utilizing beamforming antennas for wireless multi-hop networks

1

Utilizing Beamforming Antennas for Wireless Multi-hop Networks

Romit Roy Choudhury

Dept. of Computer ScienceUniversity of Illinois at Urbana-Champaign

romit

Thank you for the introduction, Prof. XXX, and thank you all for taking the time to come to this talk. My area of research is in networking and systems, and today I would be presenting my work on "Utilizing Beamforming ..."

2

Wireless Multihop Networks

B

Collection of wireless hosts Relay packets on behalf of each other Together form an arbitrary topology May be connected to wired infratructure

2 reasons to prefer multihop Capacity and Power constraint

A

C

D

romit

Let me begin by talking a lil bit about wireless multihop networks, what it is now, and what it might be in the future. By definition, wireless multihop networks are formed by a set of wireless stations, that cooperate among themselves to relay packets from any source to its desired destination. Together, these wireless hosts form an arbitrary topology, and the entire multihop network may or may not be connected to a wired infrastructure, like the internet.Here is a simple pictorial representation of a wireless network. If node A intends to communicate to node D, it transmits the packet to node B. Node B relays the packet to node C, which in turn relays the packet to D. When node D receives the packet, it realizes that the packet is meant for itself, and thus does not relay it further.Now, an important question arises at this point. Why must A relay the packets through B and C, as opposed to shouting to D, and communicating the packet in one shot.There are 2 good reasons that motivate relaying.One, a recent result by Gupta and Kumar from UIUC has shown that relaying improves the capacity of wireless networks. The intuition is as follows: If A shouts to D, then while A is shouting, no other nodes will be able to communicate. However, if A whispers to B, several nodes can whisper in parallel, and the results prove that the overall gains from whispering is larger, even though the packet has to go over multiple hops.The second reason for multihoping comes from power considerations. Observe that A would require a lot of power to communicate to D. Since several wireless devices are often battery constrained, such communication may not be feasible, making relaying the only feasible option.This made the study of wireless multihop networks interesting, and in the last deacde or so, its popularity has increased immensely.next slide

3

Applications

Multihop networks gained wide popularity Began with military applications

New applications proving to be commercial Already beginning to appear around us

For Example …

romit

Now, while the initial popularity of these networks came from military applications where soldiers were moving around with wireless devices and communicating among themselves, newly envisioned applications are proving to be more commercial in nature. These applications are already beginning to appear around us, and will only proliferate in the future, over a variety of sizes and scale. For example ...

4


RFID

RFID

Sensors

RFID Readers

romit

... at the smallest scale, RFIDs and sensors are being embeded in the environment, and they together form a multihop network capable of sensing, tracking, and monitoring applications.People are injecting RFID tags under their skins and opening doors with a wave of their hand. Sensor networks are being mounted around houses as a means of security, or in hospitals for health monitoring systems. And then, these mini-scale networks are being dynamically programmed through a smart phone or PDA, based on the need of the occasion.The future is likely to see applications where every object in the environment can be electronically queried or programmed, changing the way in which we do the simplest of things

5


romit

At the next level, wireless personal area networks are connecting multiple gadgets within the home, and offering the ability to access information anywhere and anytime. For example, a bluetooth headset might be able to stream music from the IPOD in the neighboring apartment, or a music clip will be played from the camcorder on several LCDs in the house. A large fraction of the communication will prove to be over wireless.

6


romit

At a level yet higher, routers are being mounted on street lampposts and rooftops of houses, forming a mesh network that together illuminates the city with ubiquitous connectivity. Mobile users in vehicles or mobile pedestrians would be able to access information anywhere-anytime.The mesh network will form the wireless backbone of the city, and will eventually ...next slide

7


Internet

romit

be connected to the wired internet infrastructure, feeding urban and rural mesh networks with internet connectivity.While this is a projection of the future, realizing such a networking system is associated with a large number of challenges. These challenges are non-trivial and diverse.For example, ...

8

Protocol Design

Numerous challenges Connectivity (nodes can be mobile) Capacity (increasing demand) Reliability (channels fluctuate) Security QoS …

Many protocols designed

One commonality among most protocols

9

Omnidirectional Antennas

Internet

romit

All these protocols, implicitly or explicitly, assumes an omnidirectional antenna at its physical layer. As shown in this animation, energy radiated using an omnidirectional antenna scatters almost equally in all directions. This can prove to be a constraint, especially in multihop networks where the number of parallel communication is the key to good performance. Moreover, with denser deployments of wireless networks in the future, the situation can degrade significantly. Managing wireless interference efficiently emerged to be a very important problem, and a scalable solution became necessary.

10

CTS = Clear To Send

RTS = Request To Send

IEEE 802.11 with Omni Antenna

D

Y

S

M

K

RTS

CTS

X

11


D

Y

S

X

M

Ksilenced

silenced

silenced

silencedData

ACK

12


DS

X

M

Ksilenced

silenced

silenced

Y

silencedData

ACK

D

silenced

E

silencedA

silencedC

silenced

F

silenced

B

silenced

G

silenced

`` Interference management ``A crucial challenge for dense multihop networks

13

Managing Interference

Several approaches Dividing network into different channels Power control Rate Control …

Our focus Exploiting antenna capabilities

•Switched Beam Antennas•Steerable Antennas•Reconfigurable Antennas, etc.

Many becoming commercially available

For example …

14

Electronically Steerable Antenna [ATR Japan]

Higher frequency, Smaller size, Lower cost Capable of Omnidirectional mode and Directional

mode

romit

I had the opportunity to work on a testbed at the Advanced Telecommunications Research Labs in Japan, where they have built an electronically steerable antenna, called ESPAR antennas. This is the ESPAR antenna prototype for the 2.4 Ghz band. While this seems quite big, here is the same antenna for the .....

15

Switched and Array Antennas

On poletop or vehicles Antennas bigger No power constraint

romit

.....Now, given that there are several types of antennas, it appears necessary to generalize the antenna characteristics in a manner that is relevant to higher layer researchers. In other words, we need an abstraction that can decouple protocol design from the exact details of antennas.

16

Antenna Abstraction

3 Possible antenna modes Omnidirectional mode Single Beam mode Multi-Beam mode

Higher Layer protocols select1. Antenna Mode2. Direction of Beam

17

Antenna Beam

Energy radiated toward desired direction

A

Pictorial Model

A

Main Lobe (High gain)

Sidelobes (low gain)

romit

Let me try to present to you a 2-slide overview of antenna characteristics. With directional antennas, energy is radiated toward a desired direction. An example radiation pattern has been shown in this figure - Node A transmits most of the energy towards the right, and some energy leaks out in other directions. The amount of energy that is radated in a given direction is called the gain of the antenna, and is proportional to the length of this line joining the node and the radiation pattern in that direction.

18

Directional Reception

Directional reception = Spatial filtering Interference along straight line joining

interferer and receiver

A BC

D

Signal

Interference

No Collision at A

A B

C

D

Signal

Interference

Collision at A

19

Will attaching such antennas at the radio layer yield most of the benefits ?

Or

Is there need for higher layer protocol support ?

20

We design a simple baseline MAC protocol(a directional version of 802.11)

We call this protocol DMAC and investigate its behavior through simulation

21

Remain omni while idle Nodes cannot predict who will trasmit to it

DMAC Example

D

Y

S

X

22

RTS

DMAC Example

D

Y

S

X

Assume S knows direction of D

23

RTS

RTSCTS

DMAC Example

D

Y

S

X

DATA/ACK

X remembers not to initiate transmission toward D

24

Intuitively

Performance benefits appear obvious

25

However …

0

100

200

300

400

0 500 1000 1500 2000 2500

Sending Rate (Kbps)

Agg

rega

te T

hrou

ghpu

t (K

bps)

802.11

DMAC

Sending Rate (Kbps)

Th

rou

gh

pu

t (K

bp

s)

Romit Roy Choudhury

what are beamwidths? sidelobes? (see mobicom paper)

26

Clearly, attaching sophisticated antenna hardware is not sufficient

Simulation traces revealed various new challenges

Motivates higher layer protocol design

27

Research Contribution

Identified several new challenges

Designed MAC and Routing protocols MMAC, ToneDMAC, CaDMAC DDSR, CaRP Cross-Layer protocols – Anycasting Improved understanding of theoretical capacity

Developed prototype testbed

28





29

Self Interferencewith Directional MAC

New Challenges [Mobicom 02]

30

Unutilized Range [Best Paper, PWC 03]

Longer range causes interference downstream Offsets benefits

Network layer needs to utilize the long range Or, MAC protocol needs to reduce transmit power

A B CData

D

route

31

New Hidden Terminal Problemswith Directional MAC

New Challenges II …

32

New Hidden Terminal Problem [IEEE TMC]

Due to gain asymmetry

Node A may not receive CTS from C i.e., A might be out of DO-range from C

B CAData

RTSCTS

33

New Hidden Terminal Problem


Node A later intends to transmit to node B A cannot carrier-sense B’s transmission to C

B CAData

Carrier Sense

RTSCTS

34

New Hidden Terminal Problem


Node A may initiate RTS meant for B A can interfere at C causing collision

B CADataRTS

Collision

35

Deafnesswith Directional MAC

New Challenges III …

36

Deafness [ICNP 04]

Node N initiates communication to S S does not respond as S is beamformed toward

D N cannot classify cause of failure Can be collision or deafness

S D

N

Data

RTS

M

37

Channel Underutilized

Collision: N must attempt less often Deafness: N should attempt more often

Misclassification incurs penalty (similar to TCP)

S D

N

Data

RTS

M

Deafness not a problem with omnidirectional antennas

Romit Roy Choudhury

talk about a 2-hop flow here itself so that graph on next slide can be of immediate relevance

38

Deafness and “Deadlock”

Directional sensing and backoff ... Causes S to always stay beamformed to D X keeps retransmitting to S without success Similarly Z to X a “deadlock”

DATA

RTS

X

DS

Z

RTS

Romit Roy Choudhury

klara's comment on this case being a starvation and not a deadlock. A deadlock is A -> B -> C -> A. Incorporate this later.

39

MAC-Layer CaptureThe bottleneck to spatial reuse

New Challenges IV …

40

Typically, idle nodes remain in omni mode When signal arrives, nodes get engaged in

receiving the packet Received packet passed to MAC If packet not meant for that node, it is dropped

Capture [HotNets 03]

Wastage because the receiver could accomplish useful communication

instead of receiving the unproductive packet

41

Capture Example

A B

C

D

Both B and D are omni whensignal arrives from A

A B

C

D

B and D beamform to receivearriving signal

42





43

Impact of Capture

Beamforming for transmission and reception only is not sufficient

Antenna control necessary during idle state also

A B

C

DA B

C

D

44

Capture-Aware MAC (CaDMAC) D monitors all incident traffic Identifies unproductive traffic

Beams that receive onlyunproductive packets are turned off

However, turning beams offcan prevent useful communication in future

MAC Layer Solution

A B

C

D

45

CaDMAC turns off beams periodically Time divided into cycles Each cycle consists of

1.Monitoring window + 2. Filtering window

1 2

CaDMAC Time Cycles

1 12 2

cycle

time

All beams remain ON,monitors unproductive beams

Node turns OFF unproductivebeams while it is idle.

Can avoid capture

46

CaDMAC Communication

Transmission / Reception uses only necessary single beam

When node becomes idle, it switches back to appropriate beam pattern Depending upon current time window

47

During Monitoring window, idle nodes are omni

Spatial Reuse in CaDMAC

A B

C

EF

D

48

At the end of Monitoring window CaDMAC identifies unproductive links


A B

C

EF

D

49

During Filtering window use spatial filtering


A B

C

EF

D

Parallel CommunicationsCaDMAC : 3 DMAC & others : ≤ 2Omni 802.11 : 1

50

Network Capacity [TechReport ‘05]

Existing results show Capacity improvement lower bounded by

Results do not consider side lobes of radiation patterns

We consider main lobe and side lobe gains (gm

and gs)

We find capacity upper bounded by

i.e., improvement of

2

O

2

1

1

21

s

m

g

g

n

W

2

s

m

g

gO

CaDMAC still below achievable capacity

51

CaDMAC cannot eliminate capture completely

Happens because CaDMAC cannot choose routes Avoiding capture-prone links A routing

problem

Discussion

YX

AB

52

Routing using Directional AntennasIncorporating capture-awareness

53

Motivating Capture-Aware Routing

YX

AB

YX

AB

D

S

Z DZ

S

Find a route from S to D, given AB exists Options are SXYD, SXZG

Capture No Capture

54

1. Sum capture costs of all beams on the route Capture cost of a Beam j =

how much unproductive traffic incident on Beam j

2. Route’s hop count3. Cost of participation

How many intermediate nodes participate in cross traffic

Measuring Route Cost

X

DS

55

Uroute = Weighted Combination of1. Capture cost (K)2. Participation cost (P)3. Hop count (H)

Weights chosen based on sensitivity analysis

Unified Routing Metric

ijipijrouteij

kroute HPU

56

CaRP Vs DSR

1

2

34

57

CaRP Vs DSR

58

CaRP Vs DSR

59

CaRP Vs DSR

60

CaRP Vs DSR

61

CaRP Vs DSR

62

CaRP Vs DSR

63

CaRP Vs DSR

64

CaRP Vs DSR

DSR CaRP

CaRP prefers a traffic-free direction“Squeezes in” more traffic in given area

65

Performance of CaDMAC

CaDMAC

DMAC

802.11

CBR Traffic (Mbps)

Ag

gre

gate

Th

rou

gh

pu

t (M

bp

s)

CMAC

66

Throughput with CaRP CaRP +CaDMAC

DSR +CaDMAC

DSR +802.11

Ag

gre

gate

Th

rou

gh

pu

t (M

bp

s)

Topology Number

Random Topologies

67





68

Testbed Prototype [VTC 05, Mobihoc05Poster]

Network of 6 laptops using ESPAR antennas ESPAR attached to external antenna port Beams controlled from higher layer via USB

Validated basic operations and tradeoffs Neighbor discovery

•Observed multipath

•60 degrees beamwidth useful

Basic link state routing • Improves route stability

•Higher throughput, less delay

69

Summary

Proliferation of wireless devices motivate better interference management

Most research assume omni antennas Energy radiated in all directions – inefficient

70

Omnidirectional Antennas

Internet

romit

All these protocols, implicitly or explicitly, assumes an omnidirectional antenna at its physical layer. As shown in this animation, energy radiated using an omnidirectional antenna scatters almost equally in all directions. This can prove to be a constraint, especially in multihop networks where the number of parallel communication is the key to good performance. Moreover, with denser deployments of wireless networks in the future, the situation can degrade significantly. Managing wireless interference efficiently emerged to be a very important problem, and a scalable solution became necessary.

71

Summary

Our focus on exploiting antenna capabilities Several challenges appear Existing protocols incapable of leveraging benefits

Designed protocols at MAC and Network layer MMAC, ToneDMAC, CaDMAC Directional DSR, CaRP Analysed the upper bounds on capacity

Developed prototype testbed Capable of utilizing beamforming antennas

72

Summary

Internet

73

Other Work

Sensor Networks Reliable network-wide broadcast [submitted ICDCS]

Exploiting mobility as a carrier of gossip [StoDis 05]

K-Coverage problems

Location management in mobile networks Designed distributed algorithms [IPDPS], [Mobihoc]

Scheduling protocols for 802.11n Combination of CSMA + TDMA schemes [WTS 04]

74

Future Work

Next generation radios (software, cognitive) PHY layer not be sufficient to harness flexibility

Example•When should a radio toggle between TDMA and CSMA

?

•Dynamic channel access needs coordination

Higher layer protocols necessary for decisions

75

Future Work

Exploiting Diversity Opportunistically Especially in the context of improving

reliability

•Link diversity

•Route diversity

•Antenna diversity

•Channel diversity …

My previous work on Anycasting – a first step I intend to continue in this direction

S

A

B

C

D

76

Future Work

Sensor Networks

Strong guarantees

Weak guarantees

Solu

tion C

om

ple

xit

ySensor networks ammenableto probabilistic guarantees.

Need to design protocols that exploit this possibility

77

Future Work

Experimental Testbeds and Prototypes Evaluate protocols in real conditions Intend to build prototype testbed for sensor

and mesh networks

78

Thank You

AcknowledgmentsProf. Nitin Vaidya (advisor)

Members of my research group

CollaboratorsRam Ramanathan (BBN)

Tetsuro Ueda (ATR Labs, Japan)Steve Emeott (Motorola Labs)

Lily Yang, Guoquing Li (Intel Labs)

79

Wireless Networks

Gained popularity over last decade Cellular networks Mesh networks Wireless LANs

Several advantages Users can be mobile Low deployment cost Information access anytime-anywhere

80

Capacity Result

Wireless network capacity is shared by users

Per-user capacity upper bounded by

Implication: Relaying better than shouting

n

WO

B

A

C

D

relay relay

81

Modify

Similar to TCP on deafness – no Typo – reduscing 802.11 – say request to send Multihop in sensor MIMO – and multipath 1 slide on antenna beam/mode

Say “Antenna has 3 modes – omni, single, multi” Higher layer sets 1 mode, and direction before initiating

communication 802.11 just before DMAC 1 minute on MMAC, ToneDMAC, etc. Turning off beams has to be done before comm starts Say shortest hop first – then say DSR

82

RTS = Request To Send

RTS

CTS = Clear To Send

CTS


D

Y

S

X

M

K

83

Enhancing MAC [Mobicom02]

MMAC Transmit multi-hop RTS to far-away receiver Synchronize with receiver using CTS

(rendezvous) Communicate data over long links

84

Routing with Higher Range [PWC03 Best Paper]

Directional routes offer Better connectivity, fewer-hop routes

However, broadcast difficult Sweeping necessary to emulate broadcast

Evaluate tradeoffs Designed directional DSR

85

ToneDMAC’s Impact

Another possible improvement:

Backoff Counter for DMAC flows

Backoff Counter for ToneDMAC flows

time

Ba

cko

ff V

alu

es

86

2-hop flow

DMAC

IEEE 802.11

Dea

fnes

s

87

Neighbor Discovery

Non LOS and multipath important factors However, wide beamwidth (60 degrees)

reasonable envelope

Anechoic Chamber Office Corridor

88

Route Reliability

Routes discovered using sweeping – DO links Data Communication using DD links Improved SINR improves robustness against

fading

89

Optimal Carrier Sense Threshold

When sidelobe abstracted to sphere with gain Gs

T

Sidelobe Gs

Mainlobe Gd

Provided, optimal CS threshold is above the Rx sensitivity thresholdi.e., min {CS_calculate, RxSensitivity}

90

Commercial Antennas …

Paratek (DRWin scanning smart antennas) Beamforming in the RF domain (instead of digital) Multiple simultaneous beams possible, each

steerable http://www.mobileinfo.com/news_2002/issue08/Paratek_Antenna.htm

Motia Inc. (Javelin appliqué to 802.11 cards) Blind beamforming in RF domain (< 2us, within pilot)

CalAmps (DirectedAP offers digital beamforming) Uses RASTER beamforming technology http://www.calamp.com/pro_802_directedap.html

91

Commercial Antennas

Belkin (Pre-N smart antenna router – Airgo tech.) Uses 3 antenna elements for adaptive

beamforming http://www.techonline.com/community/tech_group/37714

Tantivy Communications (switching < 100 nanosec) http://www.prism.gatech.edu/~gtg139k/papers/11-03-025r0-WNG-bene

fitsofSmartAntennasin802.11Networks.pdf

92

While node pairs communicate X misses D’s CTS to S No DNAV toward D

New Hidden Terminal Problem II

D

Y

S

X

DataData

93

While node pairs communicate X misses D’s CTS to S No DNAV toward D X may later initiate RTS toward D, causing

collision

New Hidden Terminal Problem II

D

Y

S

X

Data

RTS

Collision

94

Abstract Antenna Model

N conical beams

Any combination of beams can be turned on

Capable of detecting beam-of-arrival for received packet

95

Source routing protocol (like DSR) Intermediate node X updates route cost from S -

X

Destination chooses route with least cost (Uroute) Routing protocol shown to be loop-free

Protocol Design

X

DS

C1

C2

C3

C5

USX

USD = USX + C2 + C5 + PD + 1

utilizing beamforming antennas for wireless multi-hop networks

Documents