chidume, chidiebere s.€¦ · chidume, chidiebere s. pg/m.eng/2013/65159 evolution (lte) downlink...

IMPROVED SCHEDULER FOR LONG TERM

Ebere Omeje

FACULTY OF ENGINEERINGENGINEERING

DEPARTMENT OF ELECTRONIC ENGINEERING

CHIDUME, CHIDIEBERE S.

PG/M.ENG/2013/65159

IMPROVED SCHEDULER FOR LONG TERM EVOLUTION (LTE) DOWNLINK

TRANSMISSION

Ebere Omeje Digitally Signed by: Content manager’s Name

DN : CN = Webmaster’s name

O= University of Nigeria, Nsukk

OU = Innovation Centre



1


PG/M.ENG/2013/65159

IMPROVED SCHEDULER FOR LONG TERM EVOLUTION (LTE) DOWNLINK

Digitally Signed by: Content manager’s Name

DN : CN = Webmaster’s name

O= University of Nigeria, Nsukka



2

TITLE PAGE

IMPROVED SCHEDULER FOR LONG TERM EVOLUTION

(LTE) DOWNLINK TRANSMISSION

BY


PG/M.ENG/2013/65159

DEPARTMENT OF ELECTRONIC ENGINEERING,

UNIVERSITY OF NIGERIA, NSUKKA

OCTOBER, 2015

3

APPROVAL PAGE

IMPROVED SCHEDULER FOR LTE DOWNLINK TRANSMISSION


(PG/M.ENG/2013/65159)

A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE RE QUIREMENTS

FOR THE AWARD OF MASTER OF ELECTRONIC ENGINEERING

(TELECOMMUNICATION OPTION) IN THE DEPARTMENT OF ELE CTRONIC

ENGINEERING, UNIVERSITY OF NIGERIA, NSUKKA.

CHIDUME CHIDIEBERE,S

(STUDENT)

SIGNATURE: ………………….... DATE… ……….

PROF. C.I. ANI

(SUPERVISOR)

SIGNATURE: ………………….... DATE………….

EXTERNAL EXAMINER

SIGNATURE: ………………….... DATE………….

DR. M.A AHANEKU

4

(HEAD OF DEPARTMENT) SIGNATURE: ………………….... DATE……… ....

PROF. E.S. OBE

(CHAIRMAN, FACULTY POSGRADUATE COMMITTEE)

SIGNATURE: ………………….... DATE……….…

CERTIFICATION

This is to certify that Chidume,Chidiebere S. a master’s degree postgraduate student in the Department of Electronic Engineering and with registration number PG/M.ENG/13/65159has satisfactorily completed the requirements for the award of Master of Engineering (M.ENG) in Electronic Engineering.

______________________

PROF. C.I. ANI

_________________________

DR. M.A. AHANEKU

(SUPERVISOR) (HOD)

_____________________________________________

PROF. E.S. OBE

5

(CHAIRMAN, FACULTY POSTGRADUATE COMMITTEE)

DECLARATION

I, Chidume, ChidiebereS. a postgraduate student of the Department of Electronic Engineering, University of Nigeria, Nsukka declare that the work embodied in this thesis is original and has not been submitted by me in part or in full for any other diploma or degree of this or any other

University.

___________________________


PG/M.ENG/13/65159

_______________

DATE

6

DEDICATION

I dedicate this work to Almighty God for his love and mercies during this research work.

7

ACKNOWLEDGEMENT

I want to first of all express my profound gratitude to the Almighty God; through his infinite mercies, I have been able to bring this project work to a successful conclusion. My special gratitude goes to Prof C.I. Ani for his teachings, patience, love, participation and fatherly guidance throughout the entire period of my course work and the project.

My acknowledgement and gratitude also goes to the Head of Department, Engr.Dr.M.AAhaneku for his deep concern, Engr.Martins, Engr.Obinna, Engr.Ezeja, Engr.Augustin and the entire members of staff of the Department of Electronic Engineering, University of Nigeria Nsukka, for their participation, assistance, encouragement and guidance, May God Almighty bless them all.

I also want to thank Dr.Joe Mom for his contribution and advice. Also my appreciation goes to my friends: Eli Jiya, Nathaniel,Udora,Kelechukwu etc. and all the 2013/2014 set of Master’s degree students of Electronic Engineering Department UNN Nsukka.

8

I will not forget my lovely parents, Mr. and Mrs. Chidume, and my siblings for their love, care, support both financially and spiritually all these years. May God reward all your effort towards me.

ABSTRACT

Long term evolution (LTE) Network is often faced with the challenge of meeting up with the quality of

service (QoS) requirement of the different services supported in the Network. Maintaining a trade-off

between system throughput and fairness among users when making radio resources scheduling

decisions is a very sensitive issue. Several algorithms have been proposed to this effect by different

researchers in an attempt to manage the limited radio resources. One of well-known packet schedulers

known as M-LWDF algorithm is known to support both real time and non-real time services. This

algorithm has been found not to support real time services at a sufficient level. This is due to the fact

that head of line (HOL) delay and packet delay not sufficient to balance the scheduling decision to real

time services thereby degrading its performance. This research work was an attempt to improve the

performance of MLWDF by incorporating bandwidth of flow,�, which is directly proportional to flow

9

weight and reserved rate .This approach used the uncertainty principle of fuzzy logic to calculate new

weight for the different flows, by considering two input parameters from the network which are latency

requirement for real time traffic and throughput for non-real time traffic .The performance of this

algorithm was compared with PF,M-LWDF and EXP/PF schemes using system throughput, packet loss

rate, delay and fairness as a performance indices in an LTE simulator. Results showed that when I-

MLWDF was used, the aggregate throughput for video and VOIP increases considerably by 9.8% and

0.1% respectively when the cell was loaded by 60 users and the throughput for non- real time flow

showed degraded performance of about 90%as against MLWDF. The improved scheme showed that the

packet loss rate for video flow decreases by about 35.71% and 75% when compared with EXP/PF and

MLWDF respectively. The acceptable packet loss rate for VOIP flow is under 3% for both algorithm and

the improved scheme showing better result by about 20%compared to MLWDF. The packet loss rate for

non-real time flow for I-MLWDF showed a poor performance of about 95% as against MLWDF.The delay

for video, VOIP for the improved scheme is under 0.05s and 0.007s respectively as against MLWDF with

delays under 0.07s and 0.02s for video and VOIP respectively, when the cell was loaded with 60 users.

Also the delay for I-MLWDF showed higher delay of 98% as compared to MLWDF. Fairness index for

video flow for I-MLWDF is 20% better than MLWDF. Both algorithms showed performance level

between 98.5% and 99.5% for fairness index for VOIP flow with I-MLWDF showing improved

performance of about 0.2% increase better than M-LWDF. However, the highest fairness index was

presented by MLWDF and PF that reach a level above 99% whereas I-MLWDF showed the worst result

that reach a level of about 93%.

TABLE OF CONTENT

TITLE PAGE i

APPROVAL PAGE ii

CERTIFICATION iii

DECLARATION iv

10

DEDICATION v

ACKNOWLEDGEMENT vi

ABSTRACT vii

TABLE OF CONTENT viii

LIST OF FIGURES xi

LIST OF TABLES

MEANING OF ABREVIATIONS USED IN THE THESIS

xiii

xiv

CHAPTER ONE: INTRODUCTION

1.1 BACKGROUND OF THE STUDY 1

1.2 STATEMENT OF THE PROBLEM 3

1.3 AIMS/OBJECTIVES 4

1.4 SCOPE OF THE WORK 4

1.5 SIGNIFICANCE OF STUDY 5

1.6 METHODOLOGY 5

1.7 THESIS OUTLINE 6

CHAPTER TWO: LITERATURE REVIEW

2.1 INTRODUCTION 7

2.2 GENERAL CELLULAR CONCEPT 7

11

2.3 EVOLUTION OF LTE SYSTEM 8

2.3.1 FIRST-GENERATION MOBILE 1G 8

2.3.2 SECOND-GENERATION MOBILE 2G 8

2.3.3 THIRD-GENERATION MOBILE 3G 9

2.3.4 THE PATH TOWARD 4G 10

2.4 ADVANTAGES OF LTE TECHNOLOGY 12

2.5 MOTIVATION FOR 3GPP LTE RELEASE 8[5] 12

2.6 FEATURES OF LTE 13

2.7 SCHEDULING 16

2.7.1 DOWNLINK SCHEDULING 16

2.7.2 PROCEDURE OF DOWNLINK SCHEDULING 17

2.8 GENERAL CLASSIFICATION OF SCHEDULING 19

2.9 WHAT IS FUZZY LOGIC? 29

2.9.1 WHY USE FUZZY LOGIC? 29

2.10

2.11

FUZZY INFERENCE SYSTEM

RELATED WORK ON SCHEDULING IN LTE

30

35

CHAPTER THREE: SYSTEM MODELING

3.1 INTRODUCTION 48

3.2 PHYSICAL MODEL 48

12

3.3 PROPOSED SCHEDULING SCHEME 50

3.4 SIMULATION SCENARIO 56

3.5 SIMULATION PARAMETERS 56

3.5.1 SIMULATION TRAFFIC MODEL 56

3.6 MODEL VALIDATION 59

CHAPTER FOUR:SIMULATION AND RESULT ANALYSIS

4.1 PERFORMANCE METRICS 60

4.2 RESULT ANALYSIS 61

4.2.1 THROUGHPUT FOR VIDEO FLOW 62

4.2.2 THROUGHPUT FOR VOIP FLOWS 62

4.2.3 THROUGHPUT FOR NON-REAL TIME FLOWS 63

4.2.4 PACKET LOSS RATE FOR VIDEO FLOWS 64

4.2.5 PACKET LOSS RATE FOR VOIP FLOWS 65

4.2.6 PACKET LOSS RATE FOR NON-REAL TIME

FLOWS

66

4.2.7 DELAYS FOR VIDEO FLOWS 66

4.2.8 DELAYS FOR VOIP FLOWS 67

4.2.9 DELAYS FOR NON-REAL TIME FLOWS 68

4.2.10 FAIRNESS INDEX FOR VIDEO FLOWS 68

4.2.11 FAIRNESS INDEX FOR VOIP FLOWS 69

13

4.2.12

CHAPTER

5.1

5.2

FAIRNESS INDEX FOR NON-REAL TIME FLOWS

FIVE:CONCLUSION AND RECOMMENDATION

CONCLUSION

RECOMMENDATION

REFERENCES

APPENDIX

FUZZY INFERENCE PROCESSES TO DETERMINE

FLOW WEIGHT

70

71

71

73

79

79

14

LIST OF FIGURES

Figure 2.1 Evolution of wireless and wire line user 11

Figure 2.2 Peak data rate evolution of 3gpp technologies 12

Figure 2.3 Driving force for LTE development 13

Figure 2.4 General model of packet scheduling 19

Figure 2.5 Generalized model of packet scheduling 29

Figure 2.6 Block diagram of a fuzzy inference system 31

Figure 2.7 Schematic diagram of a fuzzy inference system 32

Figure 2.8 Fuzzy inference 34

Figure 3.1 Physical model of a downlink resource allocator 49

Figure 3.2 Physical model of a downlink Scheduler 50

Figure 3.3 fuzzy inputs 53

Figure 3.4 Applying membership function 54

Figure 3.5 Applying rules 54

Figure 3.6 Aggregating all outputs/defuzzification 55

15

Figure 3.7 Surface viewer 55

Figure 3.8 Model Validation with MLWDF Scheme 59

Figure 4.1 Average throughput per video flow 62

Figure 4.2 Average throughput per VOIP flow 63

Figure 4.3 Average throughput per NRT flow 64

Figure 4.4 Packet loss Ratio for Video flows 65

Figure 4.5 Packet loss ratio for VOIP flows 65

Figure 4.6 Packet loss ratio for NRT flows 66

Figure 4.7 Delay for Video flows 67

Figure 4.8 Delay for VOIP flows 67

Figure 4.9 Delay for NRT flows 68

Figure 4.10 Fairness index for Video flows 69

Figure 4.11 Fairness index for VOIP flows69

Figure 4.12 Fairness index for NRT flows 70

17

LIST OF TABLES

Table 3.1 Simulation parameters

58

18

MEANING OF THE ABBREVIATIONS AS USED IN THE WORK.

QoS quality of service

MLWDF modified largest weight delay first

EXP/PF exponential/proportional fair

I-MLWDF improved MLWDF

HOL head of line delay

PF proportional fair

1G first generation

2G second generation

3G third generation

4G fourth generation

Gbps gigabit per second

MS mobile station

BS base station

MSC mobile switching center

3GPP third generation partnership project

AMPS advanced mobile phone system

NMT Nordic mobile telephone system

19

TACS total access communication system

CDMA code division multiple access

LTE long term evolution

GSM global system for mobile communication

TDMA time division multiple access

AMC adaptive modulation and coding

ARQ automatic repeated request

BER bit error rate

CDMA code division multiple access

CP cyclic prefix

CQI channel quality indicator

EDGE enhanced data rates for global evolution

eNodeBEnhanced NodeB

EPC Evolved packet core

EUTRAN Evolved UMTS terrestrial radio access network

GPRS general packet radio services

HSCSD high-speed circuit-switched data

HSDPA high-speed downlink data packet access

MIMO multiple input multiple output

UE user equipment

VOIP voice over IP

TT1 transmission time interval

PLR packet loss rate

CBR constant bit rate

RT real-time

20

NRT non-real time

GUI graphic user interphase

HOL head of line delay

TCP transmission control protocol

UDP user datagram protocol

AMC adaptive modulation and coding

21

CHAPTER ONE

INTRODUCTION

1.1BACKGROUND OF THE STUDY Cellular communication was designed to ameliorate the flaws of traditional wireless and wired

communication networks providing communication between moving units and/or stationary

units and vice versa[1]. The moving units are referred to as mobile stations while the stationary

units are referred to as land fixed units. Some of the flaws encountered in traditional wireless

communication networks range from their inability to cope with the travelling speed of fast

mobile units; to its low capacity and poor data rate.

Cellular communication divides its coverage region into small regions called cells; hence, the

terminology “cellular”. This concept of cellular communication began to appear in Bell System

proposals during the late 1970s [1, 2]. Cell size is dynamic and is dependent on the population

(traffic) in the region to be covered. However, it is set by the transmitter power and the

frequency of operation of the cell. To manage scarce frequency resources and meet the

requirement of users, cells apply techniques like frequency hopping, frequency re-use and cell

splitting [3].

However, technological advancement and modernization led to an increase in the speed and

capacity of communication gadgets and increase in the growth and demands of cellular networks

respectively[1]. Furthermore, with the current trend of an integrated network, the limits of

cellular communications were expanded and subsequently, cellular communication began to

compete favourably with conventional wideband and broadband systems. Cellular network can

be viewed as a favorable interface between the user and the integrated network [3]. These

22

developments in cellular communication are better appreciated when a study of its evolutional

stages is undertaken.

The 0G may be referred to as the period from the crude communication stage to the point where

cellular communication is discovered[1]. Some researchers do not consider this as a generation

in the evolution of cellular communication. Others believe that nature enabled man to

communicate in small regions; as such one can only speak to another who is close. These small

regions may be referred to as “cells”. A cell is an area or region covered by a fixed location

transceiver known as base station[4]. The farther one goes, the more difficult it is to

communicate [5]. Also considering that the natural medium for this communication is wireless

(low frequency and power electromagnetic wave). This era seen as the origin of cellular

communication may therefore be referred to as 0G.The first generation of analogue mobile

telephony was followed by the second digital generation. Then, the third generation was

envisaged to enable full multimedia data transmission as well as voice communications. In

parallel to these activities related to the evolution of current wireless technologies, there is also

an increased research effort on future access referred to as fourth-generation (4G) radio access.

Such future radio access is anticipated to take the performance and service provisioning of

wireless systems a step further, providing data rates up to 100Mbps with wide-area coverage and

up to 1 Gbps with local-area coverage[6].

Resource allocation in LTE is described as the sharing of frequency,time,antenna ports and

power between users which are aimed at achieving spectral efficiency, fairness and high/standard

Quality of Service (QOS) [6]. Physical Resource Blocks (PRBs) are seen as the combined

sharing of frequency and time to users (UEs) in the LTE network in order to meet a reasonable

compromise amongst QOS,fairness, and spectral efficiency. Radio resources in LTE are

composed of PRBs, Modulation and coding scheme (MCS) and power allocation. The MCS

23

determines the bit rate and thus, the capacity of PRBs. Radio resource allocation is valid for one

Transmission Time Interval (TTI) which is equal to 1ms [7].

Fundamentally, LTE experiences great competition for resources considering the multiple but

limited channels it supportsin the presence of large number of users and their varied QoS

requirements; hence the need for optimum resource allocation strategy. It is established that

differences among resource allocation techniques are mainly based on trade-off between decision

optimality and computational complexity. Key design aspects range from complexity and

scalability, spectral efficiency, fairness to QoS provisioning[8]. The parameters for the

evaluation of this optimality vary for different research goal.Despite these variations, users must

be scheduled first before the assignment of resources [8].As such, several schemes for

scheduling users were developed considering the channel status of the user, the QoS demand by

the user and other scheduling parameters.

This work will focus attention on resource allocation in downlink system for real time services in

LTE networks. An improved scheme will be developed by modifying one of the earlier

algorithms known as MLWDF. The performance of the new algorithm will be compared with

PF, MLWDF, EXP/PF algorithms using four performance metrics namely: throughput, packet

loss rate, fairness index and delay. By incorporating bandwidth of flows (β) to MLWDF scheme,

the MLWDF scheme was modified.

Practically non-real time application has no constraint in delay and therefore has more packet in

the buffer of eNodeB.This is also possible because it does not require the presence of the called

party for communication to exist. Whereas it is not the same for real time application which has

constraint in delay and has tendency of missing their deadline and therefore lost. This led to the

considerable priority non-real time services get in accessing resources as can be seen from the

original algorithm. The new algorithm considered bandwidth of flow which was incorporated to

the old algorithm. By this, the scheme considered the bandwidth of the flow from the users to the

buffer in the eNodeB for the different users. The aim is to improve the chances of real time

services to access resources since they may exceed their deadline before the next transmission

time interval (TTI).

24

1.2 STATEMENT OF PROBLEM Opportunistic Algorithm such as PF, M-LWDF, EXP-RULE and EXP/PF are mostly used in 3G

technologies for resource allocation[9]. QoS delivery in LTE is mainly focused on real time

services, therefore it is clear that scheduler must target packet delays in order to have a standard

QoS for both EXP-RULE and M-LWDF schemes [9]. However, the HOL and packet delay

metrics are not enough to focus the scheduling priority on real time services[9]. The thesis

considered the bandwidth of flow (�) which is directly proportional to flow weight and users

reserved rate. Resources are allocated to flows depending on its bandwidth which could be real

time flow or non-real time flow. Non real time services has larger flows in the network and as

such has a higher flow bandwidth. The introduced parameter willfavor real time services during

the computation. This is because the larger the flow bandwidth, the lower the scheduling metric.

The Flowweight is determined by taking latency for real time services and throughput for non-

real time services determined by the uncertainty principles of fuzzy logic. These, together with

packet loss rate and fairness index are important constraints to take into account when

performing the scheduling task. By incorporating the above parameters, the aim is to modify the

opportunistic scheduler, M-LWDF in order to balance the scheduling decision to real time

services.

1.3 AIMS/OBJECTIVES This work seeks to achieve the following objectives

� Support real time and non-real time services in a scheduler

� Development of Scheduling scheme that would maintain trade-off between Quality of

Service and Spectral efficiency

� Focus on balancing scheduling decisions to real time services.

25

� Compare the performance of the proposed algorithm with some earlier

schedulingalgorithms(M-LWDF,PF,EXP/PF)

1.4 SCOPE OF THE WORK This dissertation has a limited scope. LTE as a 4G network was adopted. It was specific to

modelling of the LTE downlink scheduler and implementation of scheduling schemes. Analysis

was based on the computer simulation model of the network using Mat lab, LTE simulator as a

modelling tool and excel to plot graph.

1.5 SIGNIFICANCE OF STUDY This research will be significant to mobile wireless communication providers and researchers

especially with the current trend in total migration to 4G networks. In Nigeria, the benefits of this

work cannot be over-emphasized as the country continues to lead other West Africa countries in

the implementation of the 4G network with its large population distribution. With the

development of a near practical resource allocation scheme as proposed in this work, resource

management for large population would greatly improve. This will lead to efficient utilization of

scarce radio resources in terms of fairness and spectral efficiency; which will subsequently

minimize the cost of acquiring and running of the network.

1.6 METHODOLOGY To realize the objectives of this work, the following methodology was adopted:

� Review of LTE downlink transmission techniques and resource allocation schemes.

� Propose a scheme by modifying one of the earlier popular scheme for downlink

scheduling

� Develop computer models of scheduling scenarios of the proposed schemes.

26

� Validate model with performance of existing schemes.

� Simulate the models and obtain data.

� Analyze data in terms of the performance metrics.

� Compare the performance of the proposed scheme and their corresponding existing

scheme

1.7 THESIS OUTLINE This thesis report is organized as follows. In chapter one, the thesis work was introduced. In

chaptertwo, review of literatures was carried out on thehistory/evolution of the LTE network,the

LTE technology including its features. Chapter three presented the architecture, models and

simulation of an LTE scheduler. InChapterfour, simulation results and analysis were presented.

Chapter Five presented conclusions and relevant recommendations. Finally, the references

concluded this dissertation.

27

CHAPTER TWO

LITERATURE REVIEW

2.1 INTRODUCTION The essence of this chapter is to give a theoretical background of LTE Technology, fuzzy logic

and scheduling in LTE. The second aspect is the review of some related works in LTE

scheduling.

28

2.2GENERAL CELLULAR CONCEPT The essential of all cellular networks is that the final link between the subscriber and the fixed

network is by radio[10]. This has a number of consequences:

� Radio spectrum is a finite resource and the amount of spectrum available for mobile

communications is strictly limited;

� The radio environment is subject to multipath propagation, fading and interference

therefore it is not an ideal transmission medium;

� The subscriber is able to move and this movement must be accommodated by the

communications system.

In a cellular system, thehandsets carried by the users are called Mobile Stations(MS)[11]. The

MS communicates to the base stations (BS) through a pair of frequency channels, one for up-link

and another for downlink.All the base station, BS of a cellular system is controlled by a central

switching station called mobile switching center(MSC). The MSC is responsible for all kinds of

network management functionssuch as channel allocations, handoffs,billing, and power control

etc. The MSC is connected to other networks via gateways to allow the MS talk to other

networks.

Cellular technologies are classified into three generation;first. Second and third generation

respectively 1G, 2G, and 3G.

2.3EVOLUTION OF LTE SYSTEM The development of mobile communications has traditionally been viewed as a sequence of

successive generations. Two partnership organizations wereborn from the ITU-IMT-2000

initiative: The Third Generation Partnership project and the Third Generation Partnership Project

2. The 3GPP and 3GPP2 developed their own version of 2G, 3G, and even beyond mobile

29

systems. Here, a summary of all mobile generationsdeveloped by these two organizations as a

path to the evolution of LTE system.is presented[6].

2.3.1 FIRST-GENERATION MOBILE 1G

First-generation cellular networks (1G) were analog-based and limited to voice servicesand

capabilities only. 1G technology was vastly inferior to today’s technology.

In the late 1970s and early 1980s, various 1G cellular mobile communication systemswere

introduced; the first of such system, the Advanced Mobile Phone System(AMPS), was

introduced in the USA in the late 1970s[4]. Other 1G systems includethe NordicMobile

Telephone System (NMT) and the Total Access CommunicationsSystem (TACS)[4].While these

systems offer reasonably good voice quality, theyprovide limited spectral efficiency. This is why

the evolution toward 2G was necessaryto overcome the drawback of such technology.

2.3.2 SECOND-GENERATION MOBILE 2G The second-generation (2G) digital systems promised higher capacity and bettervoice quality

than did their analog counterparts. The two widely deployed secondgeneration(2G) cellular

systems are GSM (Global System for Mobile Communications)and CDMA (Code Division

Multiple Access) which was originally known asthe American interim standard 95, or IS-95, and

is now called CDMA One[12]. Boththe GSM and CDMA camps formed their own separate 3G

partnership projects(3GPP and 3GPP2, respectively) to develop IMT-2000-compliant standards

basedon the CDMA technology[13]. GSM differs from 1G by using digital cellular technology

and Time DivisionMultiple Access (TDMA) transmission methods and slow frequency hopping

forthe voice communication. In the USA, 2G cellular standardization process utilizeddirect-

sequence CDMA with phase shift-keyed modulation and coding.There was an evolution of main

air interface-related enhancements to GSMwhich are:

30

� higher data rates for circuit-switched services through aggregationof several time slots

per TDMA frame with High-Speed Circuit-Switched Data(HSCSD);

� General Packet Radio Service (GPRS) which had an efficient supportof non-real-time

packet data traffic. GPRS reached peak data rates up to 140 Kbpswhen a user aggregates all time

slots; and

� Enhanced Data Rates for Global Evolution(EDGE) has increased data rates up to 384

Kbps with high-level modulationand coding within the existing carrier bandwidth of 200 kHz[6].

2.3.3 THIRD-GENERATION MOBILE 3G Further evolution of the GSM-based systems is handled under 3GPP to define aglobal third-

generation Universal Mobile Telecommunications System (UMTS).The main components of this

system are the UMTS Terrestrial Radio AccessNetwork (UTRAN) based onWideband Code

Division Multiple Access (WCDMA)radio technology since it is using 5 MHz bandwidth and

GSM/EDGE Radio AccessNetwork (GERAN) based on (GSM)-enhanced data rates[14].On the

other hand, 3GPP2 implemented CDMA2000 under 1.25 MHz bandwidthwhich increased voice

and data services and supported a multitude of enhancedbroadband data applications, such as

broadband Internet access and multimediadownloads. This technology also doubled user

capacity over CDMAOne, and withthe advent of 1xRTT, packet data was available for the first

time[15].

As an evolution for CDMA2000, the 3GPP2 first introduced the High-RatePacket Data (HRPD)

which was referred to as CDMA20001xEV-DO. This standardenables high-speed, packet-

switched techniques designed for high-speed data transmissions,enabling peak data rates beyond

2 Mbps. 1xEV-DO expanded the types ofservices and applications available to end users,

enabling carriers to broadcast moremedia-rich content[16].

31

The 3GPP followed a similar direction and introduced an enhancement to theWCDMA system

providing High-Speed Downlink Packet Access (HSDPA) thatbrought spectral efficiency for

higher speed data services in 2001. Then anotherHigh-Speed Uplink Packet Access (HSUPA)

was introduced in 2005. The combinationof HSDPA and HSUPA is called HSPA[17]. The last

evolution of HSPA is the HSPA+ which was specified resulting fromadding Multiple

Input/Multiple Output (MIMO) antenna capability and 16 QAM(uplink)/64 QAM (downlink)

modulation. Coupled with improvements in the radioaccess network for continuous packet

connectivity, HSPA+ will allow uplink speedsof 11 Mbps and downlink speeds of 42 Mbps.

As the successor of CDMA2000 1xEV-DO, the CDMA2000 1xEV-DO Release 0provides peak

speeds of up to 2.4 Mbps with an average user throughput of between400 and 700 Kbps. The

average uplink data rate is between 60 and 80 Kbps. Release0 makes use of existing Internet

protocols, enabling it to support IP-based connectivityand software applications. In addition,

Release 0 allows users to expandtheir mobile experience by enjoying broadband Internet access,

music and videodownloads, gaming, and television broadcasts.

A revision of CDMA2000 1xEV-DO Release 0 is CDMA2000 Revision A (Rev-A) which is an

evolution of CDMA2000 1xEV-DO Rel-0 to increase peak rates onreverse and forward links to

support a wide variety of symmetric, delay-sensitive,real-time, and concurrent voice and

broadband data applications. It also incorporatesOFDM technology to enable multicasting (one-

to-many) for multimedia contentdelivery. As the successor of Rev-A, CDMA2000 1xEV-DO

Revision B (Rev-B)introduces dynamic bandwidth allocation to provide higher performance by

aggregatingmultiple 1.25MHz Rev-A channels[18].

2.3.4 THE PATH TOWARD 4G

4G mobile broadband technologies will allow wireless carriers to take advantageof greater

download and upload speeds to increase the amount and types of contentmade available through

32

mobile devices. 4G networks are comprehensive IPsolutions that deliver voice, data, and

multimedia content to mobile users anytimeand almost anywhere. They offer greatly improved

data rates over previousgenerations of wireless technology. Faster wireless broadband

connections enablewireless carriers to support higher level data services, including business

applications,streamed audio and video, video messaging, video telephony, mobile TV,

andgaming[19].

As a step toward 4G mobile broadband wireless, the 3GPP body began its initialinvestigation of

the Long-Term Evolution (LTE) standard as a viable technology in2004[20]. The LTE

technology is expected to offer a number of distinct advantagesover other wirelesstechnologies.

+

Figure 2.1: Evolution of wireless and wire line

33

2.4 ADVANTAES OF LTE TECHNOLOGY. These advantages include increased performanceattributes, such as:

� High spectral efficiency.

� Very low latency.

� Support of variable bandwidth.

� Simple protocol architecture.

� Compatibility and interworking with earlier 3GPP releases.

� Interworking with other systems, e.g., cdma2000.

� FDD and TDD within a single radio access technology.

� Efficient multicast/broadcast [6, 9].

2.5MOTIVATION FOR 3GPP LTE RELEASE 8[6] � Need to ensure the continuity of competitiveness of the 3G system for the future;

� User demand for higher data rates and quality of service;

� Packet Switch optimized system;

� Continued demand for cost reduction;

� Low complexity;

Figure 2.2: Peak data rate evolution of 3GPP technologies[30]

34

� Avoid unnecessary fragmentation of technologies for paired and unpaired bandoperation.

2.6FEATURES OF LTE LTE is a mobile broadband solution that offers a rich set of features with a lot of Flexibility in

terms of deployment options and potential service offerings[6]. Some ofthe most important

features are as listed below:

� OFDM for high spectral efficiency is the basis of the physical layer

OFDMis used in downlink in order to obtain a robustness against multipath interferenceand high

affinity to advanced techniques such as frequency domainchannel-dependent scheduling and

MIMO, while Single-Carrier FrequencyDivision Multiple Access (SC-FDMA) is used in uplink

in order to get alow Peak-to-Average Power Ratio (PAPR), user orthogonally in

frequencydomain, and multi-antenna application[21].

� Support for TDD and FDD

LTE supports both Time Division Duplexing(TDD) and Frequency Division Duplexing[22].

TDD is favored by a majority ofimplementations because of its advantages:

� Flexibility in choosing uplink to-downlink data rate ratios,

� Ability to exploit channel reciprocity,

Figure 2.3: Driving force for LTE development[30]

35

� Ability to implement in non-paired spectrum, and

� Less complex transceiverdesign.

� Adaptive Modulation and Coding (AMC)

LTE supports a number of modulationand Forward Error Correction (FEC) coding schemes and

allows thescheme to be changed on a per user and per frame basis, based on

channelconditions[23]. AMC is an effective mechanism to maximize throughput in atime-

varying channel. The adaptation algorithm typically calls for the useof the highest modulation

and coding scheme that can be supported by thesignal-to-noise and interference ratio at the

receiver such that each user isprovided with the highest possible data rate that can be supported

in theirrespective links.

� Support of variable bandwidth

E-UTRA shall operate in spectrum allocationsof different sizes, including 1.25, 1.6, 2.5, 5, 10,

15, and 20 MHz in boththe uplink and downlink[24]. Operation in paired and unpaired

spectrumshall be supported. This scaling may be done dynamically to supportuser roaming

across different networks that may have different bandwidthallocations.

� Very high peak data rates

LTE is capable of supporting very high peak datarates. In fact, the peak PHY data rate can be as

high as downlink peak datarate of 100 Mb/s within a 20 MHz downlink spectrum allocation (5

bps/Hz),while it provides uplink peak data rate of 50 Mb/s (2.5 bps/Hz) within a20MHz uplink

spectrum allocation.

� Mobility

36

E-UTRAN should be optimized for low mobile speed from 0 to 15km/h[25]. A higher mobile

speed between 15 and 120 km/h should be supportedwith high performance. Mobility across the

cellular network shall be maintainedat speeds from 120 to 350 km/h (or even up to 500 km/h

dependingon the frequency band).

� Link layer retransmissions

LTE supports Automatic Retransmission Requests(ARQ) at the link layer. ARQ-enabled

connections require each transmittedpacket to be acknowledged by the receiver;

unacknowledged packetsare assumed to be lost and are retransmitted. LTE also optionally

supportshybrid-ARQ, which is an effective hybrid between FEC and ARQ[26].

� Simultaneous user support

LTE provides the ability to perform twodimensionalresource scheduling (in time and frequency),

allowing supportof multiple users in a time slot; in contrast, existing 3G technology

performsone-dimensional scheduling, which limits service to one user for each timeslot. This

capability of LTE results in amuch better always-on experience andalso enables the proliferation

of embedded wireless applications/systems[27].

� Security

LTE provides enhanced security through the implementation of UICCSubscriber Identity

Module (SIM) and the associated robust and non-invasivekey storage and symmetric key

authentication using 128-bit private keys[28].

37

LTE additionally incorporates strong mutual authentication, user identityconfidentiality, integrity

protection of all signaling messages between UEand Mobility Management Entity (MME), and

optional multi-level bearerdata encryption.

� Efficient worldwide roaming

Because LTE will be the unified 4G standard formost 3GPP and 3GPP2 carriers worldwide, LTE

devices will be fundamentallyeasier to set up for worldwide roaming. The caveat is that the

actualfrequency band used by different carriers will be different (thereby retainingthe need for

multiband devices)[29]. As a result, the Verizon wireless migrationpath to LTE will provide

greater opportunities for seamless internationalroaming and for global device economies of scale

as well [6, 30].

2.7Scheduling The scheduler resides in the eNodeB to dynamically allocate uplink and downlinkresources over

the uplink and downlink shard channel U-SCH and D-SCH,respectively. Uplink scheduling is

performed per SC-FDMA while downlink isperformed for OFDMA. The eNodeB calculates the

time-frequency resourcesgiven the traffic volume and the QoS requirements of each radio

bearer[31]. However,the resources are allocated per UE and not per radio bearer. The uplink and

downlink schedulers are invoked to allocate resources every

TTI. The minimum TTI duration is of one subframe length; that is, 1 ms. However,the LTE

specification allows adaptive downlink TTI duration where multiplesubframes can be

concatenated to produce a longer TTI duration. This concatenationreduces the overhead for

higher layers. The TTI length can be set dynamicallyby the eNodeB through defining the

modulation and coding scheme used and the size of the resource blocks. Otherwise, it can be set

semi-statically through higher layer signaling. Adaptive TTI length can be used to improve the

Hybrid Automatic Repeat Request (HARQ) performance or the support of lower data ratesand

38

quality of service. In the following two sections we summarize the operationof the downlink

scheduler and uplink scheduler.

2.7.1Downlink Scheduling The unicast downlink transmission is carried over the shared downlink channel (D-SCH) and the

operation takes place at the MAC layer of eNodeB. At each TTI[6],the eNodeB has to

dynamically decide which UE is supposed to transmit, and when and using which frequency

resources. The decision depends on different factors including the cell’s nominal capacity, QoS

parameters (BER, minimum and maximum data rate and delay), and backlogged traffic waiting

for retransmission, link channel quality relayed to the eNodeB as a CQI, buffers sizes, and the

UE’scapabilities. More than one UE can be scheduled during one TTI. However, thenumber of

UEs scheduled that can be scheduled during one TTI is limited bythe signaling overhead.

Allocations are signaled to UEs on the PDCCH, and aUE with enabled downlink reception

monitors the PDCCH every TTI.In addition to the dynamic allocation, LTE standard provides

the flexibilityto what is called persistent scheduling where the time-frequency resources canbe

implicitly reused in the consecutive TTIs according to a specific periodicity[31].

Persistent scheduling reduces the overhead scheduling for applications suchas VoIP.

Scheduler design is not specified in the standard and is left for vendor implementation. An

efficient scheduler, however, should take into account the channelquality of the link from the

eNodeB to the UE and the buffer length of theradio bearers. It should also cater to fairness

among the UEs based on their servicelevel agreement (SLA), that is, subscription type and

priority level. A UEmonitors a shared reference signal broadcast to all UEs in the cell by

eNodeBto estimate the instantaneous downlink channel quality and signal it in a CQIreport. CQI

can be about either a single or multiple resource blocks, and can beeither periodic or aperiodic.

The periodic CQI report is transmitted together withuplink data on the PUSCH or on the

39

PUCCH, while the aperiodic CQI is scheduledby the eNodeB via the PDCCH and transmitted

together with uplink dataon PUSCH.

2.7.2 PROCEDURE OF DOWNLINK SCHEDULING The per-RB metrics’ comparison that serves as thetransmission priority of each user on a specific

RB is takeninto account for resource allocation for each UE. For examplethe k-th RB is allocated

to the j-th user if its metric mj;k is thelargest one among all i-UEs, i.e., if it satisfies the

equation:mj;k = maxi {m i;k} (1)The whole process of downlink scheduling can be dividedin a

sequence of operations that are repeated, in general, everyTTI as sown in figure 2.42

1) The Evolved Node B prepares the list of flows which canbe scheduled in the current TTI

.Flows could be formulatedonly if there are packets to send at MAC layer and UE atreceiving

end is not in the idle state.

2) Each UE decodes the reference signals, reports CQI(Channel Quality Indicator) to eNB which

helps to estimatethe downlink channel quality. The eNB can configure if theCQI report would

correspond to the whole downlinkbandwidth or a part of it which is called sub-band.

3) Then the chosen metric is computed for each flowaccording to the scheduling strategy using

the CQIinformation. The sub-channel is assigned to that UE thatpresents the highest metric.

4) For each scheduled flow, the eNB computes the amountof data that will be transmitted at the

MAC layer i.e. the sizeof transport block during the current TTI. The AMC(Adaptive

Modulation and Coding module) at MAC layerselects the best MCS (Modulation and Coding

Scheme) thatshould be used for the data transmission by scheduled users.Link adaptation

involves tailoring the modulation order(QPSK, 16-QAM, 64-QAM) and coding rate for each UE

inthe cell, depending on the downlink channel conditions.

5) Physical Downlink Control Channel (PDCCH) is used tosend the information about the users,

the assigned ResourceBlocks, and the selected MCS to terminals in the form of DCI

40

(Downlink Control Information).

6) Each UE reads the PDCCH payload .If a particular UEhas been scheduled; it will try to access

the proper PDSCHpayload.

The users are prioritized by packet scheduler on the basis ofa scheduling algorithm being used.

These algorithms whilemaking scheduling decisions, takes into account the

Instantaneous or average channel conditions, Head of Line(HOL) packet delays, status of

receiving buffer or type ofservice being used[32]

2.8General classification of Scheduling The scheduling strategies of wireless network can be broadly classified as shown in Figure. 2.45

Channel independent scheduling is based on the assumption that channel is time invariant and

error-free.The channel independent scheduling is first introduced in wired networks Examples of

channel independent scheduling are First-in-First-out (FIFO), Round Robin (RR), Weighted Fair

Queuing (WFQ), Earliest Deadline First (EDF), Largest Weighted Delay First (LWDF) etc. Here

some algorithms satisfy the QoS requirements and some simply schedules.With the help of CQI

reports which are periodicallysent by UEs to eNB, the scheduler can estimate thechannel quality

Figure 2.4: General Model of packet scheduler[8]

41

experienced by each UE. The schedulingperformed by these schedulers is called channel

sensitivescheduling. In this type of scheduling the scheduler maytry to maximize the QoS

requirements of each UE (QoSaware scheduling) or it may try to provide fairness amongUEs

(QoS unaware scheduling). Examples of channelsensitive scheduling are Maximum Throughput

(MT),Proportional Fairness (PF), Throughput To Average(TTA), Modified- Largest Weighted

Delay First (MLWDF),Exponential Proportional Fairness (EXP/PF),Exponential rule (EXP rule),

Logarithmic rule (LOG rule)etc. In LTE only channel sensitive scheduling is donebased on the

CQI reports from the UE.

A. Channel Independent Scheduling Strategies

Channel independent strategies were firstly introduced in wired networks and are based on the

assumption of time invariant and error-free transmission media[33]. Being unrealistic for LTE

networks, they are typically used in conjugation with channel-dependent strategies to improve

system performance.

� First in First out (FIFO): Though FIFO is the simplest of all possible scheduling

disciplines but it is inefficient and unfair. This scheduler serves the packets in the queue in order

of arrival and when the queue is full, it drops the packets that are just arriving[34]. The major

setback is that it cannot differentiate among connections; therefore all packets experience the

same delay, jitter and packet loss irrespective of which packet it is. The metric of i-th user on the

k-th RB can be translated from its behavior as shown in equation (2.1).

��,�� = � − ��2.1)

Where: t - the current time.

42

Ti- the time instant when request was issued by i-th user.

� Round Robin (RR): Round Robin allocates resources to each UE, completely neglecting

channel quality or data rate. Initially, the terminals are ordered randomly in a queue. The new

terminals are inserted at the end of the queue. The first terminal of this queue is assigned all the

available resources by scheduler, and then put it at the rear of the queue. The rest of steps follow

the same way, until no terminal applies for resources. RR metric is similar to the one defined for

FIFO. The only difference is that, in this case, Ti refers to the last serving time instant of the user.

On one hand, it seems to be a fair scheduling, since every terminal is given the same amount of

resources. On the other hand, it neglects the fact that certain terminals in bad channel conditions

need more resources to carry out the same rate, so it is absolutely unfair. This scheme is

impractical in LTE because different terminals have different service with different QoS

requirements [35].

� Weighted Fair Queuing (WFO): In Weighted Fair scheduling, the packets are grouped

into various queues [15]. A weight is assigned to each queue which determines the fraction of the

total bandwidth available to the queue. In this case, a specific weight (wi) is associated to the i-th

user (or class of users) and then it is used to modify Round-Robin metric as shown in equation

(3).

��,�� = ��,��2.2)

To assure that flows with larger packets are not allocated more bandwidth than flows with

smaller packets, it also supports variable-length packets. The Weighted Fair scheduling assigns

the bandwidth for each service based on the weight assigned to each queue and not based on the

number of packets.

43

� Blind Equal Throughput (BET): The Blind equal throughput is a channel unaware

strategy that aims at providing throughput fairness among all the users. To counteract the unfair

sharing of the channel capacity, the BET scheduler uses a priority metric which considers past

average user throughput as shown in equation (4)[36].

��,�� = 1�� − 1) �2.3)

Where

Ri(t- 1) - the average throughput of terminal iover windows in the past.

It is clear from equation (4) that the BET scheduler prioritizes users with lower average

throughput in the past. This implies that users with bad channel conditions are allocated more

resources compared to the users with good channel conditions. Thus throughput fairness is

achieved but at the cost of spectral efficiency.

� Largest Weighted Delay First (LWDF): To avoid packet drops, it is required that each

packet has to be received within a certain delay deadline in Guaranteed delay services[37]. It

incorporates the information about the specific packet timing, when the packet was created and

its deadline while calculating the priority metric. For Real-Time flow, its metric is calculated as

shown in equation (5)[37].

��,�� = ��,��2.4)

�� = − !"#�$�

�2.5)

WhereDHOL,i- waiting time of the packet at the head of the line

δi- drop probability

τi- target delay for the i-th user.

Similar to Round Robin, neglecting channel conditions leads to poor throughput in LWDF.

44

B. Channel Dependent/QOS unaware Scheduling Strategies

Channel-dependent scheduling strategies allocate resources with optimal algorithms by taking

into consideration the channel conditions. The user channel quality can be estimated from CQI

reports which help the scheduler to estimate the channel quality perceived by each UE and serves

as an indication of the data rate which can be supported by the downlink channel. Such schemes

are:

� Maximum Throughput (MT): Being a channel dependent scheduling, MT takes advantage

of multiuser diversity to carry out maximum system throughput. First, scheduler analyses CQI

reports from UEs to obtain data rate of an identical sub-channel for different terminals. This

information can be used in the priority metric to prioritize users with good channel conditions

over users with bad channel conditions. Thus scheduler assigns the resource to the user which

can achieve the highest throughput in this sub-channel based on SNR. The priority metric for the

MT scheduler is given as follows in equation (6) [36]:

��,�'� = (�� )�2.6)

Wheredik(t ) - the expected data-rate for i-th user at time t on the k-th Resource-block. It can be

calculated byconsidering the Shannon expression for the channel capacityas:

(�� ) = log -1 + /01�� )2 �2.7)

MT performs unfair resource sharing of the resources since it aims at maximization of

throughput only.

� Proportional Fair (PF): The Proportional Fair algorithm can improve the fairness among

users without losing the efficiency in terms of average (or aggregate) throughput. The terminals

are ranked according to the priority function which is defined as the ratio of the instantaneous to

45

average throughput. Then scheduler assigns resources to terminal with highest priority. Repeat

the last two steps until all the resources are used up or all the resources requirements of terminals

are satisfied [38, 39, 40, 41]. The PF was designed specifically for the Non-Real Time class and

hence does not assure any QoS requirement such as delay, jitter and latency. The preference

metric or priority function is obtained by merging the metrics of MT and BET and is given in

equation (7)[41].

��,�4� = (�� )��) �2.8)

Where, dik(t) - the throughput achievable by the user at time, t, for a particular channel, k

Ri(t) - the average throughput of the user, i, in previous transmissions.

In PF, users with good channel quality are assigned more resources while users with poor

channel quality are assigned less resources. PF is a combination of channel aware Maximum

Throughput (MT) metric and the channel unawareBlind Equal Throughput (BET) [40]. A

generalized PF approach by the addition of two novel parameters is given in equation (3).

��,�64� = 7(�� )89

[��)]< �2.9)

Wherea and b were used to modify the impact on the allocation policy of the instantaneous data

rate and the past achieved throughput respectively.

� Throughput to Average: Throughput to Average (TTA) scheduling algorithm tries to

divide the available resources between all users with the priority metric shown in equation

(12)[38].

��,��> = (�� )(��) �2.10)

The above metric performs averaging of resources evenly between the users. Here, the

achievable throughput in the current TTI is used as normalization factor of the achievable

46

throughput on the considered k-th RB. It is evident from its metric that the higher the overall

expected throughput of a user is, the lower will be its metric on a single Resource Block.

C. Channel Dependent/QOS aware Scheduling Strategies

In LTE, QoS differentiation is managed by associating aset of QoS parameters to each flow.

Minimum requiredperformance can be guaranteed by the scheduler if it knowsthe values of QOS

parameters, either in terms of guaranteeddata rates or of delivery delays.In this subsection, a

comprehensive overview on QoSawaresolutions presented in literature for LTE systems

ispresented.

� Schedulers for Guaranteed Data Rate: Several schedulers for guaranteed data rate have

been proposed in literature. These schedulers are QoS oriented time and frequency domain

schedulers that focus on GBR considerations. The proposedTime Domain Priority Set Scheduler

(TDPSS) has been devised to select users with the highest priority. Users areseparated into two

sets. Set 1’s users with bit rate belowtarget bit rate are managed by using BET and prioritized

overall the other users which form Set 2[33]. Furthermore, within eachset; prioritization is

according to priority metrics. While TDPSSwill tend to maintain the throughput of low signal

qualityusers to target bit rate; Frequency Domain- PF i.e PFscheduled (PFsch) will tend to

reduce their allocation share inthe frequency domain with priority metric given in equation

(10)[33].

��,�4�@AB = (�� )�@AB� �� − 1) �2.11)

WhereRisch(t - 1) is similar to the past average throughputdefined in equation (4), with the

difference that it is updated onlywhen thei-th user is actually served.

47

A Dynamic Hybrid Scheduler (DHS) composed bytwo basic components, corresponding to a

guaranteed and adynamic delay based rate allocation policy respectively ispresented [42]. Used

priorities are calculated, for thei-th user as shown in equation (11).

C� = ��,�$�

�2.12)

It is important to note that the transmission of the head of line packet becomes more urgent,

when the value of Piis increased. To attain the guaranteed bit-rate, the resources are allocated to

the user with the highest priority. The user with second highest priority is considered thereafter

for allocation in case the RBs are left free and so on.

In order to simplify the LTE MAC scheduling, two stages have been defined: Time Domain

(TD) and Frequency Domain (FD) schedulers [43]. The TDPS differentiates the users according

to their QoS characteristic whereas FDPS assigns the RBs between the priority users. Based on

the QoS Class Identifier (QCI), the incoming packets are categorized upon their priority order.

The priority sets are classified as GBR and non- GBR set. After this step, the FDPS orderly

assigns the best RB to each user in the GBR set, updating the achieved bitrate. When all users in

the list have reached their target bit-rate, if RBs are still available, the scheduler assigns them to

users in the non-GBR list using PF metric. Thus, all these approaches use ordered lists to

prioritize the most delayed flows and to achieve their target bit-rate. This approach is unfair to

non-GBR services.

2) Schedulers for Guaranteed Delay Requirements: Real- Time flows have more strict delay

restraint than Non-Real- Time flows resulting in the reduction of influence of error correction.

Scheduling strategies that aim to guarantee bounded delay fall in the category of the QoS-aware

48

schemes. Herein, QoS aware algorithms present in the literature that makes use of per-RB

metrics are described.

The Modified Largest Weighted Delay First (M-LWDF) combines both channel conditions and

the state of the queue with respect to delay in making scheduling decisions [44]. It ensures that

the probability of delay packets does not exceed the discarded bound below the maximum

allowable packet loss ratio i.e.

CDE��,� > $�G ≤ #� �2.13) The scheduler allocates resources to the user with the maximum priority index which is made up

of the product of the HOL packet delay of the user, the channel capacity with respect to flow and

the QoS differentiating factor. The analytical expression of the MLWDF scheduler is shown in

equation (12) [15].

��,�'�� = ��,�(�� )��) �2.14)

EXP/PF takes into account the characteristics of PF and an exponential function of the end-to-

end delay of the packet to be transmitted. EXP/PF distinguishes between real time and best effort

flows. For best effort flows, EXP/PF becomes PF while for real time flows. EXP/PF is given in

equation (13) [45].

��,��I4/4� = KLM N��O�,� − P

1 + √P R . (�,S��)��) �2.15)

Where

P = 11TU

V ��O�,�

WXY

�Z[ � 2.16)

Nrt is the number of active real time downlink flows

49

The exponential term is closer to 1 if HOL delays of all users are about the same. Thusabove

formula becomes Proportional Fair. If one of the user’s delays becomes large, the exponential

term in will override the left term in equation (13) and dominate the selection of a user.

It suffix to mention that in the implementation of MLWDF and the EXP/PF schemes, they are

primarily used as non-real time services schemes, while the PF is implemented for real time

services. However, a comparison of both schemes shows that MLWDF performs optimally at

low load scenario while EXP/PF performs optimally at high load scenario [33].

D. Dynamic and Semi-persistent Scheduling for VoIP support

Dynamic packet scheduling for VoIP traffic in the LTE Downlink is presented in literature [50,

51]. Optimizing the performance of dynamic scheduling considering a mix of VoIP traffic and

best effort flows is the major motive of this proposal. The proposed algorithm is divided into

time domain and frequency domain packet schedulers. At every TTI,scheduler called as

Required Activity Detection with Delay Sensitivity (RAD-DS) prioritizes each schedulable user

according to the time domain metric MTD[n,t],which is combination of 3 metrics given in

equation (14).

\��[], �] = �[], �]. �^UT9_[], �]. �/UT9_[], �]�2.17)

RAtraf[n, t] (i.e. the required activity) implies the time share required by user n where a user

should be scheduled. m[n, t] is a counter incremented every TTI that guarantees some fairness in

resource scheduling. Finally, DStraf[n, t] (i.e. the delay sensitivity) function imposes time

constraints to users with a delay bound that increases with HOL packet delay. The frequency

domain scheduler allocates Resource Blocks to different users using the Proportional Fair

scheduled (PFsch) metric.

50

To support high number of VoIP flows, semi-persistent allocation solutions (generally

considered as channel independent schemes) aim at increasing the VoIP capacity of the network

by maximizing the number of supported VoIP calls. One such scheme presented, improves the

VoIP capacity of the network with the use of semi-persistent scheme [46, 47]. Here, the radio

resources are divided in several groups of RBs. Each pre-configured block is associated only to

certain users. Furthermore, RB groups are associated to each user in contiguous TTIs. Resource

allocation of each RB group to the associated UEs is performed dynamically. The proposed

scheme reduces the control overhead with respect to the dynamic scheduling. Other semi-

persistent schemes for VOIP in LTE have also been proposed [48, 49, 50].

E. Energy Aware Solutions

Energy consumption is heavy in LTE due to tremendous processing load on UE. Energy

conserving solutions curb energy waste and hence extend the battery life of UE among which

Discontinuous reception (DRX) is useful. In DRX, when there are no data transmissions; UE

turn off its radio equipment UE to save energy. The light sleeping mode is introduced to further

improve the performance of DRX for QOS –aware traffic [7, 50]. The key idea is to turn off the

power amplifier .Other components in transceiver cut down their power consumption while

allowing fast wakeup. Proposed scheme reduces energy consumption while satisfying the delay

constraints.

The impact of different scheduling schemes from an energy efficiency point of view has been

analyzed[51]. It is demonstrates that the MT scheme is more energy efficient than both PF and

RR. In scenarios with low traffic load, Bandwidth Expansion Mode algorithm is used for

achieving energy savings for the eNB[33]. The eNB transmission power is reduced by assigning

51

a coding scheme with lower rate to each user. Consequently their spectrum occupation is

expanded.

2.9 What is fuzzy Logic? Fuzzy logic has two different meanings. In a narrow sense, fuzzy logic is a logical system, which

is an extension of multivalued logic. However, in a wider sense fuzzy logic (FL) is almost

synonymous with the theory of fuzzy sets, a theory which relates to classes of objects with

unsharp boundaries in which membership is a matter of degrees[52].

2.9.1 Why Use Fuzzy Logic? � Fuzzy logic is conceptually easy to understand.

The mathematical concepts behind fuzzy reasoning are very simple. Fuzzy logic is a more

intuitive approach without the far-reaching complexity.

Figure 2.5: Generalized model of packet scheduling[75]

52

� Fuzzy logic is flexible.

With any given system, it is easy to layer on more functionality without starting again from

scratch.

� Fuzzy logic is tolerant of imprecise data.

Everything is imprecise if you look closely enough, but more than that, most things are imprecise

even on careful inspection. Fuzzy reasoning builds this understanding into the process rather than

tacking it onto the end.

� Fuzzy logic can model nonlinear functions of arbitrary complexity.

You can create a fuzzy system to match any set of input-output data. This process is made

particularly easy by adaptive techniques like Adaptive Neuro-Fuzzy Inference Systems (ANFIS),

which are available in Fuzzy Logic Toolbox software.

� Fuzzy logic can be built on top of the experience of experts.

In direct contrast to neural networks, which take training data and generate opaque, impenetrable

models, fuzzy logic lets you rely on the experience of people who already understand your

system.

� Fuzzy logic can be blended with conventional control techniques.

Fuzzy systems don't necessarily replace conventional control methods. In many cases fuzzy

systems augment them and simplify their implementation.

� Fuzzy logic is based on natural language [52, 53].

The basis for fuzzy logic is the basis for human communication. This observation underpins

many of the other statements about fuzzy logic. Because fuzzy logic is built on the structures of

qualitative description used in everyday language, fuzzy logic is easy to use

53

2.10FUZZY INFERENCE SYSTEM A fuzzy inference system (FIS) essentially defines a nonlinear mapping of the input datavector

into a scalar output, using fuzzy rules[54]. The mapping process involves input/output

membershipfunctions, FL operators, fuzzy if–then rules, aggregation of output sets, and

defuzzification.An FIS with multiple outputs can be considered as a collection of independent

multiinput,single-output systems. A general model of a fuzzy inference system (FIS) is shown in

Figure 2.46. The FLS maps crisp inputs into crisp outputs. It can be seen from the figure that the

FIScontains four components: the fuzzifier, inference engine, rule base, and defuzzifier. The

rulebase contains linguistic rules that are provided by experts. It is also possible to extract rules

fromnumeric data. Once the rules have been established, the FIS can be viewed as a system that

mapsan input vector to an output vector. The fuzzifier maps input numbers into corresponding

fuzzymemberships. This is required in order to activate rules that are in terms of linguistic

variables.

The fuzzifier takes input values and determines the degree to which they belong to each of

thefuzzy sets via membership functions. The inference engine defines mapping from input

fuzzysets into output fuzzy sets. It determines the degree to which the antecedent is satisfied for

eachrule. If the antecedent of a given rule has more than one clause, fuzzy operators are applied

Figure 2.6: Block diagram of a fuzzy inference system

54

toobtain one number that represents the result of the antecedent for that rule. It is possible that

oneor more rules may fire at the same time. Outputs for all rules are then aggregated. During

aggregation,fuzzy sets that represent the output of each rule are combined into a single fuzzy

set.Fuzzy rules are fired in parallel, which is one of the important aspects of an FIS. In an FIS,

theorder in which rules are fired does not affect the output. The defuzzifier maps output fuzzy

sets into a crisp number. Given a fuzzy set that encompasses a range of output values, the

defuzzifierreturns one number, thereby moving from a fuzzy set to a crisp number. Several

methods fordefuzzification are used in practice, including the centroid, maximum, mean of

maxima, height,and modified height defuzzifier. The most popular defuzzification method is the

centroid, whichcalculates and returns the center of gravity of the aggregated fuzzy set. FISs

employ rules. However,unlike rules in conventional expert systems, a fuzzy rule localizes a

region of space alongthe function surface instead of isolating a point on the surface. For a given

input, more than onerule may fire. Also, in an FIS, multiple regions are combined in the output

space to produce acomposite region. A general schematic of an FIS is shown in Figure 2.26[54].

55

The inference process can be described completely in the five steps shown in Figure 2.27.

Step 1: Fuzzy Inputs

The first step is to take inputs and determine the degree to which they belong to each of

theappropriate fuzzy sets via membership functions.

Step 2: Apply Fuzzy Operators

Once the inputs have been fuzzified, we know the degree to which each part of the antecedent

has been satisfied for each rule. If a given rule has more than one part, the fuzzy logical

Operators are applied to evaluate the composite firing strength of the rule.

Step 3: Apply the Implication Method

The implication method is defined as the shaping of the output membership functions onthe basis

of the firing strength of the rule. The input for the implication process is a single numbergiven by

Figure 2.7: Schematic diagram of a fuzzy inference system[54]

56

the antecedent, and the output is a fuzzy set. Two commonly used methods ofimplication are the

minimum and the product.

Step 4: Aggregate all Outputs

Aggregation is a process whereby the outputs of each rule are unified. Aggregation occursonly

once for each output variable. The input to the aggregation process is the truncated outputfuzzy

sets returned by the implication process for each rule. The output of the aggregation processis the

combined output fuzzy set.

Step 5: Defuzzify

The input for the defuzzification process is a fuzzy set (the aggregated output fuzzy set),and the

output of the defuzzification process is a crisp value obtained by using

somedefuzzificationmethod such as the centroid, height, or maximum [55, 54].

57

Figure 2.8: fuzzy

58

2.21. RELATED WORKS ON SCHEDULING IN LTE Several researches over the years proposed their own algorithms on how to schedule users and

allocate resources considering the downlink part of an LTE network in an efficient way with an

objective of maintaining trade-off between system throughput and fairness among users. Herein

is a review of some of the algorithms proposed by different researchers?

In [45], the performance of exponential/proportional fair (EXP/PF) and maximum-largest

weighted delay first (M-LWDF) scheduling algorithm in the downlink 3gpp LTE system was

evaluated. They conducted performance evaluation in terms of system throughput, average real

time (RT) and non-real time (NRT) throughput, packet loss for RT service and fairness for NRT

service. Video streaming and web browsing traffic were used to model RT service and NRT

service respectively. Results of the evaluation of the algorithms in downlink showed that at

lower load, M-LWDF algorithm provides better performance than EXP/PF while as the load

increases the EXP/PF gives better performance. However it was assumed that the instantaneous

SINR reported by the users are error and delay free which is not always the case and can be

further researched upon.[44]Investigates the performance of well-known packet scheduling

algorithms developed for single carrier wireless systems from a real time video streaming

perspective. The performance evaluation is conducted using the downlink third generation

partnership project long term evolution (3GPP LTE) system as the simulation platform. This

paper contributes to the identification of a suitable packet scheduling algorithm for use in the

downlink 3GPP LTE system supporting video streaming services. Results show that in the

downlink 3GPP LTE system supporting video streaming services, maximum-largest weighted

delay first (M-LWDF) algorithm outperforms other packet scheduling algorithms by providing a

higher system throughput, supporting a higher number of users and guaranteeing fairness at a

59

satisfactory level.[56]Propose a novel and distributed algorithm in order to mitigate the inter-cell

interference and enhance the capacity of the users near the cell-edge area of a multi-cellular

network configuration. The main characteristics of the algorithm are the bandwidth allocation in

a user selective manner and the collision avoidance with sub channels in use by users in

neighboring cells. The algorithm requires coordination between base stations (BSs) in the sense

that the cell-edge bands in adjacent cells should be orthogonal and channel occupancy state of

each cell must be known to its neighboring cells. It was shown that the proposed algorithm has

better performance than existing conventional schemes. In [57],a QOS-guaranteed cross-layer

resource allocation algorithm for multiclass services in downlink LTE system is proposed, which

takes EXP rule, channel quality variance, real-time services and non-real-time services and

Minimum transmission rate into account. The key features of the proposed algorithm are that all

the users and resource blocks will be allocated step by step respectively. Numerical results

demonstrate that the proposed algorithm effectively guarantees the user QoS requirement for

multiclass services. At the same time, it mainly maintains the throughput and fairness

performances in a high level. In [32], an extensive survey on downlink packet allocation

strategies recently proposed for LTE networks, highlighting at the same time key issues that

should be considered when designing a new solution. It was also stated that a key feature of LTE

will be the adoption of advancedRadio Resource Management procedures in order to increase

thesystem performance up to the Shannon limit. In[58], a generalized approach of Proportional

Fair (GPF) scheduling and its application to OFDMA with frequency scheduling has been

presented. Based on defined fairness criteria for wireless communications, they carried out

comprehensive OFDMA system level simulations, which show fairness and system throughput

results for GPF scheduling with and without employing frequency scheduling. The obtained

60

results show, that on the one hand the trade-off between system throughput and (data-rate and

allocation) fairness can be tailored by the parameter settings for GPF scheduling. On the other

hand, the results prove that GPF frequency scheduling provides significant multiuser diversity

gain inFrequency domain. This generates gain in system throughput over GPF scheduling

employing time domain scheduling only. Furthermore, GPF frequency scheduling gives superior

short-term fairness with respect to the achieved data-rate and with respect to the allocated

resources, which makes it more suitable for delay critical data. In[59],Adaptive Fairness Control

for a Proportional FairLTE Scheduler was proposed. The parameterization of these scheduling

algorithms to achieve a certain desired fairness level is non-trivial. They show that the optimal

fairness parameter settings depend on the System State, such as the current cell load. Their main

contribution was a design of a self-optimizing scheduler architecture which includes a controller

element that dynamically adjusts the fairness parameters of the scheduler. They demonstrate that

with this design, an operator-defined reference fairness level is maintained in scenarios with

fluctuating load and thus cell throughput can be improved.[60]Explores the potential gain of joint

diversity in both frequency domain and time domain which can be exploited to achieve spectral

efficiency gains whilst simultaneously facilitating QOS/fairness in an OFDMA system

(particularly in 3GPP Long Term Evolution (LTE)). The performance of several joint time-

frequency schedulers is investigated. Simulation results show that joint time frequency

schedulers achieve significantly superior performance compared to a more conventional time

domain (only) proportional fair scheduler. The joint schemes show promising throughput gain

while meeting stringent fairness criteria. In this paper, a Two-Level Downlink Scheduling for

Real-Time Multimedia Services in LTE Networks was proposed. This work considered the

problem of packet scheduling for multimedia real-time flows in the downlink of LTE mobile

61

networks. A two-level algorithm has been designed by exploiting discrete time feedback control

theory. The properties of the proposed approach have been theoretically investigated to

demonstrate that it is suitable to provide both real-time and best-effort services. Finally, LTE-

SIM was used as the modeling tool. Numerical simulations were presented to confirm the

analytical results. The effectiveness of the proposed approach have been highlighted comparing

it with other well-known scheduling strategies and considering both the effect of the inter-cell

interference and the impact of the packet loss ratio on the QOE of real-time flows perceived by

end users.In[61], this paper, Multi-QoS-aware Fair Scheduling for LTE was studied. They

proposed a novel LTE downlink MAC scheduling algorithm. The proposed scheduler

differentiates between the different QoS classes and their requirements. Two different QoS

classifications are considered: Guaranteed Bit Rate (GBR) and non-Guaranteed Bit Rate (non-

GBR). The proposed scheduler also considers the different users channel conditions and tries to

create a balance between the QoS guarantees and the multi-user diversity in a proportional fair

manner. The simulation analysis confirms that guaranteeing the different QoS requirements is

possible.[43]Proposed a modified radio resource management-based scheduler with minimum

guarantee in the downlink following network capacity and service class attributes defined in LTE

standard. The complete service class-based scheduler design were divided into three discrete

parts which are admission control, resource allocation and packet scheduling. Complexity

reductions was achieved by following a divide and conquer strategy in which first the service

class are sorted for resource allocation and then users in a class are sorted for resource allocation

and the users in a class are sorted for packet scheduling. Discrete event simulator ‘LTE SIM’

with LTE specifications was used as the modeling tool. The results showed that the proposed

scheme performs better than the other schemes like M-LWDF, EXP-RULE etc. in terms of

62

system throughput, user mobility and fairness. To further this work, LTE QCI parameter should

be taken into account; Attributes like fairness should be tested for different traffic types to

achieve some hybrid method of allocation and scheduling. Moreover, interference from other cell

sites and throughput for different sectored cells will be explored in different frequency setups. In

[62], a review of radio resource allocation strategies were presented. Scheduling algorithms were

classified into two classes: channel independent scheduling and channel-dependent scheduling.

The formal includes first in first out (FIFO), Round Robin, Weighted fair Queuing, and Blind

equal throughput. whereas the later consist of Maximum throughput, proportional fair,

throughput to average, schedulers for guaranteed data rate, schedulers for guaranteed delay

requirements. The various key issues that should be considered when designing a new scheduling

scheme are extensively studied. Starting from channel independent strategies. Most recently

introduced QOS aware as well as energy aware solutions have also been studied.In

[63],enhancement of QOS in LTE downlink systems using frequency diversity selectivity

scheduling was proposed. In the proposed approach, structure throughput, QOS control, and

scheduling justice are jointly incorporated into a framework to energetically execute radio

resource-allocation for numerous users, and successfully prefer optimal system parameters such

as power and modulation rate to acclimatize to the unreliable channel quality of each resource

block. The simulation model was implemented in mat lab tool. Results showed that the proposed

algorithm increases spectral efficiency and also when compared with frequency diversity

selectivity scheduling showed significant performance improvement. In[64], a scheduling

mechanism is studied for downlink video traffic for different flows like video, voice and text

traffics with fixed user equipment in terms of average good put, average delay time, invalid

packet rate, spectral efficiency and packet loss ratio. The researcher undergo three phases for

63

designing RRM: initial scheduling for resource physical resource block, managing queue and

predication of packets for delays, cut in process, the result showed high performance with respect

to maintaining trade-off between throughput and fairness. In [65],this paper, a scheduler that

allocates resource block (SB) for different services in order to meet their QOS requirement was

proposed. The proposed scheduler allocates SB’s first to real time services by estimating no of

Scheduling blocs required and then allocates scheduling blocks to user according to their

priorities. Results show that new modified algorithm has better performance than existing

algorithm. [66]Discussed various scheduling algorithms and the QOS guaranteed resource block

allocation algorithm to overcome the constraints of MCS and to fulfill the requirement of Quality

of service (QOS). In this work, various algorithms such as Round Robin (RR), maximum CQI.

Proportional fair and priority scheduling algorithms were compared. The comparison showed

that priority algorithm provides better throughput. Further work can be done to increase the

throughput of the system and efficiently allocate resources. In[67], a fast algorithm was proposed

for resource allocation in wireless cellular network. An efficient optimization algorithm to

compute the optimal resource allocation in the downlink of an OFDM wireless cellular network.

It was showed that the algorithm converges to the optimal solution and has complexity of O (n)

for n users. The modeling tool was a time-varying channel model. Numerical results showed that

the algorithm converges fast in practice.[68]Proposes a novel downlink scheduling algorithm

which balances its performance between efficiency (in term of throughput) and fairness to users.

The algorithm makes use of assignment model for resource allocation to all the selected users

during each transmission time interval (TTI). The proposed algorithm is simulated using a Mat

lab based LTE link level simulator from the Vienna University that implement standard

compliant LTE downlink. The simulation results were compared with Round Robin, Best CQI,

64

Ms Algorithm and New-SCH algorithm technique. The results show that the new algorithm

balances well between throughput and fairness. In[69], this paper, resource allocation in LTE

OFDMA systems using genetic algorithms and semi-smart antenna was proposed. The semi-

smart antennas can produce flexible coverage patterns between BSs to minimize interferences for

mobile units. A system level simulations which contains 25Bss and 1000 mobile units was

developed. The GA learned based stations coverage patterns improves system total capacity and

maintains general coverage for all serviceable mobile users. The total capacity improvement is a

result of improving channel quality on mobile users in terms of maintaining intercell

interferences. Simulation results shows significant total capacity improvement compared to

traditional cells with static coverage patterns. In[70], a new downlink scheduling algorithm was

proposed. In this algorithm, it was assumed that each eNode B receives channel feedback

information in the form of CQI-feedback matrix. The matrix size equals to (No of user

equipment) multiplied by (no of resource blocks) in each TTI. The proposed method gives

preferences to those users which use less bandwidth than others. Also it evenly distributes the

resources among the users during each TTI, therefore the fairness is taken into consideration for

user by the proposed algorithm and at the same time user’s system capacity was increased. The

performance of the proposed algorithm is evaluated and compared with traditional scheduling

algorithms like PF, RR, and best CQI. Simulations result show that the proposed method could

provide fairness better than the three mentioned algorithms, also a better trade-off between

fairness and throughput was obtained.[71]Provides a comprehensive centralized RRM algorithm

for downlink OFDMA cellular fixed relay network in a way to ensure user fairness with minimal

impact on network throughput. The proposed algorithm exploits the opportunities in the

frequency, spatial and traffic diversities irrespective of the geographical relay deployment.

65

Simulation result prove the learning ability and the efficiency of the dynamic open routing

strategy which converges to better routes, even under the challenging uniform relay deployment

considered.[72]Proposed how to obtain an optimal resource allocation for wireless networks with

inter-cell interference. The optimal solution obtained by the proposed scheme can achieve

significant performance improvement for cell-edge users and desirable performance for cell-

center users compared with the reference schemes. The common theme of ICIC avoidance

schemes is to apply restrictions to the usage of downlink resources such as time/frequency and/or

transmit power resources. Such coordination of restrictions will provide an opportunity to limit

the interference generation in the area of the cellular network. The proposed resource allocation

scheme can yield balanced performance between cell-edge and cell-center users, which allows

for future wireless networks to deliver consistent high performance to any user from

anywhere.[73]Provides evaluation of the downlink (DL) transmission in LTE. The evaluation is

performed for a regular hexagonal multi-cell deployment, universal frequency reuse scheme 2*2

multiple input multiple output (MIMO) antenna configuration and varying packet scheduling

algorithm. In this paper, multiple performance scenario of SISO and (2*2) MIMO in downlink

LTE cellular network while varying bandwidth sizes and scheduling algorithms are evaluated

through a custom built system level simulator. It is found that combining MIMO techniques with

different bandwidth sizes and various scheduling algorithm will improve the capacity of the

system without adding additional infrastructure cost.[74]Investigated ways of achieving more

efficient resource allocation in wireless relay networks via spatial reuse, they first formulate a

joint spatial reuse and resource allocation problem as an integer linear programming (ILP)

model. They then considered a grouping mechanism in which relay stations are grouped together

if they do not interfere with each other’s transmission signal. A two-phase heuristic solution was

66

proposed because of the high computational complexity of the ILP model. In the first phase, the

enhanced-Resource Diminishing principle application is employed to determine the number of

resources required by the base station and a set of selected relay stations. In the second phase, a

max-coloring algorithm is employed to organize the selected relay stations into broadcast groups

and then assign the required resources to each group by exploring the maximum advantages of

spatial reuse. The results showed that the proposed solution improves system performance up to

61% as compared to the existing mechanisms. In[75], a study and performance evaluation of four

algorithms which are PF, M-LWDF, EXP-rule and LOG rule was put forth. The analysis was

carried out using random mobility model. This paper identifies the strength and weakness of

well-known algorithms in the downlink LTE system and the key aspects that should be taken

into account when designing a new algorithm. It can be inferred that PF algorithms is more

suitable for non-real time services because it does not account packet delays during its decision

making. On the other hand schedulers such as M-LWDF, EXP/PF, EXP-rule and LOG rule are

better choice for real time services. In[76], this paper, to alleviate the performance degradation

due to simultaneous multiple imperfect channel quality information (CQI), a simple and efficient

perfect scheduling is developed in downlink LTE system for real time traffic. A frequency

domain channel predictor based on kalman filter is first developed to restore the true CQI from

erroneous channel feedback. Then, a time domain grouping technique employing the joint of

proportional fair (PF) and modified Largest weight Delay first (M-LWDF) algorithms was used.

The simulation results showed that the proposed algorithms outperforms FD-PF and M-LWDF

algorithm throughout when CQI reports has delay and errors.[77]Paper proposes an efficient

algorithm that includes scheduling strategies and resource allocation mechanisms, to avoid the

latency or starvation of lower priority connections and to maintain system performance in

67

downlink of LTE. The PFPS algorithm was proposed to maintain the fairness of all the services

and avoid latency or starvation. The simulation result showed that PEPS has a higher throughput

than RR and PF, meanwhile, it has more fairness than max-rate. In[78], downlink resource

allocation for next generation wireless network using turbo code over non-linear channel was

proposed. This work target high data rate and efficient resource usage. The schemes consist of

radio resource and power allocation which are implemented separately. They proposed low

complexity heuristic algorithm to achieve the radio resource allocation, where graph base

framework and fine physical resource blocks assignment are performed to mitigate major ICI

and hence improve the network performance. A novel distributed power allocation is then

performed to optimize the performance of cell edge users under desirable conditions. The power

optimization is formulated as an interactive barrier constrained water filling problem and solved

by Lagrange method. Simulation results indicate that the proposed scheme can improve

performance of cell edge users in multicellularnetwork. In[79], a framework for scheduling and

resource allocation in LTE downlink using Nash bargaining theory was proposed. The context is

that of spectrum sharing with multiple users competing for the simultaneous access to the radio

channel. They first gave a layered system representation and then model it through a game

theoretic formulation using nash bargaining theory, where players compete to achieve a better

common payoff. A trade-off between fairness and throughput is identified and addressed, in

addition they also provide an efficient algorithm that drives the system through a pareto efficient

point by nash bargaining solution.ns3 simulator with an extension for LTE systems is used to

carry out tests. The results confirm the optimality of the solution and it’sadaptively to changes in

the scenario. However further development of this work will include the extension to the

multicellularcase, where the intra-cell scheme should be integrated with resource allocation

68

strategies among the base stations. In [80], scheduling algorithm based on deadlines to improve

performance and fairness in the distribution of radio resources was proposed. Four metrics were

used to assess performance, throughput, delay,packet loss rate and fairness index. The

performance of the proposed scheduler was carried out using LTE-SIM simulator. The

simulation results showed that their proposal presents a better performance with respect to packet

loss rates and delay, which was expected due to the use of deadlines. The proposed scheme also

provides better throughput for video flow. The gains made by the proposed scheduler showed no

negative impact on fairness in real time services. This work can however be enhanced by the use

of other indicators of QOS in the metrics scheduler and a comparison with other scheduler. Such

as exponential rule, logarithmic rule and exponential proportional fair. In[81], a scheme was

implemented that is optimized for offering improved quality of service for a diverse mix of

traffic including real-time VBR traffic in the downlink of LTE networks. This proposed scheme

aims to satisfy application quality of service requirements and improve overall system

throughput by exploiting multi-user diversity. This was achieved by minimizing the average

delay experienced by real-time packets in the network and scheduling users in the sub-channels

where they experience the best link quality, implying higher data rate. The popular discrete event

simulator network simulator was used to formulate the LTE system. The proposed VBR-

optimized scheme performs better than the well-known modified largest weighted first algorithm

and the delay prioritized scheme. In[82], this paper, they study the allocation problem from a

cooperative game theory perspective and present two algorithm to find the kalai-smorodinsky

bargaining solution to the problem. By adopting certain relaxation, they manage to convert the

bargaining theory based MIMO-OFDMA resource allocation problem into a convex

programming problem. Two algorithm was proposed to find the KSBS and the convergence of

69

the preference function based algorithm is also shown. Simulation result showed that the

algorithms can achieve a reasonable trade-off between fairness and efficiency. In[32], the paper

provides analysis of the frequency-time scheduling technique in OFDMA system. Two novels

scheduler that try to balance the trade-off fairness and capacity are proposed. The first scheduler

divides the scheduling problem into chunks allocation and chunks assignment procedures. The

chunks are assigned to users based on the instantaneous chunk’s CQI and user’s average bit rate.

The second scheduler allocates the users based on a utility function that depends upon the user’s

instantaneous and average bit rates. System level simulation was used to show that the second

scheduler results in boosting the fairness between users but out of the expense of high cell

throughput loss (loss of 35%). Consequently, the first scheduler achieves a better trade-off

between fairness and capacity than the second scheduler with low implementation cost. However

the AMC technique applied here was not fully adaptive. Here there were static pre-mapping

between MCS and CQI.In[33], they proposed a strategy for resource allocation for different

traffic classes at the medium access control(MAC) layer of wireless system based on orthogonal

frequency division multiple access(OFDMA).in order to achieve inter-class fairness, a

modification of the virtual token modified largest weighted delay first(VT-M-LWDF) and

modified largest weighted delay first(M-LWDF) rules. The proposed scenario is simulated using

LTE-SIM simulator through simulation. They show that the proposed scheduler introduces

remarkable multi-objective improvement of the quality of service performance parameters i.e.

packet loss rate, average throughput, fairness index and system spectral efficiency among

different classes of traffic such as video, VoIP and best-efforts. In[34], this paper, they

investigate the important downlink scheduling problem in LTE networks with a focus on video

streaming applications. The scheme is directly targeted on optimizing the application-layer video

70

quality with the required end-to-end delay bound. The video quality optimized scheduling is

formulated as a convex combinational optimization problem with an exponentially-growing

search space. They employ the intrinsic strength of population based solution of GA to solve

complex optimization problem because it offers a superior advantage for this type scheduling

problems. Performance of the proposed cross layer design and the GA solution is evaluated

against the well-known M-LWDF scheduling rule and a trajectory method. The simulation

results demonstrate the effectiveness of the GA based quality-optimized approach. It can

enhance the video quality significantly and satisfy the delay bound. [35]reports a new scheme for

allocating time-slots to the end users in converged EPON and LTE networks. It supports QOS by

establishing a mapping mechanism between the LTE bearers and EPON queues. The

performance of the proposed work is evaluated using OPNET simulator for the converged LTE-

EPON architecture. Simulation results reveal that throughput and other QOS services, such as

jitter, delay excel in the employed scheme than the traditional one. However the algorithm has

limitation in that it only frames a model for bandwidth allocation between OLT and ONUs, but

does not deal with the resource allocation among the services in an ONU.[36] This paper

proposed an efficient downlink radio resource allocation with carrier aggregation in LTE-

Advanced Networks. In this paper, they assume that the scheduler can reassign CCs to each UE

at each transmission time interval and formulate the downlink radio resource scheduling problem

under the modulation and coding scheme constraint, which is proved to be NP-hard.A novel

greedy-based scheme was proposed to maximize the system throughput while maintaining

proportional fairness of radio resource allocation among all UEs.An LTE-EPC network simulator

that is based on ns-3 network simulator was used to evaluate the performance of the proposed

71

scheme. Simulation results show that the proposed scheme outperforms the schemes in the

previous work.

72

CHAPTER THREE

SYSTEM MODELING

3.1 INTRODUCTION LTE represents an emerging and promising technologyfor providing broadband ubiquitous

Internet access. Forthis reason, several research groups are trying to optimizeitsperformance.

Unfortunately, no open source simulation platforms, that thescientificcommunity can use to

evaluate the performance of theentire LTE system, are freely available[87]. The lack of a

commonreference simulator does not help the work of researchersand poses limitations on the

comparison of results claimedby different research groups. In order to bridge this gap, herein,

theopen source framework LTE-Sim provides acomplete performance verification of LTE

networks. LTE-Simhas been conceived to simulate uplink and downlink schedulingstrategies in

multi-cell/multi-users environments taking intoaccount user mobility, radio resource

optimization, frequencyreuse techniques, AMC module, and other aspects very relevantfor

industrial and scientific communities. The effectiveness of theproposed simulator has been tested

and verified considering (i)the software scalability test which analyzes both memory

andsimulation time requirements; (ii) the performance evaluation ofa realistic LTE network

providing a comparison among well-knownscheduling strategies. In this work mat lab and LTE-

sim are used to achieve the aim of this work. The simulations were carried out on a computer

with these specifications: an Intel core i5 processor at 2.0GHz, 4 GB of RAM running Linus

Ubuntu 13.10.

3.2 PHYSICAL MODEL The physical model for a generic downlink resource allocation scheme is shown in Figure(3.1).

The scheduler selects the users to be allocated resources based on some scheduling parameters

such as channel conditions, packet delays, Queue length size, flow bit rates, packet queue size,

73

headof line delay, flowclasses, channel state information, flowpriorities, multiplexingmethods,

buffer status reports, interference conditions, delay budget,packet loss rate, trafficload, service

class priorities., user data rate, channel occupancy statistics,QOS requirement,

propagationcondition, power per resource block, transmission rate etc.

When the users are scheduled, the allocator allocates the selected user with RBs. The allocated

RBs are used to determine the modulation, coding and power schemes.

Once the modulation, coding and power schemes are selected, the user is informed of its status

and the MCS (if the user is selected) on the PDCCH. The user uses this information to decode its

packet at the next TTI on the PDSCH.

Subsequently, for the purpose of this work, emphasis is laid on the Scheduling section of the

downlink resource allocator model of figure 3.1. Figure 3.2 illustrates the physical representation

of a typical downlink scheduler. The scheduler receives inputs from all the users via a signaling

channel. These inputs indicate the channel quality/condition of the user per channel. This

information is referred to as the user channel quality indicator (CQI) feedback. This CQI value is

chiefly dependent on the signal to interference noise ratio (SINR) experienced by the user for

Allocator PDCCH

USER

Scheduler

PDSCH

Figure 3.1: physical model of a downlink resource allocator

74

each available channel. Other factors that influence the CQI is the mobility speed of the user

(however, this is mostly neglected in the development of research models and simulators.

Other inputs to the downlink scheduler are obtained from the eNB. Such inputs are the QoS of

the packets to be transmitted to each user, the Head on the line delay of the packets, etc.

Further, the scheduler calculates a metric for each user depending on the type of scheduling

scheme it is designed to implement. In calculating the metric per user, the scheduler uses the

parameters such as per user CQI, buffer status etc, it assembled from the user and eNB. The UEs

CQI is sent to the eNB as a control signaling information. The calculated metric per user are

compared and the user with the largest metric is selected or scheduled to use the channel for

downlink transmission at the next TTI. This information is passed via the PDCCH to the user.

3.3 PROPOSED SCHEDULING SCHEME The M-LWDF was modified to incorporate the parameters that will enhancethe performance of

thescheme. In this work anotherbandwidth of flow,� was introduced which is a function of

buffer weight and reserved rate. This weight parameter � is multiplied by the average throughput

Metric

Calculator per

user

UEs CQI

Downlink Packets Buffer

(FIFO Queue) per User

B1 B2

B3

Bn

Scheduling

decision

(comparator)

Figure. 3.2: Physical Model of a Downlink Scheduler

USER

Scheduler

75

in equation3.1 which lead to the modified form in equation3.2.Equation3.I and equation3.2

below show the M-LWDF and improved MLWDF respectively.

The scheduling problem in LTE is quite novel as it has to satisfy a number of applications and

these application tend to have different QOS satisfaction levels.This work tend to improve the

M-LWDF algorithm by using a dynamic approach that is based on the principle of fuzzy logic.

.

��,�'�� = ��,�(�� )

�� − 1) �3.1)

��,�'�� = ��,�(�� )

�� − 1)� �3.2)

Where

��=-�`Oabc)

dc

Β=�e9f ∗ hc∑ hcjckl

∑ ��m�Zn =1 0.001 ≤��≤1

All flows shall satisfy the constraint given above.

Where

��=weight parameter

��,�=head-of-line packet delay

(�� )=expected data rate for oUB user at time t on pUBresource block

�� − 1)=average throughput up to time slot t-1

76

�=bandwidth of flow

#�=maximum probability for HOL packet delay of user i to exceed the delay threshold of user i.

$�=delay threshold of user i

�e9f=user’s maximum reserved rate.

��=weight of the flow

The traffic in LTE consist of combination of real and non –real time applications categorized

into 9 different QOS service classes namely conversational voice, conversational video,Non-

conversational video, real time gaming, IMSsignaling,[voice, video(live streaming),interactive

gaming], video(buffered streaming),[TCP based(i.e.www,e-mail),chat,FTP] and P2P file sharing.

Real time traffic requires a guaranteed delivery of packet withinstringent time constraints while

maximum sustainable time requirement of non-real time traffic shall also be satisfied. Therefore

the designed scheduler shall be able to adhere to these requirements and shall provide a fair

allocation to both these classes of traffic.

The fuzzy approach proposed in this study takes as input two variables, latency requirement for

real time and throughput for non-real time traffic. The output of the fuzzy inference system is

weight, whose value will be utilized in the proposed algorithm for scheduling decisions. The

respective membership functions for all the three variables were used. The fuzzy system is built

over three linguistic variables for the input and output variables. The membershipfunctions are

defined as High, Medium, and Low. The range of all these variable is from 0 to 500ms for

latency, 0 to 50Mb/s for throughput and 0 to 1 for flow weight. The rule base consist of 18 rules.

The rule base has been defined consideringthe nature and dynamism of the input traffic.

The fuzzy system consist of three steps: fuzzification were the system reads in system input

variables i.e. throughput and latency. Fuzzy reasoning were input state variables read in previous

77

steps are manipulated on the rule base and provides an output value. Last step deffuzifization

employs center of gravitymethod to calculate a crisp value for outputweight. The output weight

is taken as the weight for real time traffic and weight of non-real time traffic is calculated by

subtracting from 1 since the total weight for all queue shall satisfy the constraint defined in

equation 3 below. The incorporated parameter which is a function of the weight ofthe flow is

gottenfromthe fuzzy inference process using mamdani inference system shown from fig (3.3) to

fig (3.7).The mamdani system using GUI was further converted to matlabcode and then

reconverted intoqrr source code to be used in equation 3.3 and 3.4 while maintaining the

constraint of equation 5 which when multiplied by the reserved rate of each user gives the value

of the parameter β. On multiplying β with average throughput led to the modified scheme as seen

in equation 3.3 and 3.4.

78

Figure 3.3: fuzzy inputs

79

Figure 3.4: applying membership

Figure 3.5: applying rules

80

Figure 3.6: aggregating all outputs/deffuzification

81

3.4SIMULATION SCENARIO In this thesis, a simulation scenario based on realistic approach was set.In this work,all the

simulations are performed by following the same scenario, they share the same parameters and

only one simulator called LTE-SIM is used. The physical models are taken from 3GPP

specifications. The LTE propagation loss model is composed by 4 different models (shadowing,

multipath, penetration loss and path loss)

3.5 SIMULATION PARAMETERS All simulation in this work sharethe same scenario with only one macro-cell. Users are

constantly moving at speed of 3kmph in random directions (random walk). The parameters of the

simulation are sown in table 3.1.

The simulation scenario for the proposed algorithm are as follows. There are 40% of users using

video flows, 40% of users using VOIP flows and the remaining 20% are using CBR flows.

In this thesis, focus is on RT services; however, to analyze the obtained results it is necessary to

compare thebehavior of RT against NRT services. To do so we consider video and VOIP flows

as RT services and CBR flows as NRT services. Since modeling NRT services (i.e. TTP, SMS,

and P2P) is extremely hard, CBR traffic as a parasite flows which send packet all thetime was

used. Besides, to model NRT flows, a simulator which support TCP is necessary, the current

version of LTE-SIM only support UDP.

Figure 3.7: surface viewer

82

3.5.1 SIMULATION TRAFFIC MODEL A video with 242kbps source video data is used in thesimulation.This traffic is a truce based

application that sends packets based on realistic video trace files. ForVoIP flow G.729 voice

flows are generated by the VOIP application. In particular, the voice flows has been modelled

with an ON/OFF model, were the ON period is exponentially distributed with mean value of 35,

and the OFF period as a truncated exponential probability density function with an upper limit of

6.9s and an average value of 3s.Duringthe ON period, the source sends 20 bytes sized packets

every 20ms (i.e.the source data rate is 8.4kbps), wereduringthe OFF period, the rate is zero

because the presence of a voice Activity detector is assumed. The CBR application generates

packets with a packet size and inter-arrival packet time.

The buffer at the scheduler is considered to be infinite i.e., the packet loss is not due to the buffer

overflow. In real time services, the maximum delay should be in therange 100-

200ms.Accordinly, thetarget delay is set 0.1. The main parameter used for the analysis is shown

below.

83

Number of users 100

Bandwidth 10MHz

Number of resource blocks 50

Scheduling time(TTI) 1ms

Number of sub-carriers 600

Number of sub-carriers per Resource block 12

Sub-carrier spacing 15kHz

Slot Duration 0.5ms

Scheduling time(TTI) 1ms

Number of OFDM symbols per slot 7

Carrier frequency 2GHz

Simulation duration 150s

Table 3.1 SIMULATION PARAMETERS

84

Flow duration 120s

Frame structure FDD

Radius 1km

CQI range 1-15

Packet generation type Exponential

Type of user queue FIFO

Maximum delay 0.1s

Video bit rate 242kbps

VOIP bitrate 8.4kbps

NRT bit rate 20kbps

Penetration loss 10dB

3.6 MODEL VALIDATION The model was validated by reducing the proposed model into the existing schemes and

compared. The proposed model was decomposed into the MLWDF schemes and was simulated.

The results generated were compared with the MLWDF and EXP/PF results shown in

literature[32]. The comparison is presented in figure3.8. The comparison shows that the

proposed model attained an average performance level and 98.7% MLWDF schemes

respectively.

85

CHAPTER FOUR

SIMULATION AND RESULT ANALYSIS

4.1 PERFORMANCE METRICS In furtherance of carrying out an evaluation of the simulation results, several metrics areused in

this work. When evaluating the QoS, it is important to focus the performanceof schedulers on

packet losses, delays, and fairness etc. the following gives the explanation of the metrics as

outlined here.

Figure 3.8: Model validation with MLWDF Scheme

86

Average Throughput per user: This metric represents the average rate of successfulmessage

delivery over physical channel. It is calculated by dividing the size ofa transmitted packets by the

time it takes to transfer the packets per each user. The metric is chosen to examine the

degradation of throughput when the number users increases.

Packet Loss Ratio (PLR): This metric aims to measure the percentage of packets of data

travelling across a physical channel which fail to reach their destination. Alsothere exist packet

losses caused by buffer overflows.

Delay: Delay measures the elapsed time between packets departing and packets destination

reached. It is at the queues where packets spend time causing undesirable delays,packet delays

must be as short as possible to perform a desirable QoS level.

Fairness Index: In order to obtain an index related to the fairness level the Jain’s fairness index

method was used. Where xi is the throughput allocated to user I among N competing flows.

Fairness index FI=�∑ fc)sW∙∑ fcs

…………………………………………4.1

Cell spectral efficiency:it is the optimized used of spectrum or bandwidth so that the maximum

amount of data can be transmitted with the fewest transmission errors. In a cellular telephone

network, spectrum efficiency equates to the maximum number of users per cell that can be

provided while maintaining an acceptable quality of service (QOS). It can also be

expressed as the totalthroughput achieved by all users divided by the available bandwidth.

87

4.2 RESULTS ANALYSIS This chapter summarizes the results obtained from the computer simulation of the proposed LTE

downlink scheduler. The results are illustrated in graphical forms in figures 4.1 to 4.12 shown

below with detailed explanation. The simulation was run using LTE-sim with the configured

parameters as listed above and results obtained using the performance indices in section 4.1 were

compared with an existing scheme.

MLWDF scheme has been chosen and respectively modified to use the incorporated weight

parameter � to improve their performance when using multimedia services such as video and

VOIP. To make an evaluation of result, the following notation is used: ‘PF’ represents the

proportional fair algorithm, ‘MLWDF’ represents the classic modified largest weight delay first

algorithm, ‘EXP/PF’ represents the classic Exponential proportional rule, ‘I-MLWDF’ represents

the improved MLWDF.

In this analysis, a percentage value is used to compare modified algorithms result to the results of

non-modified algorithms.

4.2.1 Throughput for video flows The aggregate throughput increases considerably by about 9.8% for 60 users when I-MLWDF is

used, compared to the non-modified MLWDF. An explanation for this increase is that by using

the incorporated weighty parameter, the video service having a high bit rate occupy a weightier

flow in the network and therefore larger video flow bandwidth (β), which justify the considerable

priority that they get as we can see in figure (4.1). The worst results are obtained by PF.

88

4.2.2 Throughput for VOIP flows

There is not a significant variation of throughput between modified algorithms for VOIP flows as we can

see in figure (4.2).since we use an ON/OFF model,there was no transit packet, meaning no packet

arrives at the buffer when the state is set to OFF.this explains why there is almost no variation in

throughput gain.however the modified scheme performs the expected throughput when the cell is

loaded by 60 users.

0

50000

100000

150000

200000

250000

300000

1 0 2 0 3 0 4 0 5 0 6 0

TH

RO

UG

HP

UT

NUMBER OF USERS

PF

MLWDF

EXP/PF

I-MLWDF

8900

8950

9000

9050

9100

9150

9200

9250

1 0 2 0 3 0 4 0 5 0 6 0

TH

RO

UG

HP

UT

NUMBER OF USERS

PF

MLWDF

EXP/PF

I-MLWDF

Figure 4.1: AVERAGE THROUGHPUT PER VIDEO FLOW

Figure 4.2: AVERAGE THROUGHPUT PER VOIP FLOWS

89

4.2.3Throughput for non-real time flows

MLWDF shows the most optimal results for non-real time flows. Throughput is maintained even

when the cell is totally loaded by 60 users.PF also performs good results. On the other hand, the

algorithm that use � performs the worst result-EXP/PF shows a poor performance due to a sharp

decrease of throughput. The most likely explanation is that the bandwidth of non-real time flow

is small, the scheduler grants priority to larger flows in the network such as video for instance,

therefore there might be considerable packet losses caused by buffer overflow.

4.2.4 Packet loss rate for video flow

The packet loss ratio for video flow decreases considerably by about 35.71% when using I-

MLWDF compared to the classical EXP/PF and also by about 75% when compared with

MLWDF as shown in figure (4.4). Since video flows use the highest bitrate, video queue is the

0

5,000

10,000

15,000

20,000

25,000

1 0 2 0 3 0 4 0 5 0 6 0

TH

RO

UG

HP

UT

NUMBER OF USERS

PF

MLWDF

EXP/PF

I-MLWDF

Figure 4.3: AVERAGE THROUGHPUT PER NRT FLOW

90

first to be served, so packet losses caused by buffer overflow is reduced. The worst performance

is presented by PF. This is explained by the fact that PF does not take into account packet delays.

Video service support a packet loss rate under 1%. In our scenario about 25 users reach a packet

loss rate under this threshold.

4.2.5 Packet loss rate for VOIP flows

The accepted packet loss rate for VOIP flows is under 3%.In this scenario all schedulers show

acceptable packet loss rate under 3%.when the cell is loaded by 60 users, only PF shows a packet

loss rate above 3% as shown in figure (4.5).

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

1 0 2 0 3 0 4 0 5 0 6 0

TP

AC

KE

T L

OS

S R

AT

IO

NUMBER OF USERS

PF

MLWDF

EXP/PF

I-MLWDF

Figure 4.4: PACKET LOSS RATE FOR VIDEO FLOWS

91

4.2.6 Packet loss rate for non-real time flows

The best packet loss ratio performance for non-real time is shown by MLWDF whose packet loss

ratio is under 1% when the cell is loaded by 60 users.PF also shows an acceptable packet loss

rate when the cell is loaded up to 50 users. EXP/PF shows a packet loss rate under 3% when the

cell is loaded up to 32 users.I-MLWDF shows a sharp increase of packet loss ratio.This can be

explained that because non-real time flow will be the smallest one, Therefore NRT flows will

perform high packet losses caused by buffer overflows.However, it should be noted that LTE-

sim works under UDP traffic,so considering that FTP is normally implemented under TCP, the

packet loss rate could be lower than shown in fig (4.6) due to the TCP retransmission control.

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

1 0 2 0 3 0 4 0 5 0 6 0

PA

CK

ET

LO

SS

RA

TE

NUMBER OF USERS

PF

MLWDF

EXP-RULE

I-MLWDF

Figure 4.5: PACKET LOSS RATE FOR VOIP FLOWS

92

4.2.7 Delays for video flows

Video delays are illustrated in figure (4.7). The best performance is shown by I-MLWDF with its

delay under 0.05s.the other schedulers are under 0.07s except PF which presents the worst

performance.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1 0 2 0 3 0 4 0 5 0 6 0

PA

CK

ET

LO

SS

RA

TE

NUMBER OF USERS

PF

MLWDF

EXP/PF

I-MLWDF

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1 0 2 0 3 0 4 0 5 0 6 0

DE

LAY

NUMBER OF USERS

PF

MLWDF

EXP/PF

I-MLWDF

Figure 4.6: PACKET LOSS RATE FOR NRT FLOWS

Figure 4.7: DELAY FOR VIDEO FLOWS

93

4.2.8 Delays for VOIP flows

Video delays are illustrated in figure (4.8 ).the shortest delays are performed by I-

MLWDF(under 0.007s when the cell is loaded by 60 users) the other schedulers show similar

packet delays, all of them under 0.02s, except PF(0.08s) when the cell is loaded up to 60 users.

4.2.9Delays for non-real time

The non-real time delays are shown in figure (4.9). The shorter delays is shown by MLWDF.

These results compliments the result obtained in throughput and packet loss rate figures. We

earlier assumed that the poor throughput performance is caused by packet losses. We also

assumed that packet losses is caused by buffer overflows. Now regarding fig (4.3), our

assumption about high buffer overflow are confirmed because the packet delay are in the average

range in physical layer transmission.

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

1 0 2 0 3 0 4 0 5 0 6 0

DE

LAY

NUMBER OF USERS

PF

MLWDF

EXP/PF

I-MLWDF

Figure 4.8: DELAY FOR VOIP FLOWS

94

4.2.10 Fairness index for video flow

The fairness index for video flows is shown in figure (4.10). The best results are performed by I-

MLWDF and EXP/PF that reach a fairness index above 95%. The worst results are performed by

PF that reaches a fairness index of 61%.

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

1 0 2 0 3 0 4 0 5 0 6 0

DE

LAY

NUMBER OF USERS

PF

MLWDF

EXP/PF

I-MLWDF

0

0.2

0.4

0.6

0.8

1

1.2

1 0 2 0 3 0 4 0 5 0 6 0

FA

IRN

ES

S I

ND

EX

NUMBER OF USERS

PF

MLWDF

EXP/PF

I-MLWDF

Figure 4.9: DELAY FOR NRT FLOWS

95

4.2.11 Fairness index for VOIP flow

Figure (4.11) shows the fairness index for VOIP flows. Since the VOIP flows are created by an

ON/OFF method, the curves show ups and downs. All algorithms show a fairness index

performance between 98.5% and 99.5%.

4.2.12 Fairness index for non-real time flows

The fairness index for non-real time flows shown in figure (4.12) shows that the highest fairness

index are presented by MLWDF and PF that reach above the 99%.the worst fairness index levels

are showed by I-MLWDF and that reach a level of about 93% .

0.974

0.976

0.978

0.98

0.982

0.984

0.986

0.988

0.99

1 0 2 0 3 0 4 0 5 0 6 0

FA

IRN

ES

S I

ND

EX

NUMBER OF USERS

PF

MLWDF

EXP/PF

I-MLWDF

Figure 4.10: FAIRNESS INDEX FOR VIDEO FLOWS

Figure 4.11: FAIRNESS INDEX FOR VOIP FLOWS

96

As we can see in this section, the use of the weighty parameter � improves the performance of

real time flows such as video and VOIP. On the other hand its performance for non-real time

flows is relatively poor. The algorithm I-MLWDF show the best performance for video and

VOIP flows in terms of throughput, packet loss rate, delay and fairness index. These schedulers

show considerable packet losses due to buffer overflow in non-real time flows.

0.88

0.9

0.92

0.94

0.96

0.98

1

1.02

1 0 2 0 3 0 4 0 5 0 6 0

FA

IRN

ES

S I

ND

EX

NUMBER OF USERS

PF

MLWDF

EXP/PF

I-MLWDF

Figure 4.12: FAIRNESS INDEX FOR NRT FLOWS

97

CHAPTER FIVE

CONCLUSION AND RECOMMENDATIONS

5.1 CONCLUSION Long term Evolution technology is presently the most promising global

telecommunicationsystem. LTE provides QoS support for heterogeneous classes of traffic

withdifferent QoS requirements. In this thesis, we focus our attention on the quality ofservice in

downlink system for macro cell scenarios. This work was started by highlighting the limitations

of existent solutions for packetscheduling techniques in downlink system in macro cell scenarios.

Two resource allocation schemes specifically focusing on realtime services such as video and

VoIP for macro cell scenarios was proposed. This choice can be argued by the fact that there

already exist efficient algorithms for non-real time services suchas PF and M-LWDF in 3G

technologies. The two schemesbrought in another weighty parameter to the MLWDF and

EXP/PF algorithm respectively. However this algorithm is unfair in a multiservice level, the

scheduling priority is granted preferentially to real time services while non-real time services are

forced to wait for allocation. A possible solution could be using two types of scheduling at each

TTI, one of them presented in this work for real time services and another one such as MLWDF

or PF for non-real time services. In concluding this work, it can be said that by dividing the

whole scheme into some layers, the resource allocation task can be simplified by the fact that

each layer is responsible in granting a specific characteristic.

5.2 RECOMMENDATION � The question ‘Is it possible to set the scheduler to use two different algorithms, one for

real time services and another for non-real time services?’ should be taken into

consideration.

98

� The approaches used in this work should be modified to be adapted to uplink system in

future work.

� Future evolution can lead researchers to extend this study to LTE-Advanced

� The integration and development of a scheduling scheme that will incorporate most

possible scheduling parameters to maximally serve both real time and none-real time

services in high and low load scenarios.

99

REFERENCES

[1] K.Feher, Wireless Digital Communication modulation and spread spectrum Applications. New Delhi: prentice-Hall Inc, 2006.

[2] R.Frenkiel, “A brief History of mobile communications.” [Online]. Available: www.winlab.rutgers.edu/naravan/course/wireless Revolution/vts article.pdf.

[3] B. A. Forouzan, Data communications and Networking, 5th ed. New Delhi: McGraw Hill Education;, 2013.

[4] A. Goldsmith, wireless communications. cambridge: cambridge University press, 2005.

[5] A. pashtan, Wireless terrestrial communications: cellular telephony. illinois: Eolss publishers, 2006.

[6] T. Ali-Yahiya, understanding LTE and its performance. london: springer New York Dordrecht Heidelberg, 2011.

[7] “Free scale semiconductor. Inc, Long term evolution protocol overview,White paper.” 2008.

[8] F. Capozzi, G. Piro, S. Member, L. A. Grieco, G. Boggia, S. Member, P. Camarda, and R. Robin, “Downlink Packet Scheduling in LTE Cellular Networks : Key Design Issues and a Survey,” IEEE Commun. Surv. tutorials, vol. 15, no. 2, pp. 678–700, 2013.

[9] M. Iturralde, T. Ali-yahiya, A. Wei, M. Iturralde, T. Ali-yahiya, and A. Wei, “Resource Allocation for Real Time Services in LTE Networks : Resource Allocation Using Cooperative Game Theory and Virtual Token Mechanism To cite this version :,” Wirel. Pers. Commun. verlag, vol. 72, no. 2, pp. 1415–1435, 2015.

[10] J. Dunlop and D. G. Smith, Telecommunications engineering, 3rd editio. 2-6 boundary Row,london SEI 8UN,UK: chapman and hall, 1994.

[11] T.Nilsson, “Towards a new era in mobile communication.” [Online]. Available: http://19378.100.33/(Ericson www server).

100

[12] T. Halomen, J. Romero, and J. Melero, GSM, GPRS and EDGE Performance:Evolution Towards 3G/UMTS. England: wiley, 2002.

[13] S. Jochen, Mobile communications. England: Addison Wesley, 2003.

[14] H. Kaaranen, A. Ahtiainen, L. Laitinen, S. Naghian, and V. Niemi, UMTS Networks: Architecture ,Mobility and services. England: wiley, 2005.

[15] A. Meylan, “LTE Radio Layer 2 , RRC and Radio Access Network Architecture,” 2010.

[16] M. Sauter, Beyond 3G – Bringing Networks , Terminals and the Web Together. germany, 2009.

[17] H. Halomen and A. Toskala, HSDPA/HSUPA for UMTS:High Speed Radio Access for mobile communications. England: Wiley, 2006.

[18] V. viervi, D. Aleksandar, and V. Branimir, The cdma 2000 system for mobile communications:3G wireless Evolution. U.S.A: prentice-Hall Inc, 2004.

[19] “3gpp .TS 24.301:Non-access-stratum(NAS) protocol for Evolved packet system(EPS):stage 3.”

[20] A. David, D. Erik, F. Anders, J. Ylva, L. Magnus, and P. Stefan, “LTE : The Evolution of Mobile Broadband,” IEEE Commun. Mag., vol. 47, no. 4, pp. 44–52, 2009.

[21] P. Scheduler, “LTE Radio Access , Rel RL30 , Operating Documentation , Issue 02,” no. 02, 2011.

[22] M. Andreas and M. Jens, UMTS System Architecture and Protocol Architecture UMTS / GSM Network Architecture. 2010.

[23] “Motorola,Long term evolution(LTE): overview of LTE Air-interface technical white paper.,” 2007.

[24] “3Gpp TS 36.322, ‘ Evolved Universal Terrestrial Radio Access (E-UTRA) Radio Link Control (RLC) protocol specification, ’ Release 9.”

[25] “3Gpp TS 36.321, ‘ Evolved Universal Terrestrial Radio Access (E-UTRA) Radio Link Control (RLC) protocol specification, ’ Release 9.”

[26] “IETF RFC 462: IPv6 stateless Address Auto configuration.”

[27] “3gpp TS 22.246:Multimedia Broadcast/Multicast service(MBMS) user services: stage 1.” .

[28] “3gpp TS 22.146: Multimedia Broadcast/multicast service (MBMS): stage 1.”

101

[29] “IETF RFC 3633: IPv6 prefix options for dynamic Host configuration protocol(DHCP) version 6.” .

[30] H. Harri and T. Antti, LTE for UMTS: Evolution to LTE-Advanced, 2nd ed. Technology and Enginering, 2011.

[31] M. Abd-Elhamid, H. Taha, and N. A. A. Hassanein, LTE,LTE-ADVANCED AND WIMAX TOWARDS IMT-ADVANCED NETWORK, First. united kingdom: wiley and sons, LTD, 2011.

[32] F. Capozzi, G. Piro, S. Member, L. A. Grieco, G. Boggia, S. Member, and P. Camarda, “Downlink Packet Scheduling in LTE Cellular Network-Key Design Issues and a Survey,” IEEE Commun. Surv. Tutorials, vol. 15, no. 2, pp. 678–700, 2013.

[33] D. Singh and P. Singh, “Radio Resource Scheduling in 3GPP LTE : A Review,” Int. J. Eng. trends Technol., vol. 4, no. 6, pp. 2405–2411, 2013.

[34] A. S. Tanenbaum, Modern operating system, 3rd ed. upper saddle River,Nj,USA: prentice hall press, 2007.

[35] S. Hussian, “Dynamic Radio resource management in 3gpp LTE.,” blekage Inst. Technol., 2009.

[36] P. Kela, “Dynamic packet scheduling performance in ULTRA long term evolution downlink.,” 3rd Int. Symp. Wirel. perversive Comput., pp. 308–313, 2008.

[37] A. L. Stolyar and K. Ramanan, “Largest weighted delay first scheduling: large deviations and optimality.,” Ann. Probab., vol. 11, no. 1, pp. 1–48, 2001.

[38] Choi, K. Won, J. Wha Soak, and J. Dong Geun, “Resource allocation in OFDMA wireless communications systems supporting Multimedia services.,” networking/IEEE Trans., vol. 17, no. 3, pp. 926–936, 2009.

[39] Almatarneh, Rabie, A. Mohamed, and D. Octavia, “Frequency-time scheduling Algorithm for OFDMA systems.,” Electr. Comput. Eng. 09. Can. Conf. IEEE 2009, 2009.

[40] Almatarneh, Rabie, A. Mohamed, and O. Dobre, “Performance analysis of proportional fair scheduling in OFDMA wireless systems.,” Veh. Technol. Conf. fall(VTC 2010-fall), 2010.

[41] W. Xongxin, “QOS oriented Time and frequency domain packet scheduling for the ultra long term evolution.,” Int. Conf. Wirel. Commun., vol. 3, no. 2, pp. 1–5, 2014.

[42] S. Sharmila, “Scheduling with QOS provisioning in mobile broadband wireless systems.,” Int. J. Wirel. Mob. networks(IJWMN), vol. 4, no. 1, pp. 16–18, 2013.

102

[43] Y. Zaki, T. Weerawardane, C. Görg, and A. Timm-giel, “Multi-QoS-aware Fair Scheduling for LTE,” IEEE 73rd,vehicular Technol. Conf. spring), vol. 1, pp. 1–5, 2011.

[44] H. Adibah, M. Ramli, R. Basukala, K. Sandrasegaran, and R. Patachaianand, “Performance of Well Known Packet Scheduling Algorithms in the Downlink 3GPP LTE System,” Proc. 2009 IEEE 9th Malaysia Int. Conf. Commun., vol. 1, pp. 815–820, 2009.

[45] R. Basukala, H. A. M. Ramli, and K. Sandrasegaran, “Performance Analysis of EXP / PF and M-LWDF in Downlink 3GPP LTE System,” IEEE First Asian Himalyas Conf., vol. 1, pp. 1–5, 2009.

[46] Kim, “An efficient scheduling scheme to enhance the capacity of VOIP services in Evolved UTRA uplink.,” EURASIP J. Wirel. Commun. Netw., 2008.

[47] C. Sunggu, “Mac scheduling scheme for VOIP traffic services in 3G LTE.,” Veh. Technol. Conf. 2007,VTC-2007 fall,IEEE 66th 2007, 2007.

[48] Ting and Kuo-chang, “Energy-efficient DRX scheduling for QOS traffic in LTE networks.,” IEEE 9th Int. Symp. IEEE, 2011.

[49] S. Videv and H. Harald, “Energy-efficient scheduling and bandwidth energy efficiency trade-off with low load.,” Commun. Int. Conf. IEEE, 2011.

[50] Dechene, Dan, and S. Abdullah, “Energy-aware resource allocation strategies for LTE uplink with synchronous HARQ constaints.,” IEEE Trans. Mob. Comput., vol. 13, no. 2, pp. 422–433, 2014.

[51] D.sabella, “Energy efficiency evaluation of state of the art packet scheduling algorithm for LTE,” proceeding IEEE Eur. Wirel. Conf., pp. 1–4, 2011.

[52] A.Kaufmann, Introduction to the theory of fuzzy sets. New york: Academia press, 1975.

[53] L. Zadeh, “Fuzzy logic.pdf,” IEEE Comput., vol. 21, no. 4, pp. 83–93, 1988.

[54] E. Mandani, “Application of fuzzy logic to approximate reasoning using linguistic systems.,” IEEE Trans. Comput., vol. 26, no. 12, pp. 1182–1191, 1977.

[55] L. Zadeh, “Knowledge representation in fuzzy logic.,” IEEE Trans. Knowl. Data Eng., vol. 1, pp. 89–100, 1989.

[56] I. G. Fraimis, V. D. Papoutsis, and S. A. Kotsopoulos, “A Distributed Radio Resource Allocation Algorithm with Interference Coordination for Multi-cell OFDMA Systems,” IEEE 21st Int. Symp. Pers. indoor Mob. radio Commun., pp. 1354–1359, 2010.

103

[57] Q. Ai, P. Wang, F. Liu, Y. Wang, F. Yang, and J. Xu, “QoS-Guaranteed Cross-Layer Resource Allocation Algorithm for Multiclass Services in Downlink LTE System,” Int. Conf. Wirel. Commun. signal Process., vol. 2, pp. 1–4, 2010.

[58] C. Wengerter, J. Ohlhorst, A. Golitschek, E. Von Elbwart, and P. Mcom, “Fairness and Throughput Analysis for Generalized Proportional Fair Frequency Scheduling in OFDMA,” Proc. IEEE Veh. Technol. Conf., vol. 3, no. 2, pp. 1903–1907, 2005.

[59] M. Proebster and C. M. Mueller, “Adaptive Fairness Control for a Proportional Fair LTE Scheduler,” IEEE 21st Int. symosium Pers. indoor Mob. radio Commun., vol. 1, pp. 1504–1509, 2010.

[60] K. C. Beh, S. Armour, and A. Doufexi, “Joint Time-Frequency Domain Proportional Fair Scheduler with HARQ for 3GPP LTE Systems,” Proc. IEEE Veh. Technol. Conf., vol. 1, pp. 1–5, 2008.

[61] G. Piro, S. Member, L. A. Grieco, G. Boggia, S. Member, R. Fortuna, and P. Camarda, “Two-level downlink scheduling for real-time multimedia services in LTE networks,” IEEE Trans. Multimed., vol. 13, no. 5, pp. 1052–1065, 2011.

[62] S. Ali, M. Zeeshan, and A. Naveed, “A capacity and minimum guarantee-based service class-oriented scheduler for LTE networks,” EURASIP J. Wirel. Commun. Netw., vol. 2, no. 1, pp. 1–15, 2013.

[63] C. veer. Lakshmi and S. Sharmila, “Enhancement Of QOS in LTE Downlink systems Using Frequency Diversity Selectivity Scheduling,” Int. J. Res. Eng. Technol., vol. 3, no. 7, pp. 742–747, 2014.

[64] V. Sharma and P. P. K. Chopra, “Performance Analysis of A Novel Scheduling Algorithm for Different Traffics in LTE Networks,” Int. J. Comput. Sci. Technol., vol. 5, no. 5, pp. 6631–6634, 2014.

[65] R. Kwan, C. Leung, and J. Zhang, “Downlink Resource Scheduling in an LTE System,” IEEE Signal Process. Lett., vol. 16, no. 16, pp. 461–464, 2009.

[66] Shampy and R. Shaveta, “Scheduling Algorithms of Resource Allocation in LTE system,” Int. J. Adv. Res. Comput. Sci. Manag. Stud., vol. 2, no. 6, pp. 33–38, 2014.

[67] R. Madan, S. P. Boyd, S. Lall, and S. Member, “Wireless Cellular Networks,” IEEE/ACM Trans. Netw., vol. 18, no. 3, pp. 973–984, 2010.

[68] S. . Chadchan and C. . Akki, “A Fair Downlink Scheduling Algorithm for 3GPP LTE Networks,” I.J.Computer Netw. Inf. Secur., vol. 6, no. 5, pp. 34–41, 2013.

104

[69] X. Yang and Y. Wang, “Resource Allocation in LTE OFDMA Systems using Genetic Algorithm and Semi-Smart Antennas,” Wirel. Commun. Netw. Conf., vol. 3, no. 2, pp. 1–6, 2010.

[70] M. S. Bahreyni and V. Sattari-naeini, “Fairness Aware Downlink Scheduling Algorithm for LTE Networks,” J. Math. Comput. Sci., vol. 11, no. 2, pp. 53–63, 2014.

[71] M. Salem, A. Adinoyi, M. Rahman, H. Yanikomeroglu, D. Falconer, and Y. Kim, “Fairness-Aware Radio Resource Management in Downlink OFDMA Cellular Relay Networks,” IEEE Trans. Wirel. Commun., vol. 9, no. 5, pp. 1628–1639, 2010.

[72] P. Karthik and G. Kumaran, “Optimal Resource Allocation for Wireless Networks with Inter-Cell Interference,” Int. J. Adv. Res. Comput. Commun. Eng., vol. 2, no. 11, pp. 4490–4493, 2013.

[73] Y. Barayan, I. Kostanic, and K. Rukieh, “Performance with MIMO for the Downlink 3GPP LTE Cellular Systems,” Univers. J. Commun. Netw., vol. 2, no. 2, pp. 32–39, 2014.

[74] C. Lin, R. Hwang, J. Wu, J. Lee, and Y. Lin, “Integration of Spatial Reuse and Allocation for Downlink Broadcast in LTE - Advanced and WiMAX Relay Networks,” IEEE Trans. Veh. Technol., vol. 5, no. 99, pp. 1–10, 2014.

[75] S. F. Sulthana and R. Nakkeeran, “Study of Downlink Scheduling Algorithms in LTE Networks,” J. Networks, vol. 9, no. 12, pp. 3381–3391, 2014.

[76] Lin.Cheng-chung, “frequency and time domain packet scheduling based on channel prediction with Imperfect CQI in LTE,” Int. J. Wirel. Mob. networks(IJWMN), vol. 5, no. 4, pp. 157–170, 2013.

[77] S. Wu, “A Channel Quality-aware Scheduling and Resource Allocation Strategy for Downlink LTE Systems,” J. Comput. Inf. Syst., vol. 8, no. 2, pp. 695–707, 2012.

[78] D. Keerthana and G. Vinoth, “Downlink Resource Allocation for Next Generation Wireless Network Using Turbo Codes over Non-Linear Channel,” Int. J. Innov. Res. Sci. Eng. Technol., vol. 3, no. 1, pp. 794–798, 2014.

[79] L. Anchora, L. Badia, and M. Zorzi, “A framework for scheduling and Resource Allocation in LTE Downlink Using Nash Bargaining Theory,” IEEE Trans. Commun., vol. 53, no. 8, pp. 1366–1376, 2005.

[80] P. J. Ashidani and P. R. Guardieiro, “Downlink Scheduler Based on Deadlines for LTE Networks,” IEEE Int. Conf. EXPO(ICME), 2012.

[81] J. Makara and N. Ventura, “Downlink Packet Scheduling for Variable Bitrate Traffic in LTE Networks,” Proc. South. Africa Telecommun. Networks Appl. Conf., pp. 4–7, 2011.

105

[82] J. Chen and A. L. Swindlehurst, “Downlink Resource Allocation for Multi-user MIMO-OFDMA Systems : The Kalai-Smorodinsky Bargaining Approach,” Comput. Adv. multi-sensor Adapt. Process. IEEE Int. Work. IEEE, 2009.

[83] X. Cheng and P. Mohapatra, “Quality-optimized Downlink Scheduling for Video Streaming Applications in LTE Networks,” Glob. Commun. Conf., pp. 1914–1919, 2012.

[84] G. Piro, S. Member, L. A. Grieco, G. Boggia, S. Member, R. Fortuna, and P. Camarda, “Two-Level Downlink Scheduling for Real-Time Multimedia Services in LTE Networks,” IEEE Trans. Multimed., vol. 13, no. 5, pp. 1052–1065, 2011.

[85] P. Skillermark and P. Frenger, “Enhancing Energy Efficiency in LTE with Antenna Muting,” J. Comput. Inf. Syst., vol. 3, no. 2, 2012.

[86] H. Liao, P. Chen, and W. Chen, “An efficient Downlink Radio Resource Allocation with Carrier Aggregation in LTE-Advanced Networks,” IEEE Trans. Mob. Comput., vol. 13, no. X, pp. 1–11, 2014.

[87] Q. Liu, Z. Zou, I. Ruiz, H. Mohammed, P. Plasin, G. Piro, L. A. Grieco, F. Capozzi, and A kamar et al, “Simulating LTE Cellular Systems : An Open-Source Framework,” IEEE Trans. Technol., vol. 8, no. 1, pp. 82–88, 2012.

[88] J. Huang and Z. M. Mao, “An In-depth Study of LTE : Effect of Network Protocol and Application Behavior on Performance,” pp. 363–374.

APPENDIX

FUZZY INFERENCE PROCESSES TO DETERMINE FLOW

WEIGHT #include <stdio.h> #include <stdlib.h> double ruleAccumulationMethod_max(double defuzzifierValue, double valueToAggregate) { return ( defuzzifierValue > valueToAggregate ? defuzzifierValue : valueToAggregate ); } double ruleActivationMethod_min(double degreeOfSupport, double membership) { return (degreeOfSupport < membership ? degreeOfSupport : membership); } double ruleConnectionMethod_and(double antecedent1, double antecedent2) { return (antecedent1 < antecedent2 ? antecedent1 : antecedent2); } class FunctionBlock_tipper { public: // VAR_INPUT double latency; double throughput;

106

// VAR_OUTPUT double flow weight; private: // FUZZIFY latency double latency_low; double latency_medium; double latency_high; // FUZZIFY throughput double throughput_low; double throughput_medium; double throughput_high; // DEFUZZIFY flow-weight double defuzzify_flow-weight[1000]; public: FunctionBlock_tipper(); void calc(); void print(); private: void defuzzify(); void fuzzify(); void reset(); double membership_latency_low(double x); double membership_latency_medium(double x); double membership_latency_high(double x); double membership_throughput_low(double x); double membership_throughput_medium(double x); double membership_throughput_high(double x); B.1. 97 double membership_flow-weight_low(double x); double membership_flow-weight_high(double x); void calc_No1(); }; // Constructor FunctionBlock_tipper::FunctionBlock_tipper() { Stability = 0.0; } // Calculate function block void FunctionBlock_tipper::calc() { reset(); fuzzify(); calc_No1(); defuzzify(); } // RULEBLOCK No1 void FunctionBlock_tipper::calc_No1() { // RULE 1 : IF (((throughput IS low) AND (latency IS low)) AND THEN flow-weight IS high; double degreeOfSupport_1 = 1.0 * ( ruleConnectionMethod_and(ruleConnectionMethod_and(ruleConnectionMethod_ and(throughput_low ,latency_high) , if( degreeOfSupport_1 > 0 ) { for (int i = 0 ; i < 1000 ; i++ ) { double x = 0.0 + i * 0.1; double membership = membership_flow-weight_high(x); double y = ruleActivationMethod_min( degreeOfSupport_1 , membership ); defuzzify_flow-weight[i] += ruleAccumulationMethod_max(

107

defuzzify_flow-weight[i], y ); } } // RULE 2 : IF (((throughput is low) AND (latency IS medium)) THEN flow-weight IS low; degreeOfSupport_1 = 1.0 * ( ruleConnectionMethod_and(ruleConnectionMethod_and (ruleConnectionMethod_and(throughput_low , latency_medium)); if( degreeOfSupport_1 > 0 ) { for (int i = 0 ; i < 1000 ; i++ ) { double x = 0.0 + i * 0.1; double membership = membership_flow-weight_low(x); double y = ruleActivationMethod_min( degreeOfSupport_1 , membership ); defuzzify_flow-weight[i] += ruleAccumulationMethod_max( defuzzify_flow-weight[i], y ); } } // RULE 3 : IF (((throughput IS low) AND (latency IS high)) AND THEN Stability IS high; degreeOfSupport_1 = 1.0 * ( ruleConnectionMethod_and (ruleConnectionMethod_and(ruleConnectionMethod_and(throughput_low , latency_high)); if( degreeOfSupport_1 > 0 ) { for (int i = 0 ; i < 1000 ; i++ ) { double x = 0.0 + i * 0.1; double membership = membership_flow-weight_high(x); double y = ruleActivationMethod_min( degreeOfSupport_1 , membership ); defuzzify_flow-weight[i] += ruleAccumulationMethod_max( defuzzify_flow-weight[i], y ); } } // RULE 4 : IF (((throughput IS medium) AND (latency IS low)) ) THEN flow-weight IS low; degreeOfSupport_1 = 1.0 * ( ruleConnectionMethod_and(ruleConnectionMethod_and (ruleConnectionMethod_and(throughput_medium , latency_low)); if( degreeOfSupport_1 > 0 ) { for (int i = 0 ; i < 1000 ; i++ ) { double x = 0.0 + i * 0.1; B.1. 99 double membership = membership_flow-weight_low(x); double y = ruleActivationMethod_min( degreeOfSupport_1 , membership ); defuzzify_flow-weight[i] += ruleAccumulationMethod_max( defuzzify_flow-weight[i], y ); } } // RULE 5 : IF (((throughput IS medium) AND (latency IS medium) THEN Stability IS low; degreeOfSupport_1 = 1.0 * ( ruleConnectionMethod_and(ruleConnectionMethod_and (ruleConnectionMethod_and(throughput_medium , latency_medium); if( degreeOfSupport_1 > 0 ) { for (int i = 0 ; i < 1000 ; i++ ) { double x = 0.0 + i * 0.1; double membership = membership_flow-weight_low(x);

108

double y = ruleActivationMethod_min( degreeOfSupport_1 , membership ); defuzzify_flow-weight[i] += ruleAccumulationMethod_max( defuzzify_flow-weight[i], y ); } } // RULE 6 : IF (((throughput IS medium) AND (latency IS high)) THEN flow-weight IS high; degreeOfSupport_1 = 1.0 * ( ruleConnectionMethod_and(ruleConnectionMethod_and (ruleConnectionMethod_and(throughput_medium , latency_high); if( degreeOfSupport_1 > 0 ) { for (int i = 0 ; i < 1000 ; i++ ) { double x = 0.0 + i * 0.1; double membership = membership_flow-weight_high(x); double y = ruleActivationMethod_min( degreeOfSupport_1 , membership ); defuzzify_flow-weight[i] += ruleAccumulationMethod_max( defuzzify_flow-weight[i], y ); } } // RULE 7 : IF (((throughput IS high) AND (latency IS low)) AND (Angle IS good) THEN flow-weight IS low; degreeOfSupport_1 = 1.0 * ( ruleConnectionMethod_and(ruleConnectionMethod_and (ruleConnectionMethod_and(throughput_high , latency_low)); if( degreeOfSupport_1 > 0 ) { for (int i = 0 ; i < 1000 ; i++ ) { double x = 0.0 + i * 0.1; double membership = membership_flow-weight_low(x); double y = ruleActivationMethod_min( degreeOfSupport_1 , membership ); defuzzify_flow-weight[i] += ruleAccumulationMethod_max( defuzzify_flow-weight[i], y ); } } // RULE 8 : IF (((throughput IS high) AND (latency IS medium)) THEN flow-weight IS low; degreeOfSupport_1 = 1.0 * ( ruleConnectionMethod_and(ruleConnectionMethod_and (ruleConnectionMethod_and(throughput_high , latency_medium)); if( degreeOfSupport_1 > 0 ) { for (int i = 0 ; i < 1000 ; i++ ) { double x = 0.0 + i * 0.1; double membership = membership_flow-weight_low(x); double y = ruleActivationMethod_min( degreeOfSupport_1 , membership ); defuzzify_flow-weight[i] += ruleAccumulationMethod_max( defuzzify_flow-weight[i], y ); } } // RULE 9 : IF (((throughput IS high) AND (latency IS high)) THEN flow-weight IS high; degreeOfSupport_1 = 1.0 * ( ruleConnectionMethod_and(ruleConnectionMethod_and (ruleConnectionMethod_and(throughput_high , latency_high) , Velocity_good) , Angle_good) ); if( degreeOfSupport_1 > 0 ) {

109

for (int i = 0 ; i < 1000 ; i++ ) { B.1. 101 double x = 0.0 + i * 0.1; double membership = membership_flow-weight_high(x); double y = ruleActivationMethod_min( degreeOfSupport_1 , membership ); defuzzify_flow-weight[i] += ruleAccumulationMethod_max( defuzzify_flow-weight[i], y ); } } // RULE 10 : IF (((latency IS low) AND (throughput IS low)) THEN flow-weight IS high; degreeOfSupport_1 = 1.0 * ( ruleConnectionMethod_and (ruleConnectionMethod_and(ruleConnectionMethod_and(latency_low , throughput_low)); if( degreeOfSupport_1 > 0 ) { for (int i = 0 ; i < 1000 ; i++ ) { double x = 0.0 + i * 0.1; double membership = membership_flow-weight_high(x); double y = ruleActivationMethod_min( degreeOfSupport_1 , membership ); defuzzify_flow-weight[i] += ruleAccumulationMethod_max( defuzzify_flow-weight[i], y ); } } // RULE 11 : IF (((latency IS low) AND (throughput IS medium)) THEN flow-weight IS low; degreeOfSupport_1 = 1.0 * ( ruleConnectionMethod_and(ruleConnectionMethod_and (ruleConnectionMethod_and(latency_low , throughput_medium)) ; if( degreeOfSupport_1 > 0 ) { for (int i = 0 ; i < 1000 ; i++ ) { double x = 0.0 + i * 0.1; double membership = membership_tflow-weigh_low(x); double y = ruleActivationMethod_min( degreeOfSupport_1 , membership ); defuzzify_flow-weight[i] += ruleAccumulationMethod_max( defuzzify_flow-weight[i], y ); } } // RULE 12 : IF (((latency IS low) AND (throughput IS high)) THEN flow-weight IS low; degreeOfSupport_1 = 1.0 * ( ruleConnectionMethod_and(ruleConnectionMethod_and (ruleConnectionMethod_and(latency_low , throughput_high)); if( degreeOfSupport_1 > 0 ) { for (int i = 0 ; i < 1000 ; i++ ) { double x = 0.0 + i * 0.1; double membership = membership_flow-weight_normal(x); double y = ruleActivationMethod_min( degreeOfSupport_1 , membership ); defuzzify_flow-weight[i] += ruleAccumulationMethod_max( defuzzify_flow-weight[i], y ); } } // RULE 13 : IF (((latency IS medium) AND (throughput IS low)) THEN flow-weight IS high; degreeOfSupport_1 = 1.0 * ( ruleConnectionMethod_and(ruleConnectionMethod_and (ruleConnectionMethod_and(latency_medium , throughput_low));

110

if( degreeOfSupport_1 > 0 ) { for (int i = 0 ; i < 1000 ; i++ ) { double x = 0.0 + i * 0.1; double membership = membership_flow-weight_high(x); double y = ruleActivationMethod_min( degreeOfSupport_1 , membership ); defuzzify_flow-weight[i] += ruleAccumulationMethod_max( defuzzify_flow-weight[i], y ); } } // RULE 14 : IF (((latency IS medium) AND (throughput IS medium)) THEN flow-weight IS high; degreeOfSupport_1 = 1.0 * ( ruleConnectionMethod_and(ruleConnectionMethod_and (ruleConnectionMethod_and(latency_medium , throughput_medium)); if( degreeOfSupport_1 > 0 ) { for (int i = 0 ; i < 1000 ; i++ ) { B.1. 103 double x = 0.0 + i * 0.1; double membership = membership_flow-weight_high(x); double y = ruleActivationMethod_min( degreeOfSupport_1 , membership ); defuzzify_flow-weight[i] += ruleAccumulationMethod_max( defuzzify_flow-weight[i], y ); } } // RULE 15 : IF (((latency IS medium) AND (throughput IS high)) THEN flow-weight IS low; degreeOfSupport_1 = 1.0 * ( ruleConnectionMethod_and(ruleConnectionMethod_and (ruleConnectionMethod_and(latency_medium , throughput_high) ); if( degreeOfSupport_1 > 0 ) { for (int i = 0 ; i < 1000 ; i++ ) { double x = 0.0 + i * 0.1; double membership = membership_flow-weight_low(x); double y = ruleActivationMethod_min( degreeOfSupport_1 , membership ); defuzzify_flow-weight[i] += ruleAccumulationMethod_max( defuzzify_flow-weight[i], y ); } } // RULE 16 : IF (((latency IS high) AND (throughput IS low)) THEN flow-weight IS high; degreeOfSupport_1 = 1.0 * ( ruleConnectionMethod_and(ruleConnectionMethod_and (ruleConnectionMethod_and(latency_high , throughput_low)); if( degreeOfSupport_1 > 0 ) { for (int i = 0 ; i < 1000 ; i++ ) { double x = 0.0 + i * 0.1; double membership = membership_flow-weight_high(x); double y = ruleActivationMethod_min( degreeOfSupport_1 , membership ); defuzzify_flow-weight[i] += ruleAccumulationMethod_max( defuzzify_flow-weight[i], y ); } } // RULE 17 : IF (((latency IS high) AND (tthroughput IS medium))

111

THEN flow-weight IS high; degreeOfSupport_1 = 1.0 * ( ruleConnectionMethod_and(ruleConnectionMethod_and (ruleConnectionMethod_and(latency_high ,throughput _medium) , if( degreeOfSupport_1 > 0 ) { for (int i = 0 ; i < 1000 ; i++ ) { double x = 0.0 + i * 0.1; double membership = membership_flow-weight_high(x); double y = ruleActivationMethod_min( degreeOfSupport_1 , membership ); defuzzify_flow-weight[i] += ruleAccumulationMethod_max( defuzzify_flow-weight[i], y ); } } // RULE 18 : IF (((latency IS high) AND (throughput IS high)) THEN flow-weight IS low; degreeOfSupport_1 = 1.0 * ( ruleConnectionMethod_and(ruleConnectionMethod_and (ruleConnectionMethod_and(latency_high , throughput_high)); if( degreeOfSupport_1 > 0 ) { for (int i = 0 ; i < 1000 ; i++ ) { double x = 0.0 + i * 0.1; double membership = membership_flow-weight_low(x); double y = ruleActivationMethod_min( degreeOfSupport_1 , membership ); defuzzify_flow-weight[i] += ruleAccumulationMethod_max( defuzzify_flow-weight[i], y ); } // Defuzzify void FunctionBlock_tipper::defuzzify() { double sum_flow-weight = 0.0; double wsum_flow-weight = 0.0; for (int i = 0; i < 1000 ; i++ ) { double x = 0.0 + i * 0.1; sum_Stability += defuzzify_flow-weight[i]; wsum_Stability += x * defuzzify_flow-weight[i]; } flow-wight = wsum_flow-weight / sum_flow-weight; } // Fuzzify all variables void FunctionBlock_tipper::fuzzify() { latency_low = membership_latency_low(latency); latency_medium = membership_latency_medium(latency); latency_high = membership_latency_high(latency); throughput_low = membership_throughput_low(throughput); throughput_medium = membership_throughput_medium(throughput); throughput_high = membership_throughput_high(throughput); } // Membership functions double FunctionBlock_tipper::membership_latency_low(double x) { if ( x <= 20.0 ) return 0.0; if ( x > 90.0 ) return 1.0; if ( x <= 40.0 ) return 0.0 + ( 1.0 - 0.0 ) * ( ( x - 20.0 ) / ( 40.0 - 20.0 ) ); if ( x <= 90.0 ) return 1.0 + ( 1.0 - 1.0 ) * ( ( x - 40.0 ) / ( 90.0 - 40.0 ) ); } double FunctionBlock_tipper::membership_latency_medium(double x) {

112

if ( x <= 0.0 ) return 1.0; if ( x > 30.0 ) return 0.0; if ( x <= 20.0 ) return 1.0 + ( 1.0 - 1.0 ) * ( ( x - 0.0 ) / ( 20.0 - 0.0 ) ); if ( x <= 30.0 ) return 1.0 + ( 0.0 - 1.0 ) * ( ( x - 20.0 ) / ( 30.0 - 20.0 ) ); } double FunctionBlock_tipper::membership_latency_high(double x) { if ( x <= 0.0 ) return 1.0; if ( x > 30.0 ) return 0.0; if ( x <= 20.0 ) return 1.0 + ( 1.0 - 1.0 ) * ( ( x - 0.0 ) / ( 20.0 - 0.0 ) ); if ( x <= 30.0 ) return 1.0 + ( 0.0 - 1.0 ) * ( ( x - 20.0 ) / ( 30.0 - 20.0 ) ); double FunctionBlock_tipper::membership_throughput_low(double x) { if ( x <= 10.0 ) return 0.0; if ( x > 180.0 ) return 1.0; if ( x <= 30.0 ) return 0.0 + ( 1.0 - 0.0 ) * ( ( x - 10.0 ) / ( 30.0 - 10.0 ) ); if ( x <= 180.0 ) return 1.0 + ( 1.0 - 1.0 ) * ( ( x - 30.0 ) / ( 180.0 - 30.0 ) ); } double FunctionBlock_tipper::membership_throughput_medium(double x) { if ( x <= 0.0 ) return 1.0; if ( x > 20.0 ) return 0.0; if ( x <= 20.0 ) return 1.0 + ( 0.0 - 1.0 ) * ( ( x - 0.0 ) / ( 20.0 - 0.0 ) ); } double FunctionBlock_tipper::membership_throughput_high(double x) { if ( x <= 0.0 ) return 1.0; if ( x > 3.0 ) return 0.0; if ( x <= 3.0 ) return 1.0 + ( 0.0 - 1.0 ) * ( ( x - 0.0 ) / ( 3.0 - 0.0 ) ); } double FunctionBlock_tipper::membership_flow-weight_low(double x) { if ( x <= 15.0 ) return 0.0; if ( x > 25.0 ) return 0.0; if ( x <= 20.0 ) return 0.0 + ( 1.0 - 0.0 ) * ( ( x - 15.0 ) / ( 20.0 - 15.0 ) ); if ( x <= 25.0 ) return 1.0 + ( 0.0 - 1.0 ) * ( ( x - 20.0 ) / ( 25.0 - 20.0 ) ); } double FunctionBlock_tipper::membership_flow-weight_high(double x) { if ( x <= 70.0 ) return 0.0; if ( x > 100.0 ) return 0.0; if ( x <= 85.0 ) return 0.0 + ( 1.0 - 0.0 ) * ( ( x - 70.0 ) / ( 85.0 - 70.0 ) ); if ( x <= 100.0 ) return 1.0 + ( 0.0 - 1.0 ) * ( ( x - 85.0 ) / ( 100.0 - 85.0 ) ); } // Print void FunctionBlock_tipper::print() { B.1. 111 printf("Function block tipper:\n"); printf(" Input %20s : %f\n", "latency" ,latency); printf(" %20s : %f\n", "latency_low" , latency_low); printf(" %20s : %f\n", "latency_medium" , latency_medium); printf(" %20s : %f\n", "latency_high" , latency_high); printf(" Input %20s : %f\n", "throughput" , throughput); printf(" %20s : %f\n", "throughput_low" , throughput_low); printf(" %20s : %f\n", "throughput_medium" , throughput_medium);

113

printf(" %20s : %f\n", "throughput_high" , throughput_high); } // Reset output void FunctionBlock_tipper::reset() { for( int i=0 ; i < 1000 ; i++ ) { defuzzify_flow-weight[i] = 0.0; } } int main() { // Create function blocks FunctionBlock_tipper tipper; // Parse input tipper.latency = 90 ; tipper throughput = 90 ; // Calculate tipper.calc(); // Show results tipper.print(); }

chidume, chidiebere s.€¦ · chidume, chidiebere s. pg/m.eng/2013/65159 evolution (lte) downlink...

Documents