access over a WLAN using IEEE

80211 technology are now nearly ubiquitous It is forecasted that the

cumulative installed base of WiFi hotspots worldwide will amount

to 551 million by 2018 excluding private hotspots (eg WiFi access

points deployed at home) [203] The sharp increase in the availabil-

ity of public WiFi was initially perceived by mobile cellular operators

as a threat due to the additional competition from wireline Internet

service providers or emerging crowdsourced WiFi networks such as

FON5 However as cellular operators are fighting to cope with the ex-

plosion of mobile data traffic created by the rising use of multimedia

content traffic over mobile devices [15] they are also starting to use

WLANs based on the IEEE 80211 technology to offload data from their

core and access networks In general mobile data offloading refers to

the use of complementary network technologies (in licensed or un-

licensed spectrum) for delivering data originally targeted to cellular

networks Intuitively the simplest type of offloading consists of ex-

ploiting connectivity to existing co-located WiFi networks and trans-

ferring data without any delay (Fig 12) Thus this offloading tech-

nique is known as on-the-spot offloading As a consequence of this

new trend the seamless integration of cellular (eg 3GLTE) and WiFi

technologies has attracted significant research interest in recent years

(see [204] for a survey) a few solutions have already been standard-

ised [205206] and roaming between cellular and WiFi is becoming

increasingly transparent to end users CellularWLAN interworking is

5 httpscorpfoncomen

fl

t

f

lso fostered by the support in the evolving 4G standards of heteroge-

eous network deployments (HetNets) in which the existing macro

ells are complemented with a number of small low-power base sta-

ions with the goal of increasing capacity in highly congested areas

207208] It is envisaged that small cells will be based on 4G stan-

ards (eg pico and femto cells) as well as IEEE 80211 technologies

nd multimode base stations that work simultaneously with LTE and

iFi are already entering the market

It is important to point out that LTE standards already support a

ariety of mechanisms that enable data offloading However most of

he existing solutions such as Local IP Access (LIPA) and Selected In-

ernet IP Traffic Offload (SIPTO) focus on data offloading at the core

f the network [205] For instance LIPA allows an IP-enabled mo-

ile device to transfer data to another device in the same pico or

emto cell without passing through the cellular access network while

IPTO enables the routing of selected IP data flows through different

ateways The only LTE offloading mechanism that supports seam-

ess interworking with IEEE 80211-based WLANs is IP Flow Mobil-

ty and Seamless Offload (IFOM) Specifically IFOM relies on Mobile

Pv6 technologies to allow a user terminal to simultaneously route

elected IP flows over different radio access technologies [209210]

n this way a user terminal can offload selected flows to a WLAN

ased on some operator-defined policy while keeping the LTE con-

ection running However to enable such an approach it is neces-

ary to support an entity in the cellular core network that can com-

unicate with the user terminal to exchange information about the

vailability and quality of neighbouring access networks as well as to

rovide the user terminal with predefined rules to manage the han-

over process This entity in current LTE standards is called the Ac-

ess Network Discovery and Selection Function (ANDSF) server [211]

ote that the problem of selecting the best communication technol-

gy in an heterogeneous wireless network has been extensively in-

estigated in the literature both with centralised as well as decen-

ralised approaches (see [212] for a survey) However the use of mul-

iple interfaces in parallel as well as per-flow offloading are relatively

ew concepts Thus the design of scalable and efficient network se-

ection strategies for the ANDSF framework is still an open issue For

nstance three offloading methods suitable for the ANDSF framework

re proposed in [213] which are based on coverage SNR and sys-

em load A reinforcement learning approach is designed in [214] that

llows multimode base stations to autonomously steer their traffic

ows across different access technologies depending on the traffic

ype the usersrsquo QoS requirements the network load and the inter-

erence levels A number of studies have also tried to quantify the

B Bellalta et al Computer Communications 75 (2016) 1ndash25 21

p

m

c

t

i

a

n

t

o

W

s

p

w

t

r

p

t

b

e

a

t

t

F

a

q

i

i

t

c

b

b

s

t

e

l

a

w

t

T

s

i

d

B

(

t

a

i

c

n

i

t

t

c

e

a

T

b

B

a

a

b

t

p

w

o

t

h

m

fl

[

t

a

[

o

o

t

o

r

i

a

m

o

p

a

d

r

b

s

n

t

n

d

s

t

t

o

i

c

t

[

l

s

u

p

d

o

s

s

f

p

c

n

I

s

l

u

a

d

d

s

i

t

fl

s

7

t

t

otential capacity gain of WiFi offloading in real-world WiFi deploy-

ents For instance the authors in [215] evaluate offloading effi-

iency using trace-based urban mobility patterns and WiFi connec-

ivity distributions The authors in [216] instead analyse the offload-

ng performance in an indoor scenario in which femto cells and WiFi

ccess points coexist On the other hand mathematical models are

eeded to derive performance bounds and guide the design of op-

imal offloading strategies A simple queuing model for the analysis

f the offloading efficiency is developed in [217] by assuming that the

iFi network availability is exponentially distributed A more general

cenario is considered in [218] by assuming multiple classes of access

oints that differ in transmit power deployment density and band-

idth The optimal association strategy is then derived to maximise

he fraction of time that a typical user in the network is served with a

ate greater than its minimum rate requirement A somehow related

roblem consists of deciding how to optimally deploy WiFi hotspots

o maximise offloading efficiency Traditionally this problem has

een considered from the point of view of coverage maximisation

g to ensure continuous WiFi connectivity by taking into consider-

tion user mobility characteristics [219] On the contrary in the con-

ext of mobile data offloading AP deployment is tackled to maximise

he throughput performance in an heterogeneous wireless network

or instance a heuristic algorithm is proposed in [220] that selects

s AP locations the cells with the higher frequency of download re-

uests A graph-theoretical solution for AP deployment is developed

n [221] by consorting a time-dependent graph that describes the

nterdependencies between usersrsquo mobility trajectories points of in-

erest and traffic demands An open issue in this field of research

oncerns the design of more adaptive traffic steering mechanisms

etween cellular and WiFi Furthermore the increase in the num-

er of wireless infrastructure nodes with the dense deployment of

mall cells will make the future network deployments quasi stochas-

ic Thus new methodologies such as stochastic geometry have to be

xplored to model the performance bounds of heterogeneous wire-

ess networks that allow the inter-working between cellular systems

nd WLANs [222] Finally the use of historical data to predict net-

ork conditions and user locations may also become infeasible due

o the scale of the network in terms of infrastructure nodes and users

o deal with this issue limited measurements could be coupled with

tatistical inference methods

On-the-spot offloading is the dominant but not the only form of

nterworking between cellular networks and WLANs More recently

elayed offloading has also been proposed for delay-tolerant traffic

asically if a user is willing to accept a delayed content reception

eg the download of a YouTube video) the cellular operator may in-

entionally postpone the content transfer in order to wait for WLAN

vailability or better transmission conditions The cellular network

s then used to complete the data transfer only if the content re-

eption cannot be guaranteed within a user-specified deadline A

umber of studies have explored the feasibility of delayed offload-

ng for different delay deadlines using trace-based WLAN usage pat-

erns In particular results in [215] confirm that increasing the delay-

olerance of content significantly improves the fraction of traffic that

an be offloaded The offloading performance clearly depends on sev-

ral factors including the location of IEEE 80211 hotspots and the

bility to accurately predict the future availability of WLAN coverage

hus several studies have addressed the problem of forecasting mo-

ile connectivity For instance the solution proposed in [223] called

readCrumbs tracks the movements of the mobile devicersquos owner

nd maintains a history of observed networking conditions to train

forecast model of near-term connectivity More recently a time-

ased prediction model of visited locations derived from movement

races of mobile users is given in [224] Then the authors in [225]

ropose a system called Wiffler that allows mobile users to decide

hether or not to wait for a future WiFi offloaded opportunity based

n the predicted WLAN capacity and the total data that needs to be

ransferred However the design of location prediction models that

ave low computation complexity and are suitable for short-term

obility is an open issue Also related to the problem of delayed of-

oading feasibility is the optimal placement of APs The authors of

226] develop the HotZones algorithm which selects the cells with

he highest number of daily visits as the location of additional WiFi

ccess points A similar solution called Drop Zones is proposed in

227] In the context of vehicular networks the optimal deployment

f RoadSide Units (RSU) over a given road layout to maximise the

verall system throughput is analysed in [228] In [229] it is analysed

he data offloading gain in a vehicular sensor network as a function

f the percentage of equipped vehicles of the number of deployed

oad side units and of the adopted routing protocol Although exist-

ng work established the potential of delayed offloading there is still

need for considerable research to design incentive mechanisms for

otivating users to leverage their delay tolerance for cellular traffic

ffloading For instance an auction-based pricing framework is pro-

osed in [230] to give priority to users with high delay tolerance and

large offloading potential

We conclude this section by discussing a third type of mobile

ata offloading known as opportunistic offloading which does not

ely on a WLAN infrastructure but exploits direct communications

etween mobile devices eg through the emerging WiFi Direct

tandard [231] Opportunistic offloading schemes allow saving sig-

ificant cellular bandwidth because the content spreads through

he opportunistic network formed by the users while the cellular

etwork is mainly used for signalling and triggering the content

issemination Clearly the performance of opportunistic offloading

olutions depends on several factors including the user mobility pat-

erns the user density the delay tolerance for content reception and

he popularity of the content that needs to be transferred Note that

pportunistic offloading requires a less controlled type of interwork-

ng between LTE and WLANs because user devices can setup direct

onnections and start ad hoc communication autonomously with lit-

le or no intervention from the operator For instance the authors of

226] propose a simple algorithm called MixZones to allow the cel-

ular operator to decide when the mobile users should be notified to

witch their wireless interface for data transfer with potential other

sers that they are predicted to encounter In addition most of the pa-

ers consider that content must be delivered to users within a given

eadline Most of the research in this context addresses the problem

f selecting the best (eg the smallest) set of users called seeds that

hould receive the content from the cellular network and help to dis-

eminate it over the opportunistic network Three simple algorithms

or initial seed selections are proposed in [232] The authors of [233]

ropose using social network properties eg betweenness or degree

entrality to select the most useful seeds for offloading the cellular

etwork A similar social-aware approach is also used in [234235]

t is assumed in [236] that a centralised entity keeps tracks of the

peed of the content dissemination process to decide when the cel-

ular network should directly transmit the content to the interested

sers to guarantee that the delivery deadline is met with high prob-

bility An actor-critic learning framework is designed in [237] to un-

erstand when and to how many users the cellular network should

irectly transmit the content In this context an important area of re-

earch is the design of scalable and efficient network-aided offload-

ng schemes where the cellular network guides the mobile users in

he connectivity management and dissemination phase Finally of-

oading data traffic when the requests for popular content are not

ynchronised is still an open issue [238]

Summary

In this paper we have described the main scenarios novel func-

ionalities and mechanisms that will characterise the use opera-

ion and performance of next-generation WLANs and provided an

22 B Bellalta et al Computer Communications 75 (2016) 1ndash25

extensive and thorough review of the IEEE 80211ac IEEE 80211aa

IEEE 80211ah and IEEE 80211af amendments The paper also pro-

vides an up-to-date survey of the most representative work in this re-

search area summarising the key contributions to the current status

and future evolution of WLANs Differently from other 80211-related

surveys this overview is structured with regards to the emerging

WLAN application scenarios such as M2M cognitive radios and high-

definition multimedia delivery We also describe some open chal-

lenges that require further research in coming years [239] with spe-

cial focus on software-defined MACs and the internet-working with

cellular systems

Acknowledgements

We would like to thank the anonymous reviewers for their in-

sightful comments on the paper as these comments led us to an

improvement of the work A special thanks also to Toke Hoslashiland-

Joslashrgensen and Ognjen Dobrijevic for the time they spent reviewing

the manuscript The research of Boris Bellalta was partially supported

by the Spanish Government (Project TEC2012-32354) and by the

Catalan Government (SGR-2014-1173)

References

[1] G Hiertz D Denteneer L Stibor Y Zang XP Costa B Walke The IEEE 80211Universe IEEE Commun Mag 48 (1) (2010) 62ndash70

[2] N Bhushan J Li D Malladi R Gilmore D Brenner A DamnjanovicR Sukhavasi C Patel S Geirhofer Network densification the dominant theme

for wireless evolution into 5G IEEE Commun Mag 52 (2) (2014) 82ndash89[3] M Conti S Giordano Mobile ad hoc networking milestones challenges and

new research directions IEEE Commun Mag 52 (1) (2014) 85ndash96

[4] M Conti C Boldrini S Kanhere E Mingozzi E Pagani PM Ruiz M YounisFrom MANET to people-centric networking milestones and open research chal-

lenges Comput Commun (2015) doi101016jcomcom201509007[5] H Zhu M Li I Chlamtac B Prabhakaran A survey of quality of service in IEEE

80211 networks IEEE Wirel Commun 11 (4) (2004) 6ndash14[6] B Bellalta A Vinel P Chatzimisios R Bruno C Wang Research advances and

standardization activities in WLANs Comput Commun 39 (2014) 1ndash2

[7] IEEE IEEE Std 80211n-2009 Part 11 Wireless LAN Medium Access Control(MAC) and Physical Layer (PHY) Specifications ndash Amendement 5 Enhancements

for Higher Throughput October 2009[8] IEEE IEEE Std 80211p-2010 Part 11 Wireless LAN Medium Access Control

(MAC) and Physical Layer (PHY) Specifications ndash Amendment 6 Wireless Accessin Vehicular Environments July 2010

[9] IEEE IEEE Std 80211s-2011 Part 11 Wireless LAN Medium Access Control

(MAC) and Physical Layer (PHY) Specifications September 2011[10] IEEE IEEE 80211-2012 Part 11 Wireless LAN Medium Access Control (MAC)

and Physical Layer (PHY) Specifications March 2012[11] W Sun O Lee Y Shin S Kim G Yang H Kim S Choi Wi-Fi could be much

more IEEE Commun Mag 52 (11) (2014) 22ndash29[12] E Borgia The internet of things vision key features applications and open is-

sues Comput Commun 54 (2014) 1ndash31

[13] S Tozlu M Senel W Mao A Keshavarzian Wi-Fi enabled sensors for internetof things A practical approach IEEE Commun Mag 50 (6) (2012) 134ndash143

[14] CW Park D Hwang T-J Lee Enhancement of IEEE 80211ah MAC for M2M com-munications IEEE Commun Lett 18 (7) (2014) 1151ndash1154

[15] Cisco Cisco Visual Networking Index Global Mobile Data Traffic Forecast Up-date 2013-2018 Technical report Cisco February 2014

[16] K Kosek-Szott M Natkaniec S Szott A Krasilov A Lyakhov A Safonov I Tin-

nirello Whatrsquos new for QoS in IEEE 80211 IEEE Netw 27 (6) (2013) 95ndash104[17] C-S Sum GP Villardi MA Rahman T Baykas HN Tran Z Lan C Sun Y Alem-

seged J Wang C Song C-W Pyo S Filin H Harada Cognitive communicationin TV white spaces an overview of regulations standards and technology IEEE

Commun Mag 51 (7) (2013) 138ndash145[18] ACV Gummalla JO Limb Wireless medium access control protocols IEEE

Commun Surv Tutor 3 (2) (2000) 2ndash15Second Quarter

[19] RC Carrano LCS Magalhatildees DCM Saade CVN Albuquerque IEEE 80211smultihop MAC a tutorial IEEE Commun Surv Tutor 13 (1) (First 2011) 52ndash67

[20] E Charfi L Chaari L Kamoun PHYMAC enhancements and QoS mechanismsfor very high throughput WLANs a survey IEEE Commun Surv Tutor 15 (4)

(2013) 1714ndash1735[21] ITU-T International Telecommunication Union Recommendation G114 One-

way Transmission Time ITU-T Study Group 12 International Telecommunica-tion Union May 2003

[22] H Schwarz D Marpe T Wiegand Overview of the scalable video coding exten-

sion of the H264AVC standard IEEE Trans Circuits Syst Video Technol 17 (9)(2007) 1103ndash1120

[23] E Ancillotti R Bruno M Conti The role of communication systems in smartgrids architectures technical solutions and research challenges Comput Com-

mun 36 (17ndash18) (2013) 1665ndash1697

[24] ETSI Applicability of M2M architecture to smart grid network Technical Report102 935 V211 ETSI September 2009

[25] IF Akyildiz W Su Y Sankarasubramaniam E Cayirci Wireless sensor net-works a survey Comput Netw 38 (4) (2002) 393ndash422

[26] S-Y Lien K-C Chen Y Lin Toward ubiquitous massive accesses in 3GPPmachine-to-machine communications IEEE Commun Mag 49 (4) (2011) 66ndash

74[27] B Bellalta A Checco A Zocca J Barcelo On the interactions between multiple

overlapping WLANs using channel bonding IEEE Trans Veh Technol (2015)

[28] M Nekovee A survey of cognitive radio access to TV white spaces Ultra ModernTelecommunications amp Workshops 2009 ICUMTrsquo09 International Conference

on IEEE 2009[29] IEEE IEEE Std P80211ac Part 11 Wireless LAN Medium Access Control (MAC)

and Physical Layer (PHY) specifications Enhancements for Very High Through-put for Operation in Bands below 6 GHz 2013

[30] M Gong B Hart L Xia R Want Channel bounding and MAC protection mecha-

nisms for 80211ac in Proceedings of IEEE GLOBECOMrsquo11 2011[31] M Park IEEE 80211ac dynamic bandwidth channel access in Proceedings of

IEEE ICCrsquo11 IEEE 2011[32] A Faridi B Bellalta A Checco Analysis of dynamic channel bonding in dense

networks of WLANs 2015 arXiv150900290[33] J Ong EH Kneckt O Alanen Z Chang T Huovinen T Nihtila IEEE 80211ac En-

hancements for very high throughput WLANs in Proceedings of IEEE PIMRCrsquo11

2011 pp 849ndash853[34] Y Zeng PH Pathak P Mohapatra A first look at 80211ac in action energy effi-

ciency and interference characterization in Proceedings of IFIP Networkingrsquo142014

[35] E Aryafar N Anand T Salonidis EW Knightly Design and experimental evalu-ation of multi-user beamforming in wireless LANs in Proceedings of ACM MO-

BICOMrsquo10 2010 pp 197ndash208

[36] B Bellalta J Barcelo D Staehle A Vinel M Oliver On the performance of packetaggregation in IEEE 80211 ac MU-MIMO WLANs IEEE Commun Lett 16 (10)

(2012) 1588ndash1591[37] O Sharon Y Alpert MAC level Throughput comparison 80211ac vs 80211n

Phys Commun 12 (2014) 33ndash49[38] M Yazid A Ksentini L Bouallouche-Medjkoune D Aissani Performance Analy-

sis of the TXOP Sharing Mechanism in the VHT IEEE 80211ac WLANs IEEE Com-

mun Lett 18 (9) (2014) 1599ndash1602[39] G Redieteab L Cariou P Christin J-F Helard PHY+MAC channel sounding in-

terval analysis for IEEE 80211ac MU-MIMO in Proceedings of IEEE ISWCSrsquo122012 pp 1054ndash1058

[40] O Bejarano E Magistretti O Gurewitz E Knightly MUTE sounding inhibitionfor MU-MIMO WLANs in Proceedings of IEEE SECONrsquo14 2014

[41] Q Wang L Greenstein L Cimini D Chan A Hedayat Multi-user and single-user

throughputs for downlink MIMO channels with outdated channel state informa-tion IEEE Wirel Commun Lett 3 (3) (2014) 321ndash324

[42] T Hiraguri K Nishimori Survey of transmission methods and efficiency us-ing MIMO technologies for wireless LAN systems IEICE Trans Commun 98 (7)

(2015) 1250ndash1267[43] J Cha H Jin BC Jung DK Sung Performance comparison of downlink user

multiplexing schemes in IEEE 80211ac multi-user MIMO vs frame aggregationin Proceedings of IEEE WCNCrsquo12 2012 pp 1514ndash1519

[44] R Liao B Bellalta J Barcelo V Valls M Oliver Performance analysis of IEEE

80211ac wireless backhaul networks in saturated conditions EURASIP J WirelCommun Netw 2013 (1) (2013) 1ndash14

[45] O Aboul-Magd U Kwon Y Kim C Zhu Managing downlink multi-user MIMOtransmission using group membership in Proceedings of IEEE CCNCrsquo13 IEEE

2013 pp 370ndash375[46] C Zhu C Ngo A Bhatt Y Kim Enhancing WLAN backoff procedures for down-

link MU-MIMO support in Proceedings of IEEE WCNCrsquo13 IEEE 2013 pp 368ndash

373[47] K Hanada K Yamamoto M Morikura K Ishihara K Riichi Game-theoretic

analysis of multibandwidth channel selection by coordinated APs in WLANs IE-ICE Trans Commun 96 (6) (2013) 1277ndash1287

[48] W-S Jung K-W Lim Y-B Ko Utilising partially overlapped channels for OFDM-based 80211 WLANs Comput Commun 63 (2015) 77ndash86

[49] IEEE IEEE 80211 TGax Status of IEEE 80211 HEW Task Group httpwww

ieee802org11Reportstgax_updatehtm 2014[50] MX Gong B Hart S Mao Advanced wireless LAN technologies IEEE

80211ac and beyond ACM SIGMOBILE Mob Comput Commun Rev 18 (4)(2015) 48ndash52

[51] B Bellalta IEEE 82011ax high-efficiency WLANs IEEE Wirel Commun (2015)arXiv 150101496(in press)

[52] EH Ong Performance analysis of fast initial link setup for IEEE 80211ai WLANs

in Proceedings of IEEE PIMRCrsquo12 IEEE 2012 pp 1279ndash1284[53] D Camps-Mur A Garcia-Saavedra P Serrano Device-to-device communica-

tions with Wi-Fi Direct overview and experimentation IEEE Wirel Commun20 (3) (2013)

[54] L Suresh J Schulz-Zander R Merz A Feldmann T Vazao Towards pro-grammable enterprise WLANS with Odin in Proceedings of ACM HotSDNrsquo12

ACM 2012 pp 115ndash120

[55] M Fang D Malone KR Duffy DJ Leith Decentralised learning MACs forcollision-free access in WLANs Wirel Netw 19 (1) (2013) 83ndash98

[56] L Sanabria-Russo A Faridi B Bellalta J Barcelo M Oliver Future evolutionof CSMA protocols for the IEEE 80211 standard in Proceedings of IEEE ICCrsquo13

2013 pp 1274ndash1279

22 B Bellalta et al Computer Communications 75 (2016) 1ndash25

extensive and thorough review of the IEEE 80211ac IEEE 80211aa

IEEE 80211ah and IEEE 80211af amendments The paper also pro-

vides an up-to-date survey of the most representative work in this re-

search area summarising the key contributions to the current status

and future evolution of WLANs Differently from other 80211-related

surveys this overview is structured with regards to the emerging

WLAN application scenarios such as M2M cognitive radios and high-

definition multimedia delivery We also describe some open chal-

lenges that require further research in coming years [239] with spe-

cial focus on software-defined MACs and the internet-working with

cellular systems

Acknowledgements

We would like to thank the anonymous reviewers for their in-

sightful comments on the paper as these comments led us to an

improvement of the work A special thanks also to Toke Hoslashiland-

Joslashrgensen and Ognjen Dobrijevic for the time they spent reviewing

the manuscript The research of Boris Bellalta was partially supported

by the Spanish Government (Project TEC2012-32354) and by the

Catalan Government (SGR-2014-1173)

References

[1] G Hiertz D Denteneer L Stibor Y Zang XP Costa B Walke The IEEE 80211Universe IEEE Commun Mag 48 (1) (2010) 62ndash70

[2] N Bhushan J Li D Malladi R Gilmore D Brenner A DamnjanovicR Sukhavasi C Patel S Geirhofer Network densification the dominant theme

for wireless evolution into 5G IEEE Commun Mag 52 (2) (2014) 82ndash89[3] M Conti S Giordano Mobile ad hoc networking milestones challenges and

new research directions IEEE Commun Mag 52 (1) (2014) 85ndash96

[4] M Conti C Boldrini S Kanhere E Mingozzi E Pagani PM Ruiz M YounisFrom MANET to people-centric networking milestones and open research chal-

lenges Comput Commun (2015) doi101016jcomcom201509007[5] H Zhu M Li I Chlamtac B Prabhakaran A survey of quality of service in IEEE

80211 networks IEEE Wirel Commun 11 (4) (2004) 6ndash14[6] B Bellalta A Vinel P Chatzimisios R Bruno C Wang Research advances and

standardization activities in WLANs Comput Commun 39 (2014) 1ndash2

[7] IEEE IEEE Std 80211n-2009 Part 11 Wireless LAN Medium Access Control(MAC) and Physical Layer (PHY) Specifications ndash Amendement 5 Enhancements

for Higher Throughput October 2009[8] IEEE IEEE Std 80211p-2010 Part 11 Wireless LAN Medium Access Control

(MAC) and Physical Layer (PHY) Specifications ndash Amendment 6 Wireless Accessin Vehicular Environments July 2010

[9] IEEE IEEE Std 80211s-2011 Part 11 Wireless LAN Medium Access Control

(MAC) and Physical Layer (PHY) Specifications September 2011[10] IEEE IEEE 80211-2012 Part 11 Wireless LAN Medium Access Control (MAC)

and Physical Layer (PHY) Specifications March 2012[11] W Sun O Lee Y Shin S Kim G Yang H Kim S Choi Wi-Fi could be much

more IEEE Commun Mag 52 (11) (2014) 22ndash29[12] E Borgia The internet of things vision key features applications and open is-

sues Comput Commun 54 (2014) 1ndash31

[13] S Tozlu M Senel W Mao A Keshavarzian Wi-Fi enabled sensors for internetof things A practical approach IEEE Commun Mag 50 (6) (2012) 134ndash143

[14] CW Park D Hwang T-J Lee Enhancement of IEEE 80211ah MAC for M2M com-munications IEEE Commun Lett 18 (7) (2014) 1151ndash1154

[15] Cisco Cisco Visual Networking Index Global Mobile Data Traffic Forecast Up-date 2013-2018 Technical report Cisco February 2014

[16] K Kosek-Szott M Natkaniec S Szott A Krasilov A Lyakhov A Safonov I Tin-

nirello Whatrsquos new for QoS in IEEE 80211 IEEE Netw 27 (6) (2013) 95ndash104[17] C-S Sum GP Villardi MA Rahman T Baykas HN Tran Z Lan C Sun Y Alem-

seged J Wang C Song C-W Pyo S Filin H Harada Cognitive communicationin TV white spaces an overview of regulations standards and technology IEEE

Commun Mag 51 (7) (2013) 138ndash145[18] ACV Gummalla JO Limb Wireless medium access control protocols IEEE

Commun Surv Tutor 3 (2) (2000) 2ndash15Second Quarter

[19] RC Carrano LCS Magalhatildees DCM Saade CVN Albuquerque IEEE 80211smultihop MAC a tutorial IEEE Commun Surv Tutor 13 (1) (First 2011) 52ndash67

[20] E Charfi L Chaari L Kamoun PHYMAC enhancements and QoS mechanismsfor very high throughput WLANs a survey IEEE Commun Surv Tutor 15 (4)

(2013) 1714ndash1735[21] ITU-T International Telecommunication Union Recommendation G114 One-

way Transmission Time ITU-T Study Group 12 International Telecommunica-tion Union May 2003

[22] H Schwarz D Marpe T Wiegand Overview of the scalable video coding exten-

sion of the H264AVC standard IEEE Trans Circuits Syst Video Technol 17 (9)(2007) 1103ndash1120

[23] E Ancillotti R Bruno M Conti The role of communication systems in smartgrids architectures technical solutions and research challenges Comput Com-

mun 36 (17ndash18) (2013) 1665ndash1697

[24] ETSI Applicability of M2M architecture to smart grid network Technical Report102 935 V211 ETSI September 2009

[25] IF Akyildiz W Su Y Sankarasubramaniam E Cayirci Wireless sensor net-works a survey Comput Netw 38 (4) (2002) 393ndash422

[26] S-Y Lien K-C Chen Y Lin Toward ubiquitous massive accesses in 3GPPmachine-to-machine communications IEEE Commun Mag 49 (4) (2011) 66ndash

74[27] B Bellalta A Checco A Zocca J Barcelo On the interactions between multiple

overlapping WLANs using channel bonding IEEE Trans Veh Technol (2015)

[28] M Nekovee A survey of cognitive radio access to TV white spaces Ultra ModernTelecommunications amp Workshops 2009 ICUMTrsquo09 International Conference

on IEEE 2009[29] IEEE IEEE Std P80211ac Part 11 Wireless LAN Medium Access Control (MAC)

and Physical Layer (PHY) specifications Enhancements for Very High Through-put for Operation in Bands below 6 GHz 2013

[30] M Gong B Hart L Xia R Want Channel bounding and MAC protection mecha-

nisms for 80211ac in Proceedings of IEEE GLOBECOMrsquo11 2011[31] M Park IEEE 80211ac dynamic bandwidth channel access in Proceedings of

IEEE ICCrsquo11 IEEE 2011[32] A Faridi B Bellalta A Checco Analysis of dynamic channel bonding in dense

networks of WLANs 2015 arXiv150900290[33] J Ong EH Kneckt O Alanen Z Chang T Huovinen T Nihtila IEEE 80211ac En-

hancements for very high throughput WLANs in Proceedings of IEEE PIMRCrsquo11

2011 pp 849ndash853[34] Y Zeng PH Pathak P Mohapatra A first look at 80211ac in action energy effi-

ciency and interference characterization in Proceedings of IFIP Networkingrsquo142014

[35] E Aryafar N Anand T Salonidis EW Knightly Design and experimental evalu-ation of multi-user beamforming in wireless LANs in Proceedings of ACM MO-

BICOMrsquo10 2010 pp 197ndash208

[36] B Bellalta J Barcelo D Staehle A Vinel M Oliver On the performance of packetaggregation in IEEE 80211 ac MU-MIMO WLANs IEEE Commun Lett 16 (10)

(2012) 1588ndash1591[37] O Sharon Y Alpert MAC level Throughput comparison 80211ac vs 80211n

Phys Commun 12 (2014) 33ndash49[38] M Yazid A Ksentini L Bouallouche-Medjkoune D Aissani Performance Analy-

sis of the TXOP Sharing Mechanism in the VHT IEEE 80211ac WLANs IEEE Com-

mun Lett 18 (9) (2014) 1599ndash1602[39] G Redieteab L Cariou P Christin J-F Helard PHY+MAC channel sounding in-

terval analysis for IEEE 80211ac MU-MIMO in Proceedings of IEEE ISWCSrsquo122012 pp 1054ndash1058

[40] O Bejarano E Magistretti O Gurewitz E Knightly MUTE sounding inhibitionfor MU-MIMO WLANs in Proceedings of IEEE SECONrsquo14 2014

[41] Q Wang L Greenstein L Cimini D Chan A Hedayat Multi-user and single-user

throughputs for downlink MIMO channels with outdated channel state informa-tion IEEE Wirel Commun Lett 3 (3) (2014) 321ndash324

[42] T Hiraguri K Nishimori Survey of transmission methods and efficiency us-ing MIMO technologies for wireless LAN systems IEICE Trans Commun 98 (7)

(2015) 1250ndash1267[43] J Cha H Jin BC Jung DK Sung Performance comparison of downlink user

multiplexing schemes in IEEE 80211ac multi-user MIMO vs frame aggregationin Proceedings of IEEE WCNCrsquo12 2012 pp 1514ndash1519

[44] R Liao B Bellalta J Barcelo V Valls M Oliver Performance analysis of IEEE

80211ac wireless backhaul networks in saturated conditions EURASIP J WirelCommun Netw 2013 (1) (2013) 1ndash14

[45] O Aboul-Magd U Kwon Y Kim C Zhu Managing downlink multi-user MIMOtransmission using group membership in Proceedings of IEEE CCNCrsquo13 IEEE

2013 pp 370ndash375[46] C Zhu C Ngo A Bhatt Y Kim Enhancing WLAN backoff procedures for down-

link MU-MIMO support in Proceedings of IEEE WCNCrsquo13 IEEE 2013 pp 368ndash

373[47] K Hanada K Yamamoto M Morikura K Ishihara K Riichi Game-theoretic

analysis of multibandwidth channel selection by coordinated APs in WLANs IE-ICE Trans Commun 96 (6) (2013) 1277ndash1287

[48] W-S Jung K-W Lim Y-B Ko Utilising partially overlapped channels for OFDM-based 80211 WLANs Comput Commun 63 (2015) 77ndash86

[49] IEEE IEEE 80211 TGax Status of IEEE 80211 HEW Task Group httpwww

ieee802org11Reportstgax_updatehtm 2014[50] MX Gong B Hart S Mao Advanced wireless LAN technologies IEEE

80211ac and beyond ACM SIGMOBILE Mob Comput Commun Rev 18 (4)(2015) 48ndash52

[51] B Bellalta IEEE 82011ax high-efficiency WLANs IEEE Wirel Commun (2015)arXiv 150101496(in press)

[52] EH Ong Performance analysis of fast initial link setup for IEEE 80211ai WLANs

in Proceedings of IEEE PIMRCrsquo12 IEEE 2012 pp 1279ndash1284[53] D Camps-Mur A Garcia-Saavedra P Serrano Device-to-device communica-

tions with Wi-Fi Direct overview and experimentation IEEE Wirel Commun20 (3) (2013)

[54] L Suresh J Schulz-Zander R Merz A Feldmann T Vazao Towards pro-grammable enterprise WLANS with Odin in Proceedings of ACM HotSDNrsquo12

ACM 2012 pp 115ndash120

[55] M Fang D Malone KR Duffy DJ Leith Decentralised learning MACs forcollision-free access in WLANs Wirel Netw 19 (1) (2013) 83ndash98

[56] L Sanabria-Russo A Faridi B Bellalta J Barcelo M Oliver Future evolutionof CSMA protocols for the IEEE 80211 standard in Proceedings of IEEE ICCrsquo13

2013 pp 1274ndash1279

B Bellalta et al Computer Communications 75 (2016) 1ndash25 23

[57] Y Xiao IEEE 80211 performance enhancement via concatenation and piggybackmechanisms IEEE Trans Wirel Commun 4 (5) (2005) 2182ndash2192

[58] R Liao B Bellalta M Oliver Z Niu MU-MIMO MAC protocols for wireless localarea networks a survey IEEE Commun Surv Tutor (2015)

[59] B Li Q Qu Z Yan M Yang Survey on OFDMA based MAC protocols for the nextgeneration WLAN in Proceedings of IEEE WCNCWrsquo15 2015 pp 131ndash135

[60] JI Choi M Jain K Srinivasan P Levis S Katti Achieving single channel fullduplex wireless communication in Proceedings of ACM MOBICOMrsquo10 2010

pp 1ndash12

[61] MS Afaqui E Garcia-Villegas E Lopez-Aguilera G Smith D Camps Evaluationof dynamic sensitivity control algorithm for IEEE 80211ax in Proceedings of

IEEE WCNCrsquo15 2015 pp 1060ndash1065[62] I Jamil L Cariou J-F Helard Improving the capacity of future IEEE 80211 high

efficiency WLANs in Proc IEEE ICTrsquo14 2014 pp 303ndash307[63] VP Mhatre K Papagiannaki F Baccelli Interference mitigation through power

control in high density 80211 WLANs in Proceedings of IEEE INFOCOMrsquo07

2007 pp 535ndash543[64] X Liu A Sheth M Kaminsky K Papagiannaki S Seshan P Steenkiste Pushing

the envelope of indoor wireless spatial reuse using directional access points andclients in Proceedings of ACM MOBICOMrsquo10 2010 pp 209ndash220

[65] IEEE IEEE Std 80211aa-2012 Specific requirements Part11 Wireless LANMedium Access Control (MAC) and Physical Layer (PHY) Specifications Amend-

ment 2 MAC Enhancements for Robust Audio Video Streaming May 2012

[66] K Maraslis P Chatzimisios AC Boucouvalas IEEE 80211aa improvements onvideo transmission over wireless LANs in Proceedings of IEEE ICCrsquo12 2012

pp 115ndash119[67] IEEE IEEE Std 80211v-2011 Part 11 Wireless LAN Medium Access Control

(MAC) and Physical Layer (PHY) specifications Amendment 8 IEEE 80211 Wire-less Network Management February 2011

[68] A Banchs A De La Oliva L Eznarriaga DR Kowalski P Serrano Performance

analysis and algorithm selection for reliable multicast in IEEE 80211aa wirelessLAN IEEE Trans Veh Technol 63 (8) (2014) 3875ndash3891

[69] IEEE IEEE Std 8021D-2004 IEEE Standard for Local and metropolitan area net-works Media Access Control (MAC) Bridges June 2004

[70] P Pancha ME Zarki MPEG coding for variable bit rate video transmission IEEECommun Mag 32 (5) (1994) 54ndash66

[71] MA Santos J Villalon L Orozco-Barbosa A novel QoE-aware multicast mecha-

nism for video communications over IEEE 80211 WLANs IEEE J Sel Areas Com-mun 30 (7) (2012) 1205ndash1214

[72] N Bhushan J Li D Malladi R Gilmore D Brenner A DamnjanovicRT Sukhavasi C Patel S Geirhofer Network densification the dominant theme

for wireless evolution into 5G IEEE Commun Mag 52 (2) (2014) 82ndash89[73] M Benveniste Wireless LANs and lsquoneighborhood capturersquo in Proceedings of

IEEE PIMRCrsquo02 vol 5 2002 pp 2148ndash2154

[74] IEEE IEEE Std 8021BA-2011 IEEE Standard for Local and metropolitan areanetworksndashAudio Video Bridging (AVB) Systems September 2011

[75] GM Garner H Ryu Synchronization of audiovideo bridging networks usingIEEE 8021AS IEEE Commun Mag 49 (2) (2011) 140ndash147

[76] IEEE IEEE Std 8021Qav-2009 IEEE Standard for Local and Metropolitan AreaNetworks - Virtual Bridged Local Area Networks Amendment 12 Forwarding

and Queuing Enhancements for Time-Sensitive Streams January 2009[77] IEEE IEEE Std 8021Qat-2010 IEEE Standard for Local and Metropolitan Area

NetworksmdashVirtual Bridged Local Area Networks Amendment 14 Stream Reser-

vation Protocol (SRP) September 2010[78] J Kuri SK Kasera Reliable multicast in multi-access wireless LANs Wirel Netw

7 (4) (2001) 359ndash369[79] D Dujovne T Turletti Multicast in 80211 WLANs an experimental study in

Proceedings of ACM MSWiMrsquo06 2006 pp 130ndash138[80] M-T Sun L Huang A Arora T-H Lai Reliable MAC layer multicast in IEEE

80211 wireless networks in Proceedings of IEEE ICCPrsquo02 2002 pp 527ndash536

[81] A Lyakhov M Yakimov Analytical Study of QoS-oriented multicast in wirelessnetworks EURASIP J Wirel Commun Netw 11 (2011) 1ndash13

[82] SKS Gupta V Shankar S Lalwani Reliable multicast MAC protocol for wirelessLANs in Proceedings of IEEE ICCrsquo03 volume 1 2003 pp 93ndash97

[83] MA Santos J Villalon LO Barbosa F Ramirez-Mireles A new ARQ mechanismfor multicast traffic over IEEE 80211 WLANs in Proceedings of IEEE WMNCrsquo11

2011 pp 1ndash8

[84] J Qureshi CH Foh J Cai Online XOR packet coding Efficient single-hop wire-less multicasting with low decoding delay Comput Commun 39 (2014) 65ndash77

[85] E Ancillotti R Bruno M Conti Design and performance evaluation ofthroughput-aware rate adaptation protocols for IEEE 80211 wireless networks

Perform Eval 66 (12) (2009) 811ndash825[86] A Basalamah H Sugimoto T Sato Rate adaptive reliable multicast MAC proto-

col for WLANs in Proceedings of IEEE VTC-Spring 2006 vol 3 2006 pp 1216ndash

1220[87] S Choi N Choi Y Seok T Kwon Y Choi Leader-based rate adaptive multicas-

ting for wireless LANs in Proceedings of IEEE GLOBECOMrsquo07 2007 pp 3656ndash3660

[88] J Villalon P Cuenca L Orozco-Barbosa Y Seok T Turletti ARSM a cross-layerauto rate selection multicast mechanism for multi-rate wireless LANs IET Com-

mun 1 (5) (2007) 893ndash902

[89] G-H Liaw K-H Tsai S-Y Wang T-L Kao L-C Hwang Y-C Lin Adaptive ratecontrol for broadcasting multimedia streams in IEEE 80211 networks in Pro-

ceedings of IEEE ISNErsquo13 2013 pp 296ndash300[90] M Lacage MH Manshaei T Turletti IEEE 80211 rate adaptation a practical

approach in Proceedings of ACM MSWiMrsquo04 2004 pp 126ndash134

[91] IEEE IEEE Std 80211-2012 Specific requirements Part 11 Wireless LANMedium Access Control (MAC) and Physical Layer (PHY) Specifications March

2012[92] G Boggia P Camarda LA Grieco S Mascolo Feedback-based control for pro-

viding real-time services with the 80211e MAC IEEEACM Trans Netw 15 (2)(2007) 323ndash333

[93] K-Y Lee K-S Cho W Ryu Efficient QoS scheduling algorithm formultimedia services in IEEE 80211e WLAN in Proceedings of IEEE VTC-Fallrsquo11

2011 pp 1ndash6

[94] Y Xiao Performance analysis of priority schemes for IEEE 80211 and IEEE80211e wireless LANs IEEE Trans Wirel Commun 4 (4) (2005) 1506ndash1515

[95] H Zhu I Chlamtac Performance analysis for IEEE 80211e EDCF service differ-entiation IEEE Trans Wirel Commun 4 (4) (2005) 1779ndash1788

[96] Z Tao S Panwar Throughput and delay analysis for the IEEE 80211e enhanceddistributed channel access IEEE Trans Commun 54 (4) (2006) 596ndash603

[97] D Xu T Sakurai HL Vu T Sakurai An access delay model for IEEE 80211e

EDCA IEEE Trans Mob Comput 8 (2) (2009) 261ndash275[98] Q Zhao DHK Tsang T Sakurai A scalable and accurate nonsaturated IEEE

80211e EDCA model for an arbitrary buffer size IEEE Trans Mob Comput 12(12) (2013) 2455ndash2469

[99] ND Taher YG Doudane BE Hassan N Agoulmine Towards voicevideo appli-cation support in 80211e WLANs A model-based admission control algorithm

Comput Commun 39 (2014) 41ndash53

[100] I Kadota A Baiocchi A Anzaloni Kalman filtering estimate of the numbers ofactive queues in an 80211e EDCA WLAN Comput Commun 39 (2014) 54ndash64

[101] R-G Cheng C-J Chang C-Y Shih Y-S Chen A new scheme to achieve weightedfairness for WLAN supporting multimedia services IEEE Trans Wirel Commun

5 (5) (2006) 1095ndash1102[102] P Patras A Banchs P Serrano A control theoretic scheme for efficient video

transmission over IEEE 80211e EDCA WLANs ACM Trans Multimed Comput

Commun Appl 8 (3) (2012) 291ndash2923[103] C Cano B Bellalta A Sfairopoulou J Barcelo Tuning the EDCA parameters in

WLANs with heterogeneous traffic a flow-level analysis Comput Netw 54 (13)(2010) 2199ndash2214

[104] Y Xiao FH Li B Li Bandwidth sharing schemes for multimedia traffic in theIEEE 80211e contention-based WLANs IEEE Trans Mob Comput 6 (7) (2007)

815ndash831

[105] Y Xiao L Li Voice and video transmissions with global data parameter con-trol for the IEEE 80211e enhance distributed channel access IEEE Trans Parallel

Distrib Syst 15 (11) (2004) 1041ndash1053[106] H Liu Y Zhao Adaptive EDCA algorithm using video prediction for multimedia

IEEE 80211e WLAN in Proceedings of ICWMCrsquo06 July 2006 pp 1ndash10[107] A Ksentini M Naimi A Gueroui Toward an improvement of H264 video trans-

mission over IEEE 80211e through a cross-layer architecture IEEE Commun

Mag 44 (1) (2006) 107ndash114[108] W He K Nahrstedt X Liu End-to-end delay control of multimedia applications

over multihop wireless links ACM Trans Multimed Comput Commun Appl 5(2) (2008) 161ndash1620

[109] IEEE IEEE 80211 TGah Status of Project IEEE 80211ah httpwwwieee802org11Reportstgah_updatehtm 2015

[110] T Adame A Bel B Bellalta J Barcelo M Oliver IEEE 80211ah the WiFi ap-proach for M2M communications IEEE Wirel Commun Mag 21 (6) (2014) 144ndash

152

[111] T Adame A Bel B Bellalta J Barcelo J Gonzalez M Oliver Capacity analysisof IEEE 80211ah WLANs for M2M communications Proceedings of MACOMrsquo13

Springer 2013 pp 139ndash155[112] E Khorov A Lyakhov A Krotov A Guschin A survey on IEEE 80211ah An en-

abling networking technology for smart cities Comput Commun 58 (53ndash69)(2015)

[113] M Park IEEE 80211ah sub-1-GHz license-exempt operation for the internet of

things IEEE Commun Mag 53 (9) (2015) 145ndash151[114] M Meo EL Rouzic R Cuevas C Guerrero Research challenges on energy-

efficient networking design Comput Commun 50 (2014) 187ndash195[115] S Aust T Ito Sub 1 GHz wireless LAN propagation path loss models for

urban smart grid applications Proceedings of IEEE ICNCrsquo12 IEEE 2012 pp 116ndash120

[116] J Lansford IEEE 80211-11498r0 TGah Channel Models April 2011

[117] A Hazmi J Rinne M Valkama Feasibility study of IEEE 80211 ah radio tech-nology for IoT and M2M use cases Proceedings of IEEE GC Workshopsrsquo12 IEEE

2012 pp 1687ndash1692[118] W Sun M Choi S Choi IEEE 80211ah a long range 80211 WLAN at sub 1 GHz

J ICT Stand 1 (1) (2013) 83ndash108[119] S Aust R Prasad I Niemegeers Outdoor long-range WLANs a lesson for IEEE

80211 ah IEEE Commun Surv Tutor (2015)

[120] GM Dias B Bellalta S Oechsner Reducing the energy consumption in WSNSa data scientific mechanism 2015 arXiv150908778

[121] D Jung R Kim H Lim Power-saving strategy for balancing energy and delayperformance in WLANs Comput Commun 50 (2014) 3ndash9

[122] L Zheng L Cai J Pan M Ni Performance analysis of grouping strategy for denseIEEE 80211 networks in Proceedings of IEEE GLOBECOMrsquo13 2013 pp 219ndash

224

[123] A Bel T Adame B Bellalta J Barcelo J Gonzalez M Oliver CAS-based channelaccess protocol for IEEE 80211ah WLANs Proceedings of European Wirelessrsquo14

VDE 2014 pp 1ndash6[124] CW Park D Hwang T-J Lee Enhancement of IEEE 80211ah MAC for M2M

Communications IEEE Commun Lett 18 (7) (2014) 1151ndash1154

24 B Bellalta et al Computer Communications 75 (2016) 1ndash25

[125] RP Liu GJ Sutton IB Collings Power save with offset listen interval for IEEE80211ah smart grid communications in Proceedings of IEEE ICCrsquo13 IEEE 2013

pp 4488ndash4492[126] M Hasan E Hossain D Niyato Random access for machine-to-machine com-

munication in LTE-advanced networks issues and approaches IEEE CommunMag 51 (6) (2013) 86ndash93

[127] R Davies M Ghosh Field trials of DVB-T sensing for TV White Spaces in Pro-ceedings of IEEE DySPANrsquo11 May 2011 pp 285ndash296

[128] M Nekovee Quantifying the availability of TV White Spaces for cognitive radio

operation in the UK in Proceedings of IEEE ICCrsquo09 2009 pp 1ndash16[129] A Domingo B Bellalta M Oliver White Spaces in UHF band Catalonia case

study and impact of the Digital Dividend in Proceedings of EUNICErsquo12 Springer2012 pp 33ndash40

[130] Unlicensed Operation in the TV Broadcast Bands ndash Third Memorandum Opinionand Order 2012

[131] Ofcom TV White Spaces ndash a consultation on white space device requirements

httpstakeholdersofcomorgukconsultationswhitespaces November 2012[132] J van de Beek J Riihijaumlrvi A Achtzehn P Maumlhoumlnen TV White Space in Europe

IEEE Trans Mob Comput 11 (2) (2012) 178ndash188[133] L Sanabria-Russo J Barcelo A Domingo B Bellalta Spectrum Sensing with

USRP-E110 Proceedings of MACOMrsquo12 Springer 2012 pp 79ndash84[134] L Bedogni A Achtzehn M Petrova P Maumlhoumlnen Smart meters with TV gray

spaces connectivity a feasibility study for two reference network topologies in

Proceedings of IEEE SECONrsquo14 2014[135] T Novlan K Rele S Srikathyayani Coverage and density study of Wi-Fi in the

TV White Spaces 2010[136] T Rappaport Wireless Communications Principles and Practice second edition

Prentice Hall PTR 2001[137] AB Flores RE Guerra EW Knightly IEEE 80211af a standard for TV white

space spectrum sharing IEEE Commun Mag 51 (10) (2013) 92ndash100

[138] IEEE IEEE 80211af-2013 Local and metropolitan area networks Part 11 Wire-less LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications

Amendment 5 Television White Spaces (TVWS) Operation httpstandardsieeeorgfindstdsstandard80211af-2013html 2013

[139] L Simic M Petrova P Maumlhoumlnen Wi-Fi but not on steroids performance anal-ysis of a Wi-Fi-like network operating in TVWS under realistic conditions in

Proceedings of ICCrsquo12 2012 pp 1533ndash1538

[140] M Nekovee Cognitive radio access to TV White Spaces spectrum opportunitiescommercial applications and remaining technology challenges in Proceedings

of IEEE DySPANrsquo10 2010 pp 1ndash10[141] L Bedogni A Trotta MD Felice L Bononi Machine-to-machine communica-

tion over TV White Spaces for smart metering applications in Proceedings ofIEEE ICCCN13-SEPrsquo13 2013

[142] Karol Andersson Carlson Wireless Technologies Super Wi-Fi White Paper

March 2011[143] TX Brown DC Sicker Can cognitive radio support broadband wireless access

in Proceedings of IEEE DySPANrsquo07 2007 pp 123ndash132[144] IF Akyildiz BF Lo R Balakrishnan Cooperative spectrum sensing in cognitive

radio networks a survey Phys Commun 4 (1) (2011) 40ndash62[145] K Chowdhury R Doost-Mohammady W MEleis MD Felice L Bononi Coop-

eration and communication in cognitive radio networks based on TV spectrumexperiments in Proceedings of IEEE WoWMoMrsquo11 2011 pp 1ndash9

[146] IEEE IEEE 19006 Working Group on Spectrum Sensing Interfaces and Data

Structures for Dynamic Spectrum Access and other Advanced Radio Commu-nication Systems httpgrouperieeeorggroupsdyspan6 2011

[147] IEEE IEEE 80219 Task Group 1 ndash Wireless Coexistence in the TV White Spacehttpieee802org19pubTG1html

[148] IEEE IEEE 19004 Architectural Building Blocks Enabling Network-Device Dis-tributed Decision Making for Optimized Radio Resource Usage in Heterogeneous

Wireless Access Networks httpgrouperieeeorggroupsdyspan4 2009

[149] K Bian J-M Park B Gao Cognitive radio networks medium access control forcoexistence of wireless systems Springer Book Springer 2014 pp 1ndash170

[150] L Bedogni MD Felice F Malabocchia L Bononi Indoor communication overTV gray spaces based on spectrum measurements in Proceedings of IEEE

WVNCrsquo14 2014[151] C Stevenson G Chouinard S Shellhammer W Caldwell IEEE 80222 the first

cognitive radio wireless regional area network standard IEEE Commun Mag 47

(1) (2009) 130ndash138[152] R Funada F Kojima H Harada IEEE 802154m the first low rate wireless

personal area networks operating in TV white space in Proceedings of IEEEICONrsquo12 December 2012 pp 326ndash332

[153] W Webb (Ed) Understanding Weightless Cambridge University Press 2012[154] C Ghosh S Roy D Cavalcanti Coexistence challenges for heterogeneous cogni-

tive wireless networks in TV white spaces IEEE Wirel Commun 18 (4) (2011)

22ndash31[155] S Kulac A Eksim MH Sazli Effective cooperative spectrum sensing in IEEE

80222 standard with time diversity in Proceedings of IEEE ACTEArsquo09 July2009 pp 528ndash531

[156] G Villardi Y Alemseged C Sun C-S Cum T Nguyen T Baykas H Harada En-abling coexistence of multiple cognitive networks in TV white space IEEE Wirel

Commun 18 (4) (2011) 32ndash40

[157] J Wang T Baykas S Filin MA Rahman C Song H Harada Coexistence pro-tocol design for autonomous decision-making systems in TV white space in

Proceedings of IEEE WCNCrsquo12 April 2012 pp 3249ndash3254[158] T Henderson G Pei R Groves T Bosaw M Rush C Ghosh S Roy Wireless

network coexistence ndash Boeing project report Technical report 2009

[159] Y Zhang R Yu M Nekovee Y Liu S Xie S Gjessing Cognitive machine-to-machine communications visions and potentials for the smart grid IEEE Netw

26 (3) (2012) 6ndash13[160] M Fadda M Murroni V Popescu A cognitive radio indoor HD1T V multi-

vision system in the TV white spaces in Proceedings of IEEE ICCErsquo12 2012pp 27ndash28

[161] WS Hoi-Sheung J Walrand J Mo McMAC a parallel rendezvous multi-channelMAC protocol in Proceedings of WCNC 2007 IEEE 2007 pp 334ndash340

[162] Z Gu Q-S Hua W Dai Fully distributed algorithms for blind rendezvous in

cognitive radio networks in Proceedings of ACM MobiHoc 2014 IEEE 2014pp 155ndash165

[163] P Ren Y Wang Q Du J Xu A survey on dynamic spectrum access protocolsfor distributed cognitive wireless networks EURASIP J Wirel Commun Netw

(2012) 1ndash21[164] MD Felice AJ Ghandour H Artail L Bononi Integrating spectrum database

and cooperative sensing for cognitive vehicular networks in Proceedings of

IEEE VTCrsquo13 Las Vegas Nevada USA September 2013[165] MD Felice KR Chowdhury L Bononi Cooperative spectrum management in

cognitive vehicular ad hoc networks in Proceedings of IEEE VNCrsquo11 Amster-dam Netherlands November 2011 pp 14ndash16

[166] MD Felice AJ Ghandour H Artail L Bononi Analyzing the potential of cooper-ative cognitive radio technology on inter-vehicle communications in Proceed-

ings of IFIP Wireless Daysrsquo10 Venice Italy October 2010 pp 20ndash22

[167] L Bedogni MD Felice A Trotta L Bononi Distributed mobile femto-databasesfor cognitive access to TV White Spaces using PAWS in Proceedings of IEEE

VTC-Fallrsquo14 Vancouver Canada September 2014[168] MA McHenry PA Tenhula D McCloskey DA Roberson CS Hood Chicago

spectrum occupancy measurements amp analysis and a long-term studies pro-posal in Proceedings of TAPASrsquo06 2006

[169] P Avez PV Wesemael A Bourdoux A Chiumento S Pollin V Moeyaert Tuning

the LongleyndashRice propagation model for improved TV white space detection inProceedings of IEEE SCVTrsquo12 November 2012 pp 1ndash6

[170] PA Zandbergen Accuracy of iPhone locations a comparison of assisted GPSWiFi and cellular positioning Trans GIS 13 (2009) 5ndash26

[171] MD Felice KR Chowdhury W Meleis L Bononi To sense or to transmit alearning-based spectrum management scheme for cognitive radio mesh net-

works in Proceedings of IEEE WiMESHrsquo10 June 2010 pp 1ndash6

[172] R Doost-Mohammady KR Chowdhury Design of spectrum database assistedcognitive radio vehicular networks in Proceedings of ICST CROWNCOMrsquo12

2012[173] D Gurney G Buchwald L Ecklund SL Kuffner J Grosspietsch Geo-location

database techniques for incumbent protection in the TV White Space in Pro-ceedings of IEEE DySPANrsquo08 October 2008 pp 1ndash9

[174] JM Peha Spectrum sharing in the grey space Telecommun Policy 37 (2ndash3)

(2013) 167ndash177[175] LB Le E Hossain Resource allocation for spectrum underlay in cognitive radio

networks IEEE Trans Wirel Commun 7 (12) (2008) 5306ndash5315[176] ARUBA Mobility Controllers httpwwwarubanetworkscomproducts

mobility-controllers March 2014[177] CISCO Cisco Wireless Control System httpwwwciscocomcenusproducts

wirelesswireless-control-systemindexhtml March 2014[178] Y Cui X Ma J Liu L Wang Y Ismailov Policy-based flow control for multi-

homed mobile terminals with IEEE 80211u standard Comput Commun 39

(2014) 33ndash40[179] R Murty J Padhye A Wolman M Welsh Dyson an architecture for extensible

wireless LANs Proceedings of ACM USENIXrsquo10 USENIXATCrsquo10 USENIX Associa-tion Berkeley CA USA 2010 pp 1ndash15

[180] R Murty J Padhye A Wolman M Welsh An architecture for extensible wirelessLANs in Proceedings of ACM HotNetsrsquo08 October 2008

[181] R Murty J Padhye A Wolman B Zill Designing high performance enterprise

Wi-Fi networks in Proceedings of ACM USENIXrsquo08 April 2008[182] P Zerfos G Zhong J Cheng H Luo S Lu JJ-R Li DIRAC a software-based wire-

less router system in Proceedings of ACM MOBICOMrsquo03 MobiCom rsquo03 ACMNew York NY USA 2003 pp 230ndash244

[183] V Shrivastava N Ahmed S Rayanchu S Banerjee S Keshav K Papagian-naki A Mishra CENTAUR realizing the full potential of centralized WLANs

through a hybrid data path in Proceedings of ACM MOBICOMrsquo09 2009

pp 297ndash308[184] I Tinnirello G Bianchi P Gallo D Garlisi F Giuliano F Gringoli Wireless MAC

processors programming MAC protocols on commodity hardware in Proceed-ings of IEEE INFOCOMrsquo12 IEEE 2012 pp 1269ndash1277

[185] G Bianchi P Gallo D Garlisi F Giuliano F Gringoli I Tinnirello MAClets ac-tive MAC protocols over hard-coded devices in Proceedings of ACM CoNEXTrsquo12

2012 pp 229ndash240

[186] N McKeown T Anderson H Balakrishnan G Parulkar OpenFlow enabling in-novation in campus networks ACM SIGCOMM Comput Commun Rev 38 (2)

(2008) 69ndash74[187] N Gude T Koponen J Pettit B Pfaff M Casado N McKeown S Shenker NOX

towards an operating system for networks ACM SIGCOMM Comput CommunRev 38 (3) (2008) 105ndash110

[188] L Suresh J Schulz-Zander R Merz A Feldmann T Vazao Towards pro-

grammable enterprise WLANS with Odin in Proceedings of ACM HotSDNrsquo122012 pp 115ndash120

[189] J Vestin P Dely A Kassler N Bayer H Einsiedler C Peylo CloudMAC towardssoftware defined WLANs ACM SIGMOBILE Mob Comput Commun Rev 16 (4)

(2013) 42ndash45

B Bellalta et al Computer Communications 75 (2016) 1ndash25 25

[

[

[

[190] J Vestin A Kassler QoS enabled WiFi MAC layer processing as an example of aNFV service in Proceedings of the 1st IEEE Conference on Network Softwariza-

tion (NetSoft) April 2015 pp 1ndash9[191] J Vestin A Kassler QoS management for WiFi MAC layer processing in the cloud

ndash demo description in Proceedings of IEEE MSWiMrsquo15 2015[192] P Dely A Kassler L Chow N Bambos N Bayer H Einsiedler C Peylo BEST-AP

non-intrusive estimation of available bandwidth and its application for dynamicaccess point selection Comput Commun 39 (2014) 78ndash91

[193] USRP Universal Software Radio Peripheral httpwwwettuscom

[194] WARP Wireless Open-access Research Platform httpswarpprojectorgtrac[195] SORA Microsoft Research Software Radio httpresearchmicrosoftcomen-us

projectssora[196] F Gringoli L Nava OpenFWWF website httpwwwingunibsitsimopenfwwf

2009[197] P Salvador L Cominardi F Gringoli P Serrano A first implementation and eval-

uation of the IEEE 80211aa group addressed transmission service ACM SIG-

COMM Comput Commun Rev 44 (1) (2013) 35ndash41[198] L Sanabria-Russo F Gringoli J Barcelo B Bellalta Implementation and exper-

imental evaluation of a collision-free MAC Protocol for WLANs in Proceedingsof IEEE ICC 2015 2015

[199] E Magistretti O Gurewitz EW Knightly 80211ec collision avoidance withoutcontrol messages Proceedings of ACM MOBICOMrsquo12 ACM 2012 pp 65ndash76

200] J Xiong K Sundaresan K Jamieson MA Khojastepour S Rangarajan MIDAS

empowering 80211ac networks with multiple-input distributed antenna sys-tems in Proceedings of ACM CoNEXTrsquo14 ACM 2014 pp 29ndash40

[201] S Aust R Prasad I Niemegeers Advances in wireless M2M and IoT rapid SDR-prototyping of IEEE 80211ah in Proceedings of IEEE LCNrsquo14 2014 pp 290ndash292

[202] S Gollakota F Adib D Katabi S Seshan Clearing the RF smog making 80211nrobust to cross-technology interference ACM SIGCOMM Comput Commun Rev

41 (4) (2011) 170ndash181

[203] C Gabriel Wireless broadband alliance industry report 2013 global trends inpublic Wi-Fi WBA ndash Maravedis-Rethink (November 2013)

204] F Rebecchi MD de Amorim V Conan A Passarella R Bruno M Conti Dataoffloading techniques in cellular networks a survey IEEE Commun Surv Tutor

17 (2) (2014) 580ndash603[205] K Samdanis T Taleb S Schmid Traffic offload enhancements for eUTRAN IEEE

Commun Surv Tutor 14 (3) (2012) 884ndash896

206] D Astely E Dahlman G Fodor S Parkvall J Sachs LTE release 12 and beyondIEEE Commun Mag 51 (7) (2013) 154ndash160

[207] S Landstrom A Furuskar K Johansson L Falconetti F Kronestedt Heteroge-neous Networks (HetNets) ndash an approach to increasing cellular capacity and

coverage Ericsson Rev (2011)[208] JG Andrews Seven ways that HetNets are a cellular paradigm shift IEEE Com-

mun Mag 51 (3) (2013) 136ndash144

[209] 3GPP 3GPP TS 23261 IP flow mobility and seamless Wireless Local Area Net-work (WLAN) offload (Rel 10) 2011

[210] A de la Oliva CJ Bernardos M Calderon T Melia JC Zuniga IP flow mobilitysmart traffic offload for future wireless networks IEEE Commun Mag 49 (10)

(2011) 124ndash132[211] 3GPP 3GPP TS 24312 Access Network Discovery and Selection Function

(ANDSF) Management Object (MO) (Rel 10) 2011[212] L Wang G-SGS Kuo Mathematical modeling for network selection in hetero-

geneous wireless networks ndash a tutorial IEEE Commun Surv Tutor 15 (1) (2013)

271ndash292[213] DH Hagos R Kapitza Study on performance-centric offload strategies for LTE

networks in Proceedings of IEEEIFIP WMNCrsquo13 April 2013 pp 1ndash10[214] M Bennis M Simsek A Czylwik W Saad S Valentin M Debbah When cellular

meets WiFi in wireless small cell networks IEEE Commun Mag 51 (6) (June2013) 44ndash50

[215] K Lee J Lee Y Yi I Rhee S Chong Mobile data offloading how much can WiFi

deliver IEEEACM Trans Netw 21 (2) (2013) 536ndash550

[216] L Hu C Coletti N Huan IZ Kovaacutecs B Vejlgaard R Irmer N Scully Realistic in-door Wi-Fi and femto deployment study as the offloading solution to LTE macro

networks in Proceedings of IEEE VTC Fallrsquo12 2012 pp 1ndash6[217] F Mehmeti T Spyropoulos Performance analysis of ldquoon-the-spotrdquo mobile data

offloadingpdf in Proceedings of IEEE GLOBECOMrsquo13 2013 pp 1577ndash1583[218] S Singh HS Dhillon JG Andrews Offloading in heterogeneous networks mod-

eling analysis and design insights IEEE Trans Wirel Commun 12 (5) (May2013) 2484ndash2497

[219] T Wang W Jia G Xing M Li Exploiting statistical mobility models for efficient

Wi-Fi deployment IEEE Trans Veh Technol 62 (1) (2013) 360ndash373[220] E Bulut BK Szymanski WiFi access point deployment for efficient mobile data

offloading in Proceedings of ACM PINGENrsquo12 ACM New York NY USA 2012pp 45ndash50

[221] EMR Oliveira AC Viana From routine to network deployment for data offload-ing in metropolitan areas in Proceedings of IEEE SECONrsquo14 June 2014 pp 126ndash

134

[222] H Elsawy E Hossain M Haenggi Stochastic geometry for modeling analysisand design of multi-tier and cognitive cellular wireless networks a survey IEEE

Commun Surv Tutor 15 (3) (2013) 996ndash1019[223] AJ Nicholson BD Noble BreadCrumbs forecasting mobile connectivity in

Proceedings of ACM MOBICOMrsquo08 2008 pp 46ndash57[224] S Gatmir-Motahari H Zang P Reuther Time-clustering-based place prediction

for wireless subscribers IEEEACM Trans Netw 21 (5) (2013) 1436ndash1446

[225] A Balasubramanian R Mahajan A Venkataramani Augmenting mobile 3G Us-ing WiFi in Proceedings of ACM MobiSysrsquo10 ACM New York NY USA 2010

pp 209ndash222[226] N Ristanovic J-YL Boudec A Chaintreau V Erramilli Energy efficient offload-

ing of 3G Networks in Proceedings of IEEE MASSrsquo11 2011 pp 202ndash211[227] I Trestian S Ranjan A Kuzmanovic A Nucci Taming the mobile data deluge

with drop zones IEEEACM Trans Netw 20 (4) (2012) 1010ndash1023

[228] F Malandrino C Casetti C Chiasserini M Fiore Optimal content downloadingin vehicular networks IEEE Trans Mob Comput 12 (7) (2013) 1377ndash1391

[229] A Bazzi BM Masini A Zanella G Pasolini IEEE 80211p for cellular offloadingin vehicular sensor networks Comput Commun 60 (2015) 97ndash108

[230] X Zhuo W Gao G Cao H Hua An incentive framework for cellular traffic of-floading IEEE Trans Mob Comput 13 (3) (2014) 541ndash555

[231] L Lei Z Zhong C Lin X Shen Operator controlled device-to-device communi-

cations in LTE-advanced networks IEEE Wirel Commun 19 (3) (2012) 96ndash104[232] B Han P Hui VSA Kumar MV Marathe J Shao A Srinivasan Mobile data

offloading through opportunistic communications and social participation IEEETrans Mob Comput 11 (5) (2012) 821ndash834

[233] MV Barbera AC Viana MD de Amorim J Stefa Data offloading in social mo-bile networks through VIP delegation Ad Hoc Netw 19 (2014) 92ndash110

[234] X Wang M Chen Z Han D Wu T Kwon TOSS traffic offloading by social net-

work service-based opportunistic sharing in mobile social networks in Pro-ceedings of IEEE INFOCOM April 2014 pp 2346ndash2354

[235] W Peng F Li X Zou J Wu The virtue of patience offloading topical cellularcontent through opportunistic links in Proceedings of IEEE MASSrsquo13 Oct 2013

pp 402ndash410[236] J Whitbeck Y Lopez J Leguay V Conan MD de Amorim Push-and-track sav-

ing infrastructure bandwidth through opportunistic forwarding Pervasive MobComput 8 (5) (2012) 682ndash697

[237] L Valerio R Bruno A Passarella Cellular traffic offloading via opportunistic net-

working with reinforcement learning Comput Commun (2015) doi101016jcomcom201509004

[238] R Bruno A Masaracchia A Passarella Offloading through opportunistic net-works with dynamic content requests in Proceedings of IEEE CARTOONrsquo14

2014[239] M Conti Computer communications present status and future challenges

Comput Commun 37 (2014) 1ndash4