SENSOR ARRAY ASSISTED SPECTRUM SENSING AND …

SENSOR ARRAY ASSISTED SPECTRUM SENSING AND PERFORMANCE

OPTIMIZATION IN COGNITIVE RADIO NETWORKS

KIRAN SULTAN

A Thesis Submitted

in Partial Fulfillment of the Requirement

for the Degree of

Doctor of Philosophy

DEPARTMENT OF ELECTRICAL ENGINEERING

AIR UNIVERSITY

2013



Ph.D. Dissertation

SUBMITTED BY

KIRAN SULTAN

REG. NO. Ph.D.-EE-091315

SUPERVISOR

PROF. DR. IJAZ MANSOOR QURESHI


AIR UNIVERSITY

ISLAMABAD

December, 2013

CERTIFICATE

Department of Electrical Engineering

It is hereby certified that Kiran Sultan (Reg # Ph.D.-EE-091315) has successfully completed her

dissertation.

_____________________________

Dr. Ijaz Mansoor Qureshi Air University

Supervisor

____________________________ ____________________________ Dr. Fida Muhammad Khan Dr. Zafar Ali Shah

Internal Examiner 1 Internal Examiner 2 Guidance and Evaluation Committee Guidance and Evaluation Committee

____________________________ ____________________________

Dr. Abdul Jalil Dr. Noor Muhammad Khan External Examiner 1 External Examiner 2

Guidance and Evaluation Committee Guidance and Evaluation Committee

____________________________ ____________________________

Dr. Fida Muhammad Khan Dr. Zafarullah Koreshi Chair Department Dean



Ph.D. Dissertation

KIRAN SULTAN

REG. NO. Ph.D.-EE-091315

SUPERVISOR

PROF. DR. IJAZ MANSOOR QURESHI

FOREIGN RESEARCH EVALUATION EXPERTS

Prof. Dr. AJITH ABRAHAM, DIRECTOR, MIR LABS, USA

Prof. Dr. WEN HSIEN FANG, NTUST, Taiwan


AIR UNIVERSITY

ISLAMABAD

2013

ABSTRACT

Cognitive Radio has gained worldwide attention from research communities and is expected to

be a revolutionary technology for the next generation (4G) wireless systems. In this dissertation,

Amplify-and-Forward (AF) based relay-assisted cognitive radio networks (RCRNs) are studied

in an underlay spectrum sharing environment. The primary issue faced by underlay networks is

the limited transmit power ability of the secondary users (SUs) due to the interference constraints

towards the primary users (PUs), which reduces secondary throughput and allows only short-

range communication. Thus, performance enhancement of secondary communication in the

frequency bands allocated to the PUs is a major design challenge faced by the underlay RCRNs.

It requires relay selection along with the fine tuning and adjustment of the transmit power of the

secondary relays.

In this thesis, we proposed advanced multiple relay selection schemes for secondary network in

the Rayleigh flat-fading scenario considering the availability of perfect instantaneous channel

state information (CSI). The effects of variations in the instantaneous CSI, transmit power of

source and relays, interference threshold of the primary network, signal-to-noise ratio (SNR)

threshold of the secondary network and size of potential relay network on multiple relay

selection in underlay RCRNs are the main issues that are analyzed in depth in this research.

Furthermore, the performance analysis of multiple relay selection has been carried out and closed

form expressions for the outage probability and average probability of error have been derived

through the cumulative distributive function (CDF) of the received SNR at secondary

destination, which is a new contribution to the AF-based underlay RCRNs. The optimization tool

used in this study is Artificial Bee Colony (ABC) global optimizer.

Another novel idea proposed in this dissertation is Fuzzy Rule Based System (FRBS) for

multiple relay selection and transmit power allocation (RSTPA), which is a new contribution to

the underlay RCRNs. The proposed FRBS assisted RSTPA schemes aim to perform intelligent

multiple relay selection for performance enhancement of secondary communication in power

constrained RCRNs. It is proved through simulations that FRBS is an optimal choice to solve the

non-linear optimization problems of SNR maximization and transmit power minimization.

Another contribution of this research is in the field of spectrum sensing in CRNs. Spectrum

sensing faces a lot of challenges in terms of reliability and accuracy of information for detection

and estimation of primary transmissions in CRNs. The advantages and limitations of different

cooperative and non-cooperative spectrum sensing schemes have been studied in detail, and a

novel spectrum sensing scheme based on uniform linear array (ULA) of sensors is proposed,

which not only detects the number of sources, but also estimates their parameters such as

frequency, Direction-of-Arrival (DOA) and power strength. The effectiveness and reliability of

the proposed scheme is proved under low SNR conditions. Genetic Algorithm (GA) hybridized

with Pattern Search (PS) is used to optimize the results.

All the proposed algorithms have been investigated through simulations under different design

requirements, constraints and a well-defined range of different parameters to validate their

significance and effectiveness.

Copyright by

KIRAN SULTAN

2013

All rights reserved. No part of the material protected by this copyright notice may be reproduced

or utilized in any form or by any means, electronic or mechanical, including photocopying,

recording or by any information storage and retrieval system, without the permission from the

author.

DEDICATED TO

My Parents,

Brother and Sisters

i

CERTIFICATE OF APPROVAL It is certified that the research work contained in this Ph.D. dissertation has been carried out

under my supervision in the Department of Electrical Engineering, Air University, Islamabad. It

is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. It

has been set against the plagiarism and the report has been attached alongwith.

Signature: _____________________

Supervisor:

Prof. Dr. Ijaz Mansoor Qureshi Department of Electrical Engineering

Air University,

Islamabad.

ii

LIST OF PUBLICATIONS

List of Published Papers

1. Kiran Sultan, Ijaz Mansoor Qureshi, Aqdas Naveed Malik, Muhammad Zubair,

“Performance Analysis of Relay Subset Selection for Amplify-and-Forward Cognitive Relay

Networks”, The Scientific World Journal by Hindawi Publishing Corporation, 2013, (ISI

Indexed Journal with IF: 1.732).

2. Kiran Sultan, Ijaz Mansoor Qureshi, Muhammad Zubair, Aqdas Naveed Malik, “Power

Minimization through Relay Subset Selection in Underlay Cognitive Radio Networks”,

World Applied Sciences Journal, Vol. 23(5), 2013, pp. 714-717, (ISI Indexed Journal with

IF: 0.234).

3. Kiran Sultan, Ijaz Mansoor Qureshi, Muhammad Zubair, “Detection and Estimation of

Multiple far-field Primary Users using Sensor Array in Cognitive Radio Networks”, Journal

of Computing, Vol. 5(2), 2013, pp. 7-14, (ISI Indexed Journal with IF: 0.21).

4. Kiran Sultan, Ijaz Mansoor Qureshi, Bahman Ramzan Ali Alyaei, Ali Azad, “Performance

Enhancement of Secondary Communication through Multiple Relay Selection and Power

Allocation in Non-Regenerative Cognitive Radio Networks”, J. Basic Appl. Sci. (JBASR),

Vol. 3(10), 2013, pp. 416-420, (ISI Indexed Journal).

5. Kiran Sultan, Ijaz Mansoor Qureshi, Waseem Khan, Atta-ur-Rahman, “Performance

Enhancement of Secondary Network using Fuzzy Rule based System in Cognitive Relay

Networks”, European Journal of Scientific Research (EJSR), 2013, (ISI Indexed Journal).

6. Kiran Sultan, Ijaz Mansoor Qureshi, Muhammad Zubair, “SNR maximization through

Relay Selection in Cognitive Radio Networks”, Research Journal of Applied Sciences,

Engineering and Technology, 6(7), 2013, pp. 2616-2620, (ISI Indexed Journal).

7. Kiran Sultan, Ijaz Mansoor Qureshi, Muhammad Zubair, Babar Sultan, “ SNR

Maximization through CSI based Relay-Subset Selection in Amplify-and-Forward Cognitive

Radio Networks”, Presented in ICACELT, Abu Dhabi UAE, 2013, pp. 61-66.

8. Kiran Sultan, Ijaz Mansoor Qureshi, Muhammad Zubair, “SNR maximization through

Relay Selection and Power Allocation for Non-Regenerative Cognitive Radio Networks”,

Presented in INMIC IEEE, 2012, pp. 361-364.

iii

9. Kiran Sultan, Ijaz Mansoor Qureshi, Babar Sultan, “Performance Enhancement of

Secondary Communication in Underlay Cognitive Relay Networks”, accepted in INAAR,

IEEE, 2013.

10. Waseem Khan, I. M. Qureshi, Kiran Sultan, "Ambiguity Function of Phased-MIMO Radar

and its Properties", IEEE Geoscience and Remote Sensing Letters, Vol. PP(99), 2013, pp. 1-

5, (ISI Indexed Journal with IF: 1.823).

11. Ayesha Khaliq, Fawad Zaman, Kiran Sultan, Ijaz Mansoor Qurehsi, “3-D near field source

localization by using hybrid Genetic Algorithm”, Research Journal of Applied Sciences,

Engineering and Technology, Vol. 6(23),2013. (ISI Indexed Journal).

12. Shahid H. Abbassi, I. M. Qureshi, Bahman R. Alyaei, Hameer Abbasi, Kiran Sultan, “An

Efficient Spectrum Sensing Mechanism for CR-VANETs”, J. Basic Appl. Sci., (JBASR),Vol.

3(12), 2013, pp. 365-378, (ISI Indexed Journal).

13. Habibullah Jamal, Kiran Sultan, “Performance Analysis of Loss-Based High-Speed TCP

Congestion Control Algorithms”, WSEAS International Conference, Ningbo, China, 2008.

14. Habibullah Jamal, Kiran Sultan, “Performance Analysis of TCP Congestion Control

Algorithms”, International Journal of Computers and Communications, Vol. 2(1), 2008.

List of Submitted Papers

1. Kiran Sultan, Ijaz Mansoor Qureshi, Atta-ur-Rahman, and Shahid Hussain Abbassi,

“Transmit Power Minimization in Cognitive Relay Networks using Fuzzy Rule Based

System”, submitted in Journal of Intelligent and Fuzzy Systems.

2. Kiran Sultan, Ijaz Mansoor Qureshi, Atta-ur-Rahman, Waseem Khan, “SNR Maximization

using Fuzzy Rule Base System in Relay Assisted Cognitive Radio Networks”, submitted in

Journal of Multiple-valued Logic and Soft Computing.

3. Kiran Sultan, Ijaz Mansoor Qureshi, Muhammad Zubair, Aqdas Naveed Malik, “Artificial

Bee Colony Optimization for Relay Selection with SNR Maximization in Underlay

Cognitive Radio Networks”, submitted in Iranian Journal of Science and Technology,

Transactions of Electrical Engineering.

4. Waseem Khan, I. M. Qureshi, Kiran Sultan, "Ambiguity Function of Frequency-Diverse-

Array Radar and its Properties", submitted in WASJ (ISI Indexed Journal with IF 0.234).

The material presented in this dissertation is based on the published papers 1 to 8 and submitted

papers 1 to 3.

iv

ACKNOWLEDGEMENTS

I am thankful to Almighty Allah for his guidance at each step of this work through the

blessings of eeman, health, supporting family, brilliant teachers, cooperative friends and

colleagues. Peace and prayer for marvelous human and the torch bearer of wisdom Muhammad

(Peace be upon him), who’s objective was enlightenment of the whole world.

I owe thanks to many people for where I have arrived today.

First and foremost, I offer my special and sincere thanks to my supervisor Dr. Ijaz

Mansoor Qureshi for his direction, motivational guidance and support that remained with me

throughout the research work. He is a man of inducing brain waves and initiating the sparks of

good ideas in my mind. He helped me with building up a strong foundation for my future

scholarly career. Despite his hyper dimensional commitments, he never let me relax. I would like

to thank him for giving me space and freedom to glide and explore the topics I felt more

comfortable with. I also thank him for listening to me with patience and tolerance. He is one of

the best samples of a good supervisor.

I am indebted to Dr. Kamal Athar for his encouragement and administrative support. He

always stands up for us compassionately, and looks for great opportunities for us. I owe to

express my thanks to Dr. Fida Muhammad Khan for his high cooperation and strong

administrative support from time to time. He always tries to make things smooth and manageable

for us. I am grateful to Dr. Atta-ur-Rahman for his fruitful advices and profound comments on

my research work. I found him eager to help me in my research. I also thank Dr. Zafar Ali Shah

for all I learned from him that I have to be organized, precise and smart to be a good academic

scholar.

I feel very fortunate to be given an opportunity to have the company of a collection of

awesome people in the Department of Electrical Engineering, AU. I like to thank my friends,

colleagues and students for being cooperative during my Ph.D. period. They are all the most

caring and smart people I have ever had the privilege to know. My deepest thanks to Mr.

Bahman Ramzan Ali Alyaei, whom one can never stop learning from. It is really hard to count

many scholars as smart and humble as him. I cannot thank him in words for his enormously

helping and directing comments on my research. I am really thankful to Ms. Sundas Amin from

v

depth of my heart for being so nice, sincere and helpful. I would like to express my gratitude to

Mr. Shahid Abbasi for his encouragement and moral support.

At last I express my immense gratitude and respect to my lovely parents and family who

never stopped supporting, encouraging and standing by me during the ups and downs of my life,

and love to see me flourishing and progressing. I am thankful to my brother Capt. Babar for

providing me the educational and moral support during the whole Ph.D. and setting up goals for

me. I dedicate this thesis, which is the outcome of my life to my family to whom I owe every

single success in my life.

(Kiran Sultan)

vi

TABLE OF CONTENTS

LIST OF TABLES x

LIST OF FIGURES xi

LIST OF ABBREVIATIONS xiii

Chapter 1 1

Introduction

1.1 Background 1

1.2 The System Model 4

1.3 Objectives and Contributions 7

1.4 Organization of the Thesis 8

Chapter 2 11

Cognitive Radios and Cognitive Relay Networks

2.1 History of Cognitive Radios 11

2.2 Spectrum Sharing in Cognitive Radios 15

2.2.1 Underlay Spectrum Sharing 16

2.2.2 Overlay Spectrum Sharing 16

2.2.3 Interweave Spectrum Sharing 16

2.3 Cooperative Communication 17

2.3.1 Relay Protocols 18

2.3.1.1 Amplify-and-Forward Relaying 19

2.3.1.2 Decode-and-Forward Relaying 19

2.4 Cognitive Relay Networks 20

2.4.1 Relay Selection in Underlay Cognitive Relay Networks 21

2.4.1.1 Best Relay Selection Schemes Proposed in Literature 22

2.4.1.2 Multiple Relay Selection Schemes Proposed in Literature 23

2.4.1.3 Observations in the Literature Review of Relay Selection 23

Schemes

2.5 Concluding Remarks 24

Chapter 3 25

Nature Inspired Algorithms and Fuzzy Logic

3.1 The Evolutionary Algorithms 25

3.2 Artificial Bee Colony (ABC) Optimization 27

3.2.1 Optimization Phases of ABC 28

3.2.1.1 Initialization Phase 28

3.2.1.2 Best Solution Search Phase 28

3.2.2 Flow Chart of ABC Optimization 29

3.3 Genetic Algorithm 30

vii

3.3.1 Optimization Phases in GA 30

3.3.2 Flow Chart of GA Optimization 31

3.3.3 Applications of GA 32

3.4 Fuzzy Logic 32

3.4.1 History of Fuzzy Logic 32

3.4.2 Fuzzy Control System 33

3.4.3 Applications of Fuzzy Logic 34

3.4.4 Fuzzy Logic in CRNs 34


Chapter 4 36

SNR Maximization in Underlay Networks

4.1 Problem Formulation 1 36

4.2 Proposed Algorithms 37

4.2.1 Proposed Algorithm 1 37

4.2.1.1 ABC Optimization 38

4.2.1.2 Simulation Results 40

4.2.2 Proposed Algorithm 2 41


4.3 Comparison of the Proposed Algorithms 43

4.3.1 Problem Formulation II 43

4.4 Problem Formulation 2 45

4.5 Proposed Algorithms 46

4.5.1 Proposed Algorithm I 45


4.5.1.2 Concluding Remarks 48

4.5.2 Proposed Algorithm II 48



4.5.3 Algorithm III 51



4.6 Comparison of the Proposed Algorithms 55


Chapter 5 58

Outage Analysis of Multiple Relay Selection

5.1 Problem Formulation 58

5.2 The Proposed Algorithm 60

5.3 Performance Analysis 63

5.3.1 Multiple Relay Selection 63

5.3.2 Best Relay Selection 67

5.4 Simulation Results 70


viii

Chapter 6 74

Transmit Power Minimization in Underlay CRNs

6.1 Problem Formulation 74

6.2 The Proposed Algorithm 75

6.3 Simulation Results 77


Chapter 7 80

Performance Enhancement of CRNs using Fuzzy Rule Based System

7.1 SNR Maximization 80

7.2 FRBS Assisted System Design 1 81

7.2.1 Mamdani Fuzzy Control 82

7.2.1.1 Fuzzificaion 82

7.2.1.2 Rule Based Decision 85

7.2.1.3 Defuzzifier 85

7.2.2 The Proposed Algorithm 87

7.2.3 Simulation Results 88

7.2.4 Concluding Remarks 91

7.3 FRBS Assisted System Design 2 91


7.3.1.1 Fuzzification 93





7.4 Comparisons of the Proposed Algorithms 101


7.5 Transmit Power Minimization 102

7.5.1 The Proposed FLS Design 103


7.5.2.1 Fuzzification 104






Chapter 8 111

Detection and Estimation of Multiple Far-Field Primary Users using Sensor Array

8.1 Background 111

8.2 Spectrum Sensing Methods 112

ix

8.2.1 Non-Cooperative Spectrum Sensing 112

8.2.1.1 Energy Detector 112

8.2.1.2 Matched Filer Detection 112

8.2.1.3 Cyclostationary based Detection 113

8.2.2 Interference based Spectrum Sensing 113

8.2.3 Cooperative Spectrum Sensing 113

8.3 Source Localization 113

8.4 Contribution of Thesis 114

8.5 System Model and Problem Formulation 115

8.6 Proposed Algorithm for Detection of PUs 117

8.7 Simulation Results and Discussions 119

8.8 Conclusion 127

Chapter 9 128

Conclusions and Future Work

9.1 Conclusions 128

9.2 Future Work 129

References 131

x

LIST OF TABLES

Table 4.1 Pseudocode of the Proposed Algorithm 38

Table 4.2 ABC Optimization to solve SNR Maximization Problems 39

Table 4.3 The Parameter Settings 40

Table 4.4 Pseudo Code For Proposed Algorithm 2 (Power Set Algorithm) 42

Table 4.5 Comparison Of The Proposed Algorithms 44

Table 4.6 Psuedocode for Algorithm I 46

Table 4.7 No. of Selected Relays Obtained From Proposed Algorithm 1 48

Table 4.8 Pseudo Code for Algorithm II 49

Table 4.9 No. Of Selected Relays From Proposed Algorithm 2 50

Table 4.10 The Pseudo Code For The Proposed Algorithm 53

Table 4.11 No. Of Selected Relays From Proposed Algorithm 3 55

Table 4.12 Comparison Of The Proposed Algorithms 56

Table 5.1 Pseudocode for the Proposed Relay Subset Selection Algorithm 61

Table 5.2 Performance Analysis Of Best Relay, Multiple Relay and All Relays

Participation Schemes 71

Table 6.1 The Proposed Algorithm 76

Table 6.2 Total Transmit Power required For Different Values of 78

Table 6.3 Transmit Power Allocation To Relay Network For 79

Table 7.1 Corresponding Number Of Selected Relays 89

Table 7.2 Total No. of Selected Relays 101

Table 7.3 Comparisons of Proposed Algorithms in Chp 4 and Chp 7 101

Table 8.1 Parameter Settings for GA-PS 119

Table 8.2 Pseudocode of the Proposed Algorithm for Detection of Number 120

of Sources

Table 8.3 Amplitude, DOA and frequency estimation for different SNR 123

levels with M = 2, L = 20

Table 8.4 Amplitude, DOA and frequency estimation for different SNR levels 125

with M = 4, L = 25

Table 8.5 Amplitude, DOA and frequency estimation for different SNR levels 126

and different number of sensors in the array with M = 2

MandI thth ,,dBdBI thth 0,0

xi

LIST OF FIGURES

Fig. 1.1 Strict allocations of frequency bands 1

Fig. 1.2 The System Model 4

Fig. 2.1 FCC Spectrum Allocation Chart 12

Fig. 2.2 The concept of Spectrum Hole 14

Fig. 2.3 Multiple Input Multiple Output (MIMO) System 17

Fig. 2.4 AF and DF Protocols 20

Fig. 2.5 Conceptual Model for Relay(s) Selection 21

Fig. 3.1 Flow Chart for Artificial Bee Colony Optimization 29

Fig. 3.2 Flow Chart for GA Optimization 31

Fig. 3.3 Fuzzy Control System 33

Fig. 4.1 Performance Analysis of Algorithm 1 41

Fig. 4.2: Performance Analysis Of Proposed Algorithm 2 43

Fig. 4.3: Comparison Of The Proposed Algorithms 44



Fig. 4.6 Performance Analysis Of Proposed Algorithm 3 54

Fig. 4.7 Performance Analysis Of Proposed Algorithms 56

Fig. 5.1: Flowchart of the Proposed Algorithm 62

Fig 5.2: Performance Analysis Of Different Schemes 71

Fig. 5.3: Outage Behavior of Best and Multiple Relay Selection Schemes 72

Fig. 5.4: BER of Best and Proposed Multiple Relay Selection 73

Fig. 6.1: Transmit Power Allocation to Relay Network keeping dBI th 0 77

Fig. 6.2: Transmit Power Allocation to Relay Network for 79

Fig. 7.1 The Proposed System Design 1 81

Fig. 7.2 MFs Of The Antecedents And The Consequents Of FLS 1 And FLS 2 84

Fig 7.3(a) Rule Surface for FLS 1 86

Fig. 7.3(b) Rule Surface for FLS 2 86

Fig. 7.4 The Flow Chart of the Proposed Fuzzy Rule Based RSTPA Design 87

Fig. 7.5 SNR Performance of Proposed Scheme 88

Fig. 7.6 SNR Performance For Different Source Transmit Power Levels 89

Fig. 7.7 Comparison of the Proposed Scheme and the Greedy Scheme 90

Fig. 7.8 Proposed System Design 2 91

Fig. 7.9 Proposed FLS Modules 92

Fig. 7.10(a) MFs For Antecedents And Consequents of FLS1 94

Fig. 7.10(b) MFs For Antecedents And Consequents of FLS1 96

Fig. 7.11(a) Rule Surface for FLS 1 97

Fig. 7.11(b) Rule Surface for FLS 2 98

Fig. 7.11(c) Rule Surface for FLS 3 98

Fig. 7.12 Flow Chart of the Proposed Algorithm 99

Fig. 7.13 Performance Of The Proposed Scheme Vs Interference Threshold thI 100

Fig. 7.14 Proposed FRBS 103

dBdBI thth 0,0

xii

Fig. 7.15 Fuzzy Sets for the Antecedents and the Consequent 104

Fig. 7.16 The Rule Surface 105

Fig. 7.17 The Flow Chart of the Proposed Algorithm 106

Fig. 7.18(a) Total Transmit Power Vs Interference Threshold thI for 1th 107

Fig. 7.18(b) Corresponding Number Of Selected Relays 108

Fig. 7.19 Total Transmit Power Vs SNR Threshold th 108

Fig. 7.20 Total Transmit Power Of For Different Source Transmit Power Levels 109

Keeping dBI th 10 And 1th

Fig.8.1. The System Model 115

Fig 8.2(a) Detection of M = 2 PUs 121

Fig. 8.2(b) Error in DOA vs SNR for M = 2, L = 20 122

Fig. 8.2(c) Error in frequency vs SNR for M = 2, L = 20 122

Fig. 8.3(a) Detection of M = 4 PUs 123

Fig. 8.3(b) Error in DOA vs SNR for M = 4, L = 25 124

Fig. 8.3(c) Error in frequency vs SNR for M = 4, L = 25 124

Fig. 8.4(a) Error in DOA estimation for different SNR levels and different number 125

of sensors in the array considering M = 2

Fig. 8.4(b) Error in frequency estimation for different SNR levels and different 126

number of sensors in the array considering M = 2

xiii

LIST OF ABBREVIATIONS

Adaptive Fuzzy Logic AFL

Additive White Gaussian Noise AWGN

Amplify-and-Forward AF

Ant Colony Optimization ACP

Artificial Bee Colony ABC

Artificial Immune Optimization AIP

Artificial Intelligence AI

Artificial Neural Networks ANN

Binary Phase Shift Keying BPSK

Bit Error Rate BER

Channel State Information CSI

Cognitive Radio CR

Cognitive Radio Network CRN

Compress-and-Forward CF

Computer Added Design CAD

Consider for Selection CS

Cumulative Distributive Function CDF

Cyclostationary Detection CD

Decode-and-Forward DF

Differential Evolution DE

Direction of Arrival DOA

Dynamic Spectrum Access DSA

Employed Bee EB

Energy Detection ED

Evolutionary Algorithm EA

Expert System ES

Fuzzy Control system FCS

Fuzzy Inference Engine FIE

xiv

Fuzzy logic FL

Fuzzy logic system FLS

Fuzzy Rule Based System FRBS

Genetic Algorithm GA

Gradient Search GS

Hybrid Evolutionary Algorithm HEA

Independent and Identically distributed IID

Interior Point Algorithm IPA

Invasive Weed Optimization IWO

Matched Filter MF

Maximum Eigenvalue Detection MED

Maximum Eigenvalue to Trace MET

Maximum Minimum Eigenvalue MME

Membership Function MF

Memetic Particle Swarm Optimization MPSO

Multiple-Input Multiple-output MIMO

Not Selected NS

Onlooker Bee OB

Particle Swarm Optimization PSO

Pattern Search PS

Primary User PU

Probability Distributive Function PDF

Quality of Service QOS

Relay Selection Factor RSF

Relay Selection and Transmit Power allocation RSTPA

Scout Bee SB

Secondary User SU

Selected S

Sequential Quadratic Programming SQP

Signal-to-Noise Ratio SNR

Simulated Annealing SA

xv

Software Defined Radio SDR

Spectrum Sensing SS

Strong Consideration for Selection SCS

Time Division Multiple Access TDMA

Ultra-Wideband UWB

Uniform Linear Array ULA

Weak Consideration for Selection WCS

1

Chapter 1

INTRODUCTION

1.1 BACKGROUND

Conventionally, spectrum regulatory bodies follow highly inflexible and authoritative approach

in specifying the services for a particular band of frequencies and the permitted technologies to

deliver those services. Such authorized or licensed users of the spectrum are known as primary

users (PUs) [1]. The strict policies for the use of spectrum are very effective to manage the

interference as depicted in Fig. 1.1. The guard bands in the figure ensure that the neighboring

services do not interfere each other’s transmissions [2],[3]. However, as a consequence of this

“Command and Control” strategy of spectrum management, some bands are heavily loaded in

vast temporal and geographical locations, whereas, a large number of frequency bands are highly

underutilized.

Fig. 1.1: Strict Allocations Of Frequency Bands

2

Thus, broadly speaking, the spectrum bands are classified into black spaces (highly occupied

frequency bands), grey spaces (partially occupied frequency bands) and white spaces (vacant

frequency bands except for ambient noise) [3]. Joseph Mitola, pioneer of cognitive radio (CR)

technology, innovated the concept of CR, for the first time in the history of wireless

communication in 1999 [4]. Mitola defined CR as a smart radio which is “self-aware” and

“alert”. A CR is expected to operate with a clear understanding of its operating environment, the

communication requirements of other user(s), spectrum regulatory policies and its own

capabilities [2].

The cognitive users, commonly known as secondary users (SUs) in CR terminology, utilize the

licensed band using overlay, underlay and interweave approaches [5]. In the overlay spectrum

sharing, the SU uses specialized signal processing techniques for the performance enhancement

of primary transmissions, while transmitting concurrently in the frequency band of PU. In the

underlay mode, the SU is allowed to transmit in the frequency band assigned to the PU, as long

as the interference offered to the PU by secondary transmissions is below the interference

threshold of the PU. In the interweave approach, the SU looks for the spectrum opportunities

where the PU is currently absent and transmit with full power in the detected spectrum holes.

These spectrum sharing modes will be discussed in more detail in the next chapter.

CR technology has gained world-wide attention from the research communities as a potential

candidate for future wireless world. Being an emerging technology, CR faces a lot of challenges

to replace the currently deployed wireless communication systems. The hot areas of research in

CR are spectrum sensing algorithms, cooperative communication, CR architecture, dynamic

spectrum access, security issues, dynamic resource management, and development of adaptive

algorithms [6]-[7]. A lot of development has been done in designing the individual components

of CRs and building protocol architecture of cognitive radio networks (CRNs), but still the

integration of these parts for large scale deployment of CRNs is a major area of research.

Spectrum sensing (SS), aims to obtain information about the local spectrum, and is the key

enabling technology for the establishment of CR. CRs are built on the ability to sense the radio

spectrum and gain knowledge about the available frequency bands, transmit powers and

3

direction-of-arrival (DOA) of the active users, available networks, spectrum management

policies regarding the use of detected spectrum and other operating restrictions. Various features

of spectrum sensing are shown in figure [8].

Earlier spectrum sensing techniques mainly focused on detecting the underutilized or vacant

bands of spectrum, however, with new challenges and dimensions in CRNs, sensing frequency

only may not be enough [9]. Thus it requires exploration of new dimensions of direction of

arrival (DOA), frequency, strength of signal, range and a critical parameter which is the number

of active PUs. In order to ensure secure, reliable and efficient communication keeping in view

the privilege of PUs, advanced SS algorithms capable of identifying occupancy in all of the

above dimensions of spectrum space to locate spectrum holes need to be developed, which have

not been considered simultaneously in CRNs yet according to the best of our knowledge.

The concept of cooperative diversity is based on the idea of introducing multiple nodes between

source-destination pair in such a way that each intermediate node listens to the signal transmitted

by the source. These partners are known as “relays” in wireless communication terminology

[10]. The relays cooperate with each other and behave like a virtual array of transmit antennas to

facilitate the source-destination pair in their communication even in worst-case scenarios when

direct communication is not possible between them due to deep fading, shadowing etc. The

whole study considers dual-hop one-way relay network, in which each relay is equipped with a

single antenna and one complete data transmission occurs in two time-slots, to be explained in

detail in the later chapters. Relaying efficiently improves system throughput, combats channel

fading effect, reduces power consumption, increases transmission reliability, and extends

coverage area [11].

Cooperative diversity techniques have been extensively utilized in the CRNs. Cooperation

between SUs [6]-[7], and cooperation between PUs and SUs [8] are two approaches followed for

this purpose. Cognitive relay networks, inspired by CR and relay networks, exist as a versatile

choice to assist SUs and extend their coverage area, but employing all relays in a power-

restricted environment may cause high interference to the concurrent primary transmissions.

4

Relay selection has been observed as a fascinating technique to solve the aforementioned

complex interference-mitigation issue faced by multi-relay networks. In this context, several best

and multiple relay selection schemes have been proposed, with relatively less contributions

observed in the area of multiple relay selection. Furthermore, a lot of effort has been done to

investigate the performance of best relay selection in terms of outage probability and bit error

rate, but no one has carried out derivation for outage probability and bit error rate in closed form

for multiple relay selection in AF based underlay networks.

1.2 THE SYSTEM MODEL

Fig. 1.2. illustrates the cooperative dual-hop CRN comprising M randomly distributed cognitive

relays, a source S , and a destination D . The whole relay network operates in the vicinity of a PU

Q . Each node in the network is equipped with a single antenna, thus simultaneous transmission

and reception is not possible. Rayleigh flat fading is assumed for the whole scenario, in which

M

mmg1 , M

mmh1 and M

mmf 1 denote the independent and identically distributed (i.i.d.) channel

coefficients between source-relay, relay-destination and relay-PU respectively.

Fig. 1.2: The System Model

S

D

R1

R2

2

R3

RM

Q

gm

hm

fm

Source

Destination

PU

Relay Network

5

Some assumptions made for the whole study are as follows. 1) Instantaneous channel state

information (CSI) is available at each secondary node. Moreover, the relays have perfect

knowledge of their forward and backward channels. 2) Binary Phase Shift Keying (BPSK)

modulation is used, hence, the symbol SS P,Ps , where SP represents the symbol power

transmitted by the source. 3) The relays operate in half-duplex mode, so they are unable to

transmit and receive on the same frequency simultaneously. 4) Amplify-and-Forward (AF)

relaying is assumed at the relay network. 4) Line-of-sight path suffers from deep fading, thus

making direct communication between source-destination pair impossible. 5) Underlay spectrum

sharing model is assumed at the relay network, thus the secondary communication can only take

place if the total interference offered to the PU by the potential relay network remains below a

predefined threshold thI , which is the maximum tolerable interference level for the PU. 6)

Additive white Gaussian noise (AWGN) with zero mean unit variance is assumed for each hop.

Based on the third assumption, one data transmission is completed in two time-slots. The source

transmits a symbol s in time slot 1 and the received signal my at the thm relay is given as:

mmsm sgPy 1 )1.1(

where, )1,0(~1 Nm is modeled as AWGN at the thm relay. In time slot 2, the destination

receives the scaled version of the received message from the relay network while the source is

silent. The signal Dy received at the destination D is expressed as:

DmmD hyy )2.1(

where, )1,0(~ ND represents AWGN with variance 0N , received at the destination. The signal

my amplified according to AF scheme is given as:

m

ms

mm y

NgP

Py 1

0

2|| )3.1(

6

where, mP represents the transmit power of thm relay in the above eq. adjusted via AF protocol

and max0 PPm . maxP is the maximum transmit power each thm relay is allowed to transmit. The

transmission power mP of each relay is limited not only by the battery capacity due to regulations

specifying the maximum power that each node is allowed transmit, but also by the interference

threshold of the PU.

Substituting (1.1) and (1.3) in (1.2) and solving the resulting expression, end-to-end signal-to-

noise ratio (SNR) m of the thm relay link can be expressed as [12]-[14],

0

2

0

20

2

0

2

||||1

||||

N

hP

N

gP

N

hP

N

gP

mmms

mmms

m

Or in compact form,

mm

mmm

21

21

1

)4.1(

where, 0

2

1

||

N

gP msm and

0

2

2

||

N

hP mmm denote the instantaneous SNR achieved at the

source-relay and relay-destination links respectively.

The total instantaneous end-to-end SNR D at the secondary destination due to M relaying

links is then given by [15],

M

m mm

mmM

m

mD1 21

21

1 1

)5.1(

In underlay networks, sophisticated signal processing techniques are employed to mitigate the

interference offered to the PU. However, due to the inherent simplicity of AF protocol, such

7

computationally complex techniques may not be supported at the AF relays. Thus, enabling the

secondary communication exploiting the services of all relays in the potential relay set may not

be a viable idea in terms of total interference offered by the relay network to the nearby PU.

Alternatively, suppressing the transmit power of candidate relays reduces the interference power

but at the same time makes it difficult to enable the secondary communication with minimum

QoS requirements. For simultaneous primary and secondary transmissions in such energy-

constrained environment, the total interference power experienced by the PU due to the

transmissions of relay network must satisfy the predefined interference threshold given as,

th

M

m

mm

M

m

m IfPII 1

2

1

|| )6.1(

where thI is the interference threshold set by the PU.

Relay selection stands as a fascinating solution to this problem. Therefore, relay selection is

performed to choose the best combination of relays that maximizes the SNR achieved at the

destination keeping in view the privilege of the PU. Thus eq. (1.5) takes the form,

SS

Dm mm

mm

m

m21

21

1

)7.1(

where, S denotes the selected subset of relays.

This dissertation proposed multiple relay selection and power allocation schemes in this

framework to select the best subset of relays S which meets the objectives and constraints.

1.3 OBJECTIVES AND CONTRIBUTIONS

This dissertation highlights several issues in CR technology. The objectives of this dissertation

are to study the SS issues in CRs, and the design issues in AF based cognitive relay networks

operating in an underlay mode of spectrum sharing, aiming to highlight the deficiencies found in

the literature. For this purpose, a detailed study of best and multiple relay selection schemes and

8

the existing SS techniques has been carried out, paying special attention on the scenarios, in

which line-of-sight path between secondary source-destination pair undergoes deep fading, thus

making direct communication impossible. The main contributions of this dissertation are

summarized as follows:

Various multiple relay selection schemes have been proposed in the setups of Rayleigh

flat-fading scenario assuming availability of perfect instantaneous CSI. The effects of

variations in the instantaneous CSI, transmit powers of source and relays, interference

threshold of the primary network, SNR threshold of the secondary network and size of

potential relay network on multiple relay selection in underlay cognitive relay networks

are the main subjects that are studied in this research.

Outage behavior of secondary network for multiple relay selection is investigated and

closed-form expressions for outage probability and bit-error rate are derived through the

CDF of the SNR received at the destination.

FRBS assisted multiple relay selection and transmit power allocation schemes are

proposed aiming to enhance secondary performance, which is another new contribution

to underlay cognitive relay networks.

A novel idea of SS is proposed for CRNs. The proposed scheme not only detects the

number of active PUs, but also provides the estimates of their parameters such as

frequency, power strength and Direction-of-Arrival (DOA) upto high accuracy.

The performance of each proposed scheme has been evaluated under different design

requirements, assumptions and constraints.

1.4 ORGANIZATION OF DISSERTATION

The dissertation has been written in a manuscript style. The contributions in the form of

published and submitted manuscripts are included as the central body of the dissertation.

Footnotes are also added for clarification where necessary. The chapter wise distribution of the

dissertation is organized as follows:

In Chapter 2, current spectrum management policy is introduced and the fundamental knowledge

of CR starting from history to the current research challenges is discussed. After going through

9

the comprehensive survey of this emerging technology, the discussion proceeds towards

cooperative communication and develop the understanding of the main concepts involved in

cognitive relay networks. Finally, some best and multiple relay selection schemes proposed for

the interference-constrained cognitive relay networks are surveyed and their findings are

highlighted. The system model and the basic assumptions made for this study are also explained.

Chapter 3 discusses the tools used in the performance evaluation of different schemes. A review

of the nature inspired evolutionary algorithms and artificial intelligence is carried out, aiming to

understand their advantages, properties, limitations, and applications. In this context, the phases

of Artificial Bee Colony optimization, Genetic Algorithm and Fuzzy Logic are further discussed

in depth with the aid of flow charts and block diagram where necessary, since these tools are

employed to solve relay selection and spectrum sensing problems in the dissertation.

Chapter 4 highlights the contributions in the area of multiple relay selection, aiming to maximize

the secondary performance under strict interference constraints imposed on the cognitive relay

network. The effect of individual relay’s transmit power constraint, entire relay network’s

transmit power constraint and interference threshold levels are studied for in-depth analysis.

In chapter 5, a problem of transmit power minimization is formulated for cognitive relay network

under same design assumptions. A multiple relay selection scheme is proposed in this regard,

aiming to minimize the total power consumed at the relay network, while satisfying minimum

quality-of service (QoS) requirements of both primary and secondary networks.

Chapter 6 highlights the deficiencies found in literature regarding the performance analysis of

multiple relay selection in underlay networks, and derives the closed form expressions of the

outage probability and bit error rate of the SNR received at the destination, which is another new

contribution to the cognitive relay networks operating in an underlay spectrum sharing

environment.

Chapter 7 presents a novel idea of using FRBS assisted intelligent relay selection schemes for

relay-assisted CRNs. The proposed schemes takes the CSI of each candidate relay in the

10

potential relay network as an input and assigns relay selection factor (RSF) to each relay. The

RSF eventually sets the precedence in which the relays are selected, aiming to enable the

coexistence of the primary and the secondary networks.

The last topic that is investigated in chapter 8 is SS. In order to preserve the PUs’ rights of

interference-free operation, the SUs are required to sense the licensed bands at regular intervals,

and reliably detect the primary signals. A novel idea of uniform linear array (ULA) based SS is

proposed, which not only detects the number of active PUs, but also provides the estimates of

amplitude, frequency and DOA of the active users upto high accuracy.

In chapter 9, the whole research work is concluded and a comprehensive summary is provided.

We also put forward some future directions to extend the proposed techniques.

11

Chapter 2

COGNITIVE RADIOS AND COGNITIVE

RELAY NETWORKS

Objective

This chapter discusses the current spectrum management policies and the motivation behind the

CR technology. After a comprehensive review of the history of the proliferating CR technology,

the research challenges and the deficiencies in designing the CRNs are discussed. Furthermore,

cognitive relay networks are included in the discussion and different single and multiple relay

selection schemes are surveyed and their findings are highlighted for power-constrained CRNs.

2.1 HISTORY OF COGNITIVE RADIOS

The wireless applications are proliferating very rapidly and large variety of communication

systems exist for different applications in the licensed and unlicensed frequency bands [16]. The

global wireless communication standards include personal area networks (IEEE 802.15), local

area networks (IEEE 802.11), metropolitan area networks (IEEE 802.16) and wide area networks

(IEEE 802.20). The flourishing wireless technology has urged the deployment of mesh networks.

A careful survey shows that up to 1 trillion wireless devices are expected to be operational by

2020. Moreover, due to non-line-of-sight propagation, radio propagation favors the use of

spectrum below 3GHz in the entire radio spectrum ranging from 3MHz to 300GHz. According to

current spectrum regulatory framework, the licensed frequency bands have been exclusively

allocated to the specific services [17]. The stringent spectrum assignment used by Federal

Communications Commission (FCC) is provided in Fig. 2.1.

12

Fig. 2.1: FCC Spectrum Allocation Chart

This “command and control” strategy of spectrum allocation and management effectively

protects the authorized users from unwanted interference of other users of the radio spectrum on

one hand, and does not allow the users to switch between highly occupied and underutilized

bands of frequencies on the other hand [18]. This ever increasing growth of wireless services,

huge demand of internet access, evolution of smart phones and spectrum analysis demand more

and more spectrum resources, which creates a scenario of spectrum scarcity. The need of

continuous and fast technological evolution creates the demand of new dedicated spectrum

bands. Another strong observation revealed through the careful analysis of FCC in 2002 is the

underutilization of most of the licensed spectrum bands allocated to the current wireless

subscribers, in temporal and spatial domains [19]. FCC reports show that the variation in the

licensed spectrum occupancy ranges from 15% to 85% [20]. This underutilization stems from the

existing fixed spectrum allocation strategies for the valuable resource as mentioned above. For

example in US, a frequency band of 512-608 MHz is dedicated for television broadcasting for

channels 21-36, while the frequencies from 960-1215 MHz are reserved for radio navigation

[21]. Spectrum management is one of the major responsibilities of the communication regulators

and more efficient spectrum utilization can result by better spectrum management.

13

Earlier in 1993, Joseph Mitola III proposed a novel idea of dynamic spectrum sharing of the

spectrum bands exclusively allocated to the licensed users using novel devices called Cognitive

Radios [22]. Mitola envisioned CR as a combination of existing wireless technology and

artificial intelligence. The need for CRs is primarily motivated by the complexity of radio

systems itself. Conventional radio design is targeted for single purpose and single environment.

The key enabling technology for cognitive wireless networks is the software defined radio (SDR)

which first emerged in 1990 [23]. SDR aims to bring radio electronics into the digital age, thus

adding new degrees of freedom in designing wireless networks by enabling radios to adapt to the

requirements at hand [24]. These radios perform signal processing in software, thus enabling the

devices that can be reconfigured via software after deployment. SDRs find their applications in

industry, academia, government and military organizations, communication research, data

acquisition and many more. However, the radios built on SDR technology are expensive, since

they support multiple interface technologies e.g. GSM, CDMA with a single modem by

reconfiguring it in software. Thus, a CR is an SDR, that is fully programmable to interact with its

operating environment and dynamically adapts its parameters i.e. carrier frequency, modulation

technique, transmit power, channel access method and networking protocols to deliver the best

application performance. However, the fundamental differences between CRs and SRs are as

follows [25]. First, contrary to CR technology, the SDRs do not have the ability to sense and

detect the unoccupied or partially occupied slots in the spectrum. Second, CRs are capable of

operating at any frequency in the entire radio spectrum, whereas, SDRs are designed for certain

standards and their assigned frequency bands. Third, SDRs are built on the availability of a priori

knowledge of the interfering channels, whereas, the CRs have the ability to tolerate interference

at any frequency in the bandwidth defined for cognitive wireless devices.

Based on the above facts and requirements, FCC took the initiative in 2004 by allowing the

unlicensed users to utilize television spectrum in the regions, where spectrum is not in use.

Broadly speaking, the term CR refers to various solutions of the spectrum underutilization

problem by enabling transmissions from the unlicensed wireless devices in the underutilized

licensed frequency bands, in such a way that the licensed users are as uninterrupted as possible.

Thus, the CR devices are capable of detecting the spectrum holes1 and dynamically and

1Spectrum hole is a band of frequencies primarily assigned to PU, but at a specific time and location, is not in use by the PU

14

autonomously adapt themselves to the changing network conditions to deliver the best possible

performance. The networks of such nodes are recognized as CRNs. Fig. 2.2 illustrates the

concept of spectrum hole.

Fig. 2.2: The concept of spectrum hole

The licensed users are known as the primary users (PUs) in CR terminology, the unlicensed

users, who are given opportunistic access of the spectrum, are known as secondary users (SUs)

or cognitive users, the underutilized bands of frequencies are known as “white spaces” or

“spectrum holes”, and such opportunistic spectrum sharing in a noninterfering manner is known

as dynamic spectrum access (DSA).

Thus each transmission process in CRs is completed in two steps. First is spectrum sensing to

detect vacant or underutilized bands of spectrum and second is transmission of data of the source

towards the destination. Both the spectrum sensing [26]-[27] and data transmission phases have

been extensively studied.

Simon Haykin in 2005 identified three specific tasks that lie in the core of CR [3]. First, radio-

scene analysis to be performed by the SU receiver, which includes detection of spectrum holes

and estimation of interference temperature around a PU receiver. Second, channel identification

15

which is required by the SU receiver to improve spectrum utilization efficiency and to perform

coherent detection of the PU signal. Third, dynamic spectrum management and transmit power

control to be performed by the SU transmitter. For this purpose, there is an obvious need to

connect the transmitter and the receiver via dedicated feedback channel to keep harmony

between both devices and to mitigate the interference offered to both the licensed and unlicensed

users.

CR is emerging as a promising technology for future wireless world. The salient features of

CRNs are spectrum sensing, spectrum management, spectrum sharing, mission-oriented

configuration, adaptive algorithms, distributed collaboration, routing and security [6]. CRs have

made rapid transition from an idea to reality during the last decade. The motivation behind this

remarkable progress to enable CR technology is the deployment of mature DSA systems. An

extensive research is being conducted by CR research community in the areas of spectrum

sensing and management, CR architecture, cooperative communication, dynamic spectrum

access algorithms, protocol architectures for CRNs, resource management, security issues and

finally large-scale deployment [6],[28]-[29].

Moving one step further, the large-scale deployment of CR technology heavily depends on the

interconnection of these intelligent devices, which work in collaboration to enhance the overall

system performance forming CRNs. In literature, infrastructure-based CRNs, cognitive ad-hoc

networks and hybrid networks have been proposed for CRNs [30]. Being wireless in nature,

CRNs are prone to all security threats inherent in conventional wireless networks [31]. The

common security threats are confidentiality, integrity, availability and access control.

2.2 SPECTRUM SHARING IN COGNITIVE RADIOS

A natural question is to explore the means by which the SUs can be accommodated in the

licensed spectrum bands without disrupting the PUs of the spectrum. Broadly speaking, three

spectrum sharing models proposed in CRs are underlay, overlay and interweave [5],[32].

16

2.2.1 UNDERLAY SPECTRUM SHARING

The underlay spectrum sharing follows the ultra-wideband (UWB) system strategy to support

simultaneous primary and secondary transmissions over same frequency bands. Underlay CRNs

guard the primary transmissions by enforcing a spectral mask on the secondary signals so that

the PUs remains undisturbed of the interference caused by the secondary transmissions. To

compensate the spectral masking, secondary signals are spread and de-spread over a wide

bandwidth to provide desired SNR at the secondary receiver. The main problem faced by

underlay networks is the limited transmit power ability of the secondary users due to the

interference constraints, which reduces secondary throughput and allows only short-range

communication. Thus, enabling secondary communication with minimum QoS in the frequency

spectrum allocated to the PUs is hot area of underlay research, and it requires fine tuning and

adjustment of the transmit power of the SUs. In [33], the authors suggested a transmit power

allocation scheme for dual-hop CRNs operating in AF mode, under transmit power constraints

and interference constraints. First, the optimization problem was simplified by relaxing the

transmit power constraint to obtain sub-optimal solution, which was then further utilized to

propose a power allocation scheme in order to satisfy both constraints all the time.

2.2.2 OVERLAY SPECTRUM SHARING

The overlay spectrum sharing enables simultaneous primary and secondary transmissions. For

this purpose, the SUs use a part of their transmit power for secondary communication, and the

remaining power to relay primary signals. This versatile mode of communication offsets the

decrease in PU’s SNR due to the interference caused by the SU’s transmit power by exactly

increasing the PU’s SNR due to the relaying services provided by SU. Sophisticated signal

processing and coding techniques are also employed for interference mitigation depending on the

available side information.

2.2.3 INTERWEAVE SPECTRUM SHARING

The interweave spectrum sharing is primarily based on the idea of opportunistic spectrum access,

i.e. to exploit the spectrum holes, that are temporarily not in use by the primary users. These

17

spectrum gaps highlight the spectrally underutilized regions in terms of time and space, and can

be used to accommodate the secondary users to enhance the spectral efficiency. The spectrum

holes vary with time and geographic location. Thus, they need to be periodically monitored to

intelligently detect unoccupied parts of the spectrum. The periodic scanning of the spectrum aims

to avoid the interference to the primary users once they are active again. Due to these reasons,

interweave mode is also known as interference avoidance mode of spectrum sharing.

2.3 COPERATIVE COMMUNICATION

Multiple-Input Multiple-Output (MIMO) systems [34], an extension of developments in antenna

array communication, employ multiple transmit and receive antennas as shown in Fig. 2.3, and

provide a number of advantages over single-antenna-to-single-antenna communication. These

advantages are: less sensitivity to fading effects of communication channel and increased

resistance to local interference due to existence of multiple spatial paths between transmit-

receive antennas, improved gain, reduced power requirements, capacity enhancement and many

more. However, these advantages are gained at the cost of implementation complexity and

increased size, which many wireless devices may not be able to support. Cooperative diversity

was proposed as an alternative potential solution.

Fig. 2.3: Multiple Input Multiple Output (MIMO) System

1

2

M

1

2

N

Transmitter Receiver

MIMO Channel

18

Cooperative communication [35]-[36] is an effective means of increasing the spatial diversity of

a signal in wireless communication networks. It has been recorded in [37]-[38] that cooperative

diversity networks can achieve a diversity order equal to the number of end-to-end source-

destination routes, however, capacity enhancement is limited by the transmission of symbols in a

time-division multiple access (TDMA) manner. It lays down the foundation of adhoc networks

and holds the promise of supporting sensor networks, communication networks for providing

public safety, strategic networks, cellular networks which are hierarchical in nature and military

applications [39].

Such communication strategy efficiently improves system throughput, combats channel fading,

reduces power consumption, increases transmission reliability and coverage area [40]-[43].

Cooperative communication techniques follow such approaches as collaborative signal

processing, cooperative coding and relaying [44]. Relaying [45] is a powerful cooperative

diversity technique in which multiple spatially distributed terminals, commonly known as relays,

assist the source by relaying its information to the destination. In such communication strategy,

the probability that all links are simultaneously down is much smaller than that for a single link

[39]. Relay networks [46], introduced by Van Der Meulen [3], have attracted tremendous

research attention and particularly find their applications in the networks with transmit power

constraints and portable mobile terminals, where mounting multiple antennas is difficult. A relay

network operates in either full-duplex or half-duplex mode [46]. In full-duplex mode of

operation, a relay can simultaneously transmit and receive at the same frequency, whereas, in

half-duplex mode, simultaneous transmission and reception on the same frequency band is not

supported on the relay.

2.3.1 RELAYING PROTOCOLS

In order to implement a cooperative communication network, efficient relaying strategies and

received signal combining schemes have been developed. Depending on the signal processing

strategy employed at the relay, two primary and most widely employed relaying techniques are

Amplify-and-Forward (AF) and Decode-and-Forward (DF) [41],[46]-[47].

19

2.3.1.1 Amplify-and-Forward Relaying

In AF relaying, also known as non-regenerative relaying, a relays simply adjusts the

amplification of the received signal and retransmits it, thus requiring less processing and low

power consumption at the relay [46]-[47]. AF relaying aims to overcome the power loss on the

source-relay link, however, a very important issue encountered by AF protocol is the

amplification factor required at the relay for scaling the received signal, which may result in an

unbounded power at the relay. In practice, peak transmit power constraints apply on the analog

circuitry involved in the communication devices. [48]. This unboundedness of the relay

amplification factor may result in peak power saturation and harmful interference to other

cochannel users, for example, CRs.

2.3.1.2 Decode-and-Forward Relaying

DF relays are more complex than the AF relays since they decode the received message, perform

error correction, encode the corrected message and retransmit it. Generally speaking, in DF

mode, source broadcasts its data which is heard by the destination and the relays. The set of

relays which are able to decode the received signal successfully constitute a set called the

decoding set [49]. DF protocol can be applied for both the coded sequences and the uncoded

signals. For coded sequences, coded DF is used in which error correction codes are added to the

symbols at the time of transmission [50]. The relays validate the reliability of the received signal

upon decoding via known error correction code. For uncoded symbols, a relay becomes a

member of the decoding set, if the received signal-to-noise (SNR) ratio at the relay exceeds a

predefined SNR threshold. Different variants of AF and DF protocols are already mature in

literature, proposed to enhance their accuracy and efficiency.

In addition, another relaying strategy, although less popular than AF and DF relaying is

Compress-and-Forward (CF), which is employed in the situations when a relay is unable to

decode the received signal and it sends an estimate of the source’s message to the destination

[51]. Fig. 2.4 shows the basic principle of AF and DF relaying.

20

Fig. 2.4: AF and DF Protocols

2.4 COGNITIVE RELAY NETWORKS

The combination of CR and user cooperation emerged as cognitive relay networks [31],[52],

which considerably improve the bandwidth efficiency, tackle unfavorable effects of wireless

channels and improve performance tradeoffs for both the PUs and the SUs [53]-[55]. In CR

research, cooperative communication is being extensively applied in spectrum sensing and

sharing, dynamic resource allocation, interference management. Cognitive relay networks follow

one of three approaches [56]. The first approach involves mutual understanding of primary and

secondary users whereby and the SUs act as relays to assist the PUs in their transmission, which

in turn provides more transmission opportunities to the SUs. The second approach is based on

the collaboration among SUs and in this scenario, the SUs relay signals for each other. In third

approach, spectrum-rich SUs help spectrum-short SUs and such communication strategy is

known as cooperative relaying. In addition to the design challenges inherent in the single-hop

CRNs, multi-hop CRNs also face research challenges related to resource sharing among different

nodes and performance enhancement of secondary network keeping in view the privilege of PUs

[55].

In order to deliver the best performance in cognitive relay networks operating in an underlay

spectrum sharing environment, the SUs, including the source and the relays, are allowed to

transmit concurrently with the primary transmissions over the same frequency band as long as

21

the interference offered by the SUs remain below a predefined threshold [52]. The interference

threshold can be defined by average or instantaneous interference power received at the primary

receiver [56]. The instantaneous or peak interference power requires the knowledge about

instantaneous channel gains of the interference channels and it is suitable for real-time traffic.

The average interference power applies to non-real time traffic where average SNR determines

the QoS. However, in practice, it is very hard to determine the interference threshold. In [57], the

instantaneous CSI between the secondary transmitter and the primary receiver along with the

instantaneous CSI of PU link determine the interference threshold.

2.4.1 RELAY SELECTION IN COGNITIVE RELAY NETWORKS

As mentioned above, the performance of underlay CRNs is significantly enhanced by

incorporating cognitive relays, which convey the message transmitted by the source to the

destination, but engaging the whole relay network may not be a feasible idea because the

interference produced by the relays to the concurrent primary communication may exceed the

threshold [58]. This practical limitation demands efficient alternatives of all-relays participation

in cognitive relay networks. Fortunately, cognitive relay networks offer a fascinating solution to

this problem in the form of relay selection. Relay selection aims to select the best combination of

relays or single best relay keeping in view the objectives and constraints of the system under

consideration. Owing to the half-duplex mode of communication [56], relay selection is

performed in two time-slots as shown in Fig. 2.5. In time slot 1, the message broadcast by the

source is heard by the potential relay network, whereas, in time slot 2, the selected relay(s)

retransmit the received message to the destination after necessary processing, while the source is

silent.

(a) (b)

Fig. 2.5: Conceptual Model for Relay(s) Selection, (a): Time Slot 1, (b): Time Slot 2

Source

broadcast

s Source Destination

Destination

22

However, a critical parameter to be considered in relay selection is the selection speed. Owing to

the time-varying nature of communication links, relay selection must be performed no slower

than the channel coherence time which is defined as the time duration over which the channel

impulse response is considered to be flat. If relay selection scheme fails to satisfy this constraint,

this might degrade the overall system performance by selecting a wrong relay because in this

case, the selection entirely based on the old CSI, while the channel conditions are changed at the

time selection is performed.

In this dissertation, relay selection in the context of AF based cognitive relay networks operating

in underlay mode is investigated in detail. Several best and multiple relay selection schemes have

been studied in this regard based on different design requirements and assumptions.

Note: “Multiple Relay Selection” and “Relay Subset Selection” terms will be used

interchangeably in the whole dissertation, where Relay Subset Selection aims to select multiple

relays from a potential relay set.

2.4.1.1 Best Relay Selection Proposed In Literature

Few research contributions involving best relay selection in underlay CRNs under interference

constraints are highlighted as follows. In best relay selection, only the single “best” relay, which

satisfies an index of merit, is nominated as a selected relay to participate in the communication.

In [59], Fredj et. al presented a scenario in which a secondary transmitter used the services of

intermediate relays to communicate to its receiver. In this scenario, best relay was selected from

the potential relay set to enable secondary communication under interference constraints.

Furthermore, end-to-end SNR statistics were derived and BER was evaluated for different

modulation schemes. D. Li investigated best relay selection based on full and partial CSI in [60],

and compared the performance of both schemes by deriving the closed-form expressions for

outage probability. For this purpose, a cluster of cognitive relays assisting a single source-

destination pair was considered. It was proved that partial-CSI-based relay selection was

outperformed by the full-CSI based relay selection. In [61], Seyfi et. al proposed a best relay

selection scheme for dual-hop cognitive relay network under transmit power constraints and

interference constraints. Furthermore, the outage probability of the secondary network with relay

selection was derived while considering the effect of PU interference. The derived results were

23

tested through simulations. In [62], Hussain et. al considered a cognitive source communicating

with its destination through direct transmission path and also with the help of intermediate fixed-

gain cognitive relays. In this scenario, best relay selection criteria was proposed aiming to

maximize the SNR achieved on relay-destination link. Furthermore, outage probability and BER

were derived in closed form for performance analysis. In [63], Bao et. al proposed best relay

selection and considered tight lower bound of the end-to-end SNR to derive the closed-form

expressions for CDF and probability density function (PDF) over non-identical Rayleigh fading

channels. The derived results were used to investigate the outage probability and average symbol

error probability of proposed system. The performance was evaluated against some key

parameters. The asymptotic analysis of the scenario showed that interference constraint does not

affect the diversity gain. In [64], the authors carried out derivations of the outage probability and

symbol error rate for cognitive relay networks under interference and transmit power constraints

over Rayleigh fading channels. It was shown through numerical results that the transmit power

constraint and interference power constraint cause the outage saturation phenomenon. The

analytical results were validated through Monte Carlo simulations.

2.4.1.2 Multiple Relay Selection Proposed In Literature

Research contributions in the area of multiple relay selection are, however, quite limited. In [65],

the authors proposed optimal and two suboptimal schemes of multiple relay selection in

cognitive relay networks with an objective to maximize the SNR received at the destination

under interference constraints. The comparative analysis of all schemes was carried out against

well-defined range of source transmit power for different interference threshold levels and

different number of candidate relays. Naeem et. al considered a dual-hop CRN, and proposed a

multiple relay selection scheme with interference awareness for underlay CR systems in [66]. It

was proved through simulations that the performance of the proposed scheme approached

exhaustive search technique, while having low implementation complexity.

2.4.1.3 Observations In The Literature Review Of Relay Selection Schemes

All the above mentioned research contributions for best and multiple relay selection are selected

for discussion, because they were built on some common assumptions which are as under. First,

underlay spectrum sharing model was assumed for each scenario. Second, all schemes, except

24

[62], assumed severe shadowing on the line-of-sight path between source-destination pair, thus

making direct communication impossible. Third, all system models were built up using single

antenna terminals. Fourth, AF relaying was assumed at the cognitive relay network. Fifth, all the

highlighted contributions for best and multiple relay selection assume the availability of CSI of

the interference channels.

Each of the proposed scheme studied above has been analyzed with interference and transmit

power constraints. However, there is a strong observation that all relay selection schemes

performing best relay selection study the outage behavior of the secondary network in detail. On

the other hand, the effect of relay subset selection on the outage probability and bit error rate of

the secondary system operating in an underlay spectrum sharing environment is not presently

available in literature, to the best of our knowledge.

These prior works have significantly improved our understanding of relay-assisted CRNs.

Inspired by the contributions in the field of relay selection, we focus on the deficiencies

highlighted in the performance evaluation of multiple relay selection schemes.

2.5 CONCLUDING REMARKS

This chapter starts with a comprehensive introduction to CRs alongwith the salient features,

different spectrum sharing modes and research challenges faced by this emerging technology.

Then the discussion proceeds towards the cooperative communication techniques highlighting

the significance of relay networks. Furthermore, cognitive relay networks have been discussed as

a merging technology of CR and cooperative communication. Finally, several contributions in

the areas of multiple relay selection and best relay selection proposed in literature for underlay

cognitive relay networks have been highlighted.

25

Chapter 3

NATURE INSPIRED ALGORITHMS AND

FUZZY LOGIC

Objective

This chapter discusses the history, advantages, optimization phases and applications of the tools

used in this disseration. After a comprehensive review of the history and applications of

Evolutionary Algorithms (EAs), two versatile EAs, namely, Artificial Bee Colony and Genetic

Algorithm are discussed in detail. Finally, the discussion moves towards Artificial Intelligence

(AI), and a well-known AI tool, the Fuzzy Logic is studied in depth.

3.1 THE EVOLUTIONARY ALGORITHMS

The difficulties associated with the mathematical modeling of large-scale engineering

optimization problems seek to develop alternative solutions. An optimization problem is the one

which aims to find out the best solution from all candidate solutions and requires specialized

problem solving techniques. When optimization is to be performed within complex domains of

available information, bio-inspired EAs based on the behavior of biological entities, have

emerged as a fascinating area in this framework [67]. The behavior of social insects, such as

finding the best food source, building of optimal nest structure, clustering etc. show intelligent

behavior on the swarm level. The swarm behavior heavily depends on the interactions among

individuals in addition to the behavior of individuals.

EAs are derivative-free methods, which belong to the class of probabilistic optimization method

[68] and perform very well to solve non-convex and non-differentiable problems [69]. The

classical and the most prominent EAs are Genetic Algorithm (GA), Artificial Bee Colony

26

(ABC), Particle Swarm Optimization (PSO), Artificial Immune Optimization (AIO), Differential

Evolution (DE), Invasive Weed Optimization (IWO), Ant Colony Optimization (ACO), and

Simulated Annealing (SA) [69],[70]-[71]. The suitability of any optimizing technique [68] for a

particular application is a compromise between accuracy, speed, complexity, memory demand,

utilization of a priori information, and balancing of global search (exploration) and local

refinement (exploitation).

During the past decades, these nature-inspired algorithms have gained tremendous attention by

research community, and have been successfully adopted in many applications requiring

optimization finding the optimal routes, scheduling, image and data analysis etc. for finding

near-optimum solutions of the optimization problems [72]-[75]. Broadly speaking, EAs find their

applications in engineering, social sciences, arts, economics, robotics and all fields of real-world.

Furthermore, most of the EAs are able to intelligently solve epistatic problems, in which the

quality of one variable is highly dependent on the other. EAs are also found to be robust in

nature, i.e. similar results are obtained from multiple runs of EA to solve the same problem.

Moving one step further, multiobjective evolutionary algorithms (MOEAs) have already been

developed to solve optimization problems involving multiple conflicting objectives in science,

economics and engineering [67],[76]. In such problems, MOEAs aim to identify a set of all

possible solutions rather than one optimal solution obtained in the case of single-objective

evolutionary algorithms (SOEAs) [67]. The set of all feasible solution represent the best

compromise between the multiple objectives defining that particular problem. These algorithms

can be classified into aggregating function algorithms, population based algorithms and Pareto

based approaches.

However, a common problem associated with all EAs is the imbalance between exploration and

exploitation. High degree of exploitation results in premature convergence to local minima, and

on the other hand excessive exploration slows down the execution [77]. Thus, a severe limitation

of the population based global optimizers is the lack of ability to do fine-tuning of the obtained

results. The global optimization tools are found to be good in exploration of the search space but

less good in the exploitation [78]. On the other hand, the local search algorithms like pattern

27

search (PS), Interior Point Algorithm (IPA) efficiently improve the accuracy of the results.

Hybridization of global optimizers with local optimizers came up as a fascinating solution to

tackle this issue, for instance, GA hybridized with PS [78],[79], Hybrid Evolutionary and

Gradient Search [41], Memetic PSO [80], DE hybridized with Sequential Quadratic

Programming (SQP) [81] are few prominent contributions in this regard.

Two famous global optimization algorithms employed to solve the optimization problems in this

research are Artificial Bee Colony and Genetic Algorithm. The history, applications and

optimization phases of these algorithms are discussed below.

3.2 ARTIFICIAL BEE COLONY OPTIMIZATION

Artificial Bee Colony (ABC) is a relatively new population based global optimization algorithm

proposed by Dervis Karaboga in 2005 [69], [74]-[75], [82]-[85] and it simulates the foraging

behavior of honey bees. In the initial phase, it was proposed to search the optimal solutions to

unconstrained problems [82]. Later, ABC and its extended versions have gained remarkable

attention of research groups due to ease of implementation, robustness, employing fewer control

parameters (mainly colony size and maximum iteration number), and good convergence

properties [85]-[87]. ABC performs smart handling of linear/non-linear constrained problems

and non-convex problems. Furthermore, it shows reduced computational overhead and does not

suffer from memory limitation problems since each candidate solution in the population is not

examined from the start to the end of optimization procedure [85].

ABC emulates honey bees intelligent behavior of searching for quality food source and sharing

that information with their fellows in the hive. A quality food source is the one, which contains

the highest amount of nector. Thus, the important decision making parameters in ABC execution

are: The amount of nector which corresponds to the fitness value associated with a particular

food source, and the number of iterations for which the algorithm is evaluated repeatedly [74].

ABC finds its applications in science, engineering and medicine [77], [88]-[89]. In order to

further enhance the accuracy of ABC and to overcome its limitations, certain modifications have

been carried out making it a versatile choice to solve multimodal and non-differentiable

28

problems [90]. These enhancements enabled ABC to outperform other state-of-the-art

metaheuristic algorithms (GA, DE, PSO) in efficiency, effectiveness and speed of convergence

[69],[74]-[75],[83],[85],[88],[90].

3.2.1 OPTIMIZATION PHASES OF ABC

ABC algorithm is based on the notion of Artificial Bees and Greedy Search procedure. The

entire optimization procedure in ABC is divided into two phases:

3.2.1.1 Initialization Phase

In the Initialization Phase, potential solutions (food sources) are randomly generated.

3.2.1.2 Best Solution Search Phase

In this optimization phase, search processes of Artificial Bees are recursively operated until the

best solution is achieved or the maximum number of iterations is expired. The best solution is

memorized by means of Greedy Search approach in ABC.

Broadly speaking, the artificial bees are divided into two categories: Employed Bees (EBs) and

the Unemployed Bees (UBs). Unemployed Bees are further classified into Onlooker Bees

(OLBs) and Scout Bees (SB)s. Thus, ABC combines local (employed and onlookers) and global

(scouts) search methods to achieve global or near-global optimum solution [85]. The tasks

performed by these foraging bees are summarized as follows.

a) Employed Bees (EBs)

Employed Bees are the “search agents” which search for the neighborhood solutions in the

vicinity of the initialized solutions and update their memory using Greedy approach by the

best solution that improves the fitness function and satisfies the constraints. There is a

dedicated EB for each potential solution.

b) Unemployed Bees (UBs)

i. Onlooker Bees (OBs)

OLBs are the “selector agents” which rely on the information shared by EBs about the

discovered solutions and exploit only those solutions chosen according to the probability

29

of their fitness function relative to the sum of all by Roulette wheel Mechanism. UBs

become EBs whenever they find a solution (food source) to act upon.

ii. Scout Bees (SBs)

SBs are the “replace agents” which carry out the random search in the whole search space

to replace the abandoned solutions by new ones. An abandoned solution is the one that

fails to improve the fitness function after several attempts w.r.t. the threshold level.

3.2.2 Flow Chart of ABC Optimization

For better understanding, main steps in ABC are summarized in the flowchart in Fig. 3.1.

Randomly initialize the population of SN number of solutions

Place EBs on the initialized solutions, kiYYYZ ikiiii ,10),(

SNi ,,2,1

Greedy Search between iY and iZ

Use Roulette Wheel to Spread Onlookers

Greedy Search to memorize the best solution

Discover new solutions via scouts

Memorize the best solution

Loop Expired or

criteria met? Return best solution

Yes No

Start

Fig. 3.1: Flow Chart of Artificial Bee Colony Optimization

30

3.3 GENETIC ALGORITHM

Genetic Algorithm (GA), was developed by Holland in 1970 on the basis of Darwin’s Theory of

natural evolution [91]. Holland’s study of natural adaptation phenomena aimed to find out ways

to apply the principles of natural evolution to the optimization problems. Later in 1975, he

presented GA as an abstraction of natural evolution in his book “Adaptation in Natural and

Artificial Systems”, and provided a theoretical framework to build the first GA. The primary

search procedures in GA like other evolutionary models are natural selection and survival of the

fittest [92]. The population of candidate solutions or chromosomes in GA is updated in each run

by selecting the best chromosome and discarding the unhealthy one.

3.3.1 OPTIMIZATION PHASES IN GA

Three steps followed in GA to produce successive generations are selection, crossover and

mutation [92]-[93]. Selection aims to produce next generation individuals, whereas, crossover

and mutation are the methods for reproduction. We briefly explain each one as follows.

a) Selection

Selection procedure aims to generate intermediate population by choosing those

chromosomes for survival in the next generation which exhibit the finest fitness scores,

while the remaining ones are discarded. The commonly used methods of selection are

stochastic uniform, remainder, roulette and tournament, rank, and scaling.

b) Crossover

Crossover combines two parents from the intermediate population to create offsprings.

Different commonly used crossover criteria are single point, two point, intermediate,

heuristic, scattered and arithmetic.

c) Mutation

Mutation functions introduce genetic diversity by making small random changes in the

individuals within a population. The purpose of mutation is spreading the search to a

broader space, thus preventing the algorithm from being stuck in the local minima.

31

Gaussian, uniform and adaptive feasible are the famous mutation function used for this

purpose.

3.3.2 FLOW CHART OF GA OPTIMIZATION

The sequence of steps followed by GA optimization can be better understood with the help of

flowchart shown in Fig. 3.2.

Fig. 3.2: Flow Chart Of GA Optimization

Create initial population

Parents’ Selection

Fitness evaluation of individuals

Fitness evaluation of children

Update population

Loop Expired

or fitness

achieved?

Return best solution

Yes No

Crossover to create children

Mutation

Start

End

32

3.3.3 APPLICATIONS OF GA

GA finds its applications in different domains including multimodal optimization, power

electronics, bioinformatics, economics, scheduling applications, robotics, security (encryption

and decryption), communication networks design, adaptive modulation, all fields of engineering,

and many more. Moreover, GAs have been extensively employed to solve miscellaneous issues

in CR systems, such as cooperative spectrum sensing [92],[93]-[94] joint channel and power

allocation [95], relay selection and resource allocation issues [96]. Efficiency, accuracy and

reliability of GA can be considerably improved by hybridization with any other competent and

well-balanced computational technique such as Interior Point Algorithm (IPA), Pattern Search

(PS) etc.

3.4 FUZZY LOGIC

Broadly speaking, Artificial intelligence (AI) is the automation of activities that are linked with

human thinking. These activities involve decision making, problem solving, learning, perception,

and reasoning [97]. CRNs encourage the use of AI tools for reconfiguration to meet the

requirements of changing network conditions. The AI tools of interest extensively applied in

huge number of real-time applications include fuzzy logic (FL), adaptive fuzzy logic (AFL),

expert systems (ESs), rough set (RS) theory and artificial neural networks (ANNs) [72]. FL,

based on human perception and cognition, is a powerful variation of crisp logic, which closely

relates the knowledge representation to human thinking. FL holds the power of natural

knowledge representation as well as strong inference capabilities of expert systems. FL is being

utilized in this research to intelligently solve the relay selection problems and is explained in

more detail.

3.4.1 HISTORY OF FUZZY LOGIC

The notion of FL, was first introduced by Lotfi A. Zadeh, in 1965 [98]. In the initial phase, it

made a very slow development, but by early 1970’s, the world paid attention to this new theory

when Zadeh delineated the motivation behind fuzzy control in 1972 [99]. An important

breakthrough in this progress was made in 1973 [100], introducing the basic idea of a linguistic

33

variable, defined as a variable whose values are linguistic terms rather than numbers. By the end

of 1970’s, FRBS captured the imagination of research groups and scientific laboratories started

implementation of fuzzy inference engines. After going through significant development over

five decades, FL systems are now used as a successful and dominant tool of AI.

3.4.2 FUZZY CONTROL SYSTEM

Block diagram of a fuzzy control system (FCS) is shown in Fig. 3.3 [101]-[103].

Fiure 3.3: Basic structure of FLC

Fig. 3.3: Fuzzy Control System

FCS is conceptually split into four components:

a) Knowledge Base

b) Fuzzifier

c) Fuzzy Inference Engine

d) Defuzzifer

The tasks performed by each component are summarized as follows.

a) Knowledge Base

The knowledge base contains all the knowledge required by the FCS and it comprises

fuzzy control rule base and a data base. The rule base illustrates the relations between the

input and the output variables through IF-THEN rules based on fuzzy reasoning. The rule

base must contain rules for every possible combination of the input space.

Crisp

Input

Crisp

Output

Fuzzy Input Sets Fuzzy Output Sets

Knowledge

Base

Fuzzification

Interface

Defuzzification

Interface

Inference

34

b) Fuzzifier

The fuzzification interface fuzzifies each input linguistic variable to produce a set of

fuzzy numbers. This is done by comparing each input variable with the predefined MFs

to allocate the variable, a value between 0 and 1.

c) Fuzzy Inference Engine (FIE)

The heart of FCS is the inference engine, fed by the fuzzy numbers from the fuzzifier and

it derives reasonable control actions on the basis of predefined rule base.

d) Defuzzifier

The fuzzy variables produced by FIE are converted into the crisp values by the

defuzzifier to represent the actual output of the system.

3.4.3 APPLICATIONS OF FUZZY LOGIC

FL has been heavily employed in power systems to solve issue related to automatic power

restoration and control, power optimization, system diagnosis and stability, classifying PQ

disturbances, protection, fault diagnosis and load forecasting [72]. FL based computer aided

design (CAD) tools are used to address issues in analog and digital circuit design. In the industry,

FL has been successfully applied in the modeling of complex systems and smart handling of

design, manufacturing and control issues. One of the most demanding applications of fuzzy set

theory is pattern recognition [104]-[105]. In addition to the above highlighted applications, fuzzy

set theory has been applied in incredibly diverse real-world applications, including engineering

design, social science, robotics, economics, management, finance, web mining, heuristic control

and regression analysis. [73], [101],[106]-[109].

3.4.4 FUZZY LOGIC IN CRNS

FL adds more degrees of freedom to make sophisticated and reliable decisions, thus it has

already been employed in CRNs, especially in the areas of spectrum sensing, interference

management and power control [111]-[119]. Fuzzy logic based transmit power control schemes

result in simple, reliable and cost effective implementations. Some methods of FL based transmit

power allocation and power management in CRNs are highlighted as under. In [120], the authors

designed a fuzzy logic system (FLS) based transmit power controller to enable coexistence of

35

primary and secondary users. It was proved through simulation results that FLS based power

control mechanism significantly reduced the average values of outage probability and transmit

power increase. In [121], Tabakovic et al. designed a simple and cost effective FL based transmit

power controller aiming to enable the secondary communication with the desired QoS

requirements, while ensuring that the interference offered to the PU is minimized and the mutual

interference of the SUs. Another fuzzy rule based opportunistic power allocation strategy

proposed in [122] for the efficient utilization of radio spectrum. However, the application of FL

to perform multiple relay selection for performance enhancement of secondary communication

in underlay cognitive radio networks has not been done so far to the best of our knowledge.


In this chapter, the literature review of Evolutionary based algorithms has been carried out with

focus on Artificial Bee Colony and Hybrid Genetic Algorithm. Furthermore an efficient

Artificial Intelligence tool, the Fuzzy Logic has been studied in detail. For each algorithm, the

optimization phases are described in detail and the applications are highlighted.

36

Chapter 4

SNR MAXIMIZATION IN UNDERLAY

CRNS

Objective

In this chapter, performance enhancement of secondary communication via AF based cognitive

relay network under Rayleigh flat-fading channel model is studied. Underlay spectrum sharing is

assumed at the intermediate relay network, and six algorithms for relay selection and power

allocation have been proposed based on the availability of perfect CSI. Each proposed algorithm

aims to maximize the SNR received at the destination under interference and transmit power

constraints. Impact of source and relay(s) transmit power, interference threshold levels, number

of candidate relays have been very well investigated for each algorithm keeping in view the

objectives and constraints.

4.1 PROBLEM FORMULATION 1

The first optimization problem is different from the other problems of SNR maximization in

defining the transmit power constraint for the relay network. No individual transmit power

constraint is imposed on any relay, rather the selected subset of relays S must constrain the total

transmit power Sm

mP below the predefined threshold maxP . The sum transmit power constraint

aims to give relaxation to individual relays to transmit at any power for secondary performance

enhancement, keeping in view the interference threshold of the PU. Hence, the mathematical

formulation of the multiple relay selection and power allocation problem can be expressed as,

37

SS m mm

mm

m

mD21

21

1max

..ts

max:1 PPC

Sm

m

thm

m

m IfPC

S

2:2

)1.4(

where maxP is the maximum power that can be transmitted by cluster of M selected relays and

thI is the maximum tolerable interference for the PU. 1C defines the transmit power constraint,

whereas 2C is the interference constraint. Both constraints must be satisfied by the relay

network to enable the secondary communication. As obvious from the constraint 2C in 4.1, sum

transmit power of the relay network Sm

mP critically effects the total interference power towards the PU.

4.2 PROPOSED ALGORITHMS

Two algorithms are proposed to solve the highlighted problem, which are explained as under.

4.2.1 PROPOSED ALGORITHM 1

The relay selection and power allocation algorithm works as follows. Let },......,2,1{ Minitial be

the initial set of relays. The proposed algorithm is a two-phase algorithm. After initialization of

transmit power of potential relay set satisfying constraint C1, the first phase performs relay

selection to satisfy constraint 2C . For this purpose, the sum interference power I due to initial

towards the PU is computed and S is updated by excluding the relays in the descending order of

individual interference offered by each relay, aiming to satisfy the interference threshold thI .

After ensuring the security of the primary communication, the second phase works on improving

the SNR achieved at the destination. To achieve this goal, each selected relay gets an equal

increment in its transmit power to increase the corresponding relay-destination link SNR. The

38

equal increment in the transmit power of each selected relay aims to reduce the computational

complexity, while satisfying sum transmit power threshold maxP . The pseudocode of the proposed

algorithm is provided in table 4.1.

Table 4.1: Pseudocode of the Proposed Algorithm 1

4.2.1.1 ABC Optimization

To solve the above mentioned non-linear constrained optimization problem, ABC is used for

optimization. ABC is preferred due to the advantages and applications highlighted in chapter 3.

Since the optimization phases have already been explained in chapter 3 with the help of flow

chart, the pseudo code of ABC algorithm to solve the SNR maximization problem is directly

presented in table 4.2.

The Proposed Algorithm

MNMmforfhgIPNPInputs SinitialmmmthS ,,,,,,,,: max0

);()( 0

22 NgPAP mSmm m

max

1

PPPM

m

msum

1// Csatisfying

thm

m

m

m

m IfPIwhile2

2// Ccheck

mm IforP max;0 ;1& MNS

endwhile

;_ m

mnewsum PP Smfor

;

_

N

PP newsumsum

; mm PP Smfor 2&1// CCsatisfying

;1 21

21

SS m mm

mm

m

mD

39

Table 4.2: ABC Optimization to solve SNR Maximization Problems

Artificial Bee colony Optimization

orithmaproposedtoaccordingsolutionspotentialSNInitialize lg

iterationsLCfor :11

)(:1 EBsBeesEmployedPhase

SNnfor ,,2,12

ntsegergeneraterandomly ...,int

;* m

n

m

n

m

n

m

n

memp PPPP m 10 n

m

21&int CandCsatisfyPonphaseialiationofnscomputatioapplyn

memp

;

1 21

21

SS m

n

newmm

n

newmm

m

n

newm

n

newD

ocedureSearchGreedyifreplace n

D

n

newD

n

D Pr//

2forend

:2Phase Onlooker Bees (OBs)

;

1

SN

n

n

D

n

Dnp

SNn 1 fitnesscheck//

SNnfor ,,2,13

methodWheelRoulettebysolutionstheontosonlooPlace ker

3forend

Phase 3: Scout Bees (SBs)

if (rem(LC,2) = =0)

for4 n = 1,2,……, SN

Generate new food source for the abandoned ones

end for4

end if

end for1

40

4.2.1.2 Simulation Results The performance of the proposed scheme for multiple relay selection is analyzed against

different interference threshold levels and different sizes of potential relay network. Furthermore,

the convergence behavior of ABC is strongly observed. The parameter settings for all

simulations of this chapter are listed in table 4.3.

Table 4.3: The Parameter Settings

Parameters Values

mg 9.03.0

mf 5.01.0

mh 9.03.0

0N 1

SP 10

maxP 10

For all simulations presented in this chapter, N denotes the number of selected relays. Fig. 4.1

evaluates the multiple relay selection algorithm at different interference threshold levels and for

different sizes of potential relay network. The interference thresholds are set to

)5,0,5( dBdBdBI th and M is set to )10,5(M . There are three strong observations First,

quick convergence of ABC, second, high SNR achieved at high interference threshold which

gives more freedom to the relays to transmit at high power, and finally SNR is further enhanced

by increasing the size of potential relay network, because large network gives more choices to

select the best possible combination of relays The NM relays offering relatively high

interference are not allowed to participate in communication.

41

Fig. 4.1: Performance Analysis of Algorithm 1

4.2.2 PROPOSED ALGORITHM 2

The second algorithm proposed to solve the same problem is the “Power Set Algorithm”, which

works as follows. Let },,2,1{ Minitial denotes the initial set of potential relays. The

proposed algorithm initializes the transmit power of each relay while satisfying individual relay’s

transmit power constraint and creates all possible non-trivial subsets of transmit power vector

],,,[ 21 MPPPP , where the number of non-trivial subset in is given by,

M

n n

MJ

1

.

Given J subsets in , the relay subset selection algorithm selects JL subsets denoted by

Lll 1

, such that, each element in the thl subset l satisfies the condition mm hf . This relay

selection criteria aims to select those relays which are less harmful to the primary

communication. Next, the interference offered by each thl subset in is computed, where

interference offered by each thm relay is defined as2|| mmm fPI . Thus, the better is the channel

coefficient mf towards the PU, the higher is the interference offered by that relay. Finally, that

0 5 10 15 200

5

10

15

20

Iterations

D(d

B)

Ith

= -5dB, M = 5, N = 2

Ith

= 0dB, M = 5, N = 3

Ith

= 5dB, M = 5, N= 3

Ith

= -5dB, M = 10, N = 2

Ith

= 0dB, M = 10, N = 3

Ith

= 5dB, M = 10, N = 3

42

subset is declared as the finally selected subset out of L selected subsets which maximizes the

SNR D received at the destination, keeping in view the constraints. The pseudo code of the

proposed algorithm is given in table 4.4 below.

Table 4.4: Pseudo Code For Proposed Algorithm 2 (Power Set Algorithm)

4.2.2.1 Simulation Results

For fair comparison, the SNR achieved by the power set algorithm is also analyzed against

different interference threshold levels and different sizes of potential relay network. The results

obtained are shown in Fig. 4.2. SNR is significantly improved because the power set algorithm

only allows those relay to participate in communication which strictly satisfy the condition

0 mm fh .

else

1//max

1

CsatisfyingPPPM

m

msum

mNgPAP mSmm )||( 0

22 1// Csatisfyingwhile

PJ

j 1][ PofsubsetstrivialnonAll //

L

ll 1

0..// mmm fhsatisfiesinelementeachts

l

th subsetleachforLlfor //:1

MNMmforfhgNPIPInputs SinitialmmmSth ,,,,,,,,: 0max

lbysatisfiedisconstraerferenceif intint

2&1 CCsatisfyingrelaysofpowertransmitadjust

lm

l

m

l

m

l

m

l

ml

D

21

21

1

ifend

forend

imumistoingcorrespondts sD

L

lls max..1

initialS N SofycardinalitthedenotesN //

SDOutputs ,:

43

Fig. 4.2: Performance Analysis Of Proposed Algorithm 2

4.3 COMPARISON OF THE PROPSOSED ALGORITHMS

The comparison of the above proposed algorithms is shown in the Fig. 4.3 alongwith table 4.5,

showing the number of selected relays for both algorithms at different interference threshold

levels and different number of potential relays. The comparisons show that increase in the size of

potential relay network does not necessarily increase the number of selected relays but increases

the SNR achieved at the destination significantly because more choices are available to select the

best combination of relays. Algorithm 2 outsmarts algorithm 1 because it exclude the relays

which exhibit bad channel coefficients towards destination as compared to the corresponding

coefficients towards the PU.

4.3.1 CONCLUDING REMARKS

Performance enhancement of secondary communication in underlay CRNs has been focused, and

two multiple relay selection and power allocation scheme are proposed for relay-assisted CRNs.

It is proved that relay selection based cooperative diversity scheme increases the SNR at the

-5 -3 -1 1 3 5

10

12

14

16

18

I th(dB)

D(d

B)

M = 5

M = 10

44

destination keeping the total interference offered by the selected relays to the PU below a certain

threshold level. Furthermore, the CSI based power set algorithm offers better results because it

makes decision about selecting a particular relay on the basis of each relay’s outgoing channel

coefficients.

Fig. 4.3: Comparison Of The Proposed Algorithms

Table 4.5: Comparison Of The Proposed Algorithms

Proposed Algorithm Total No. of

Relays “M”

No. of Selected

Relays “N”

Algorithm 1 5 -5 8.2 2

0 12.2 3

5 15.3 3

Algorithm 1 10 -5 8.7 2

0 12.9 3

5 16.1 3

Algorithm 2 5 -5 9.3 2

0 14.7 3

5 17.2 3

Algorithm 2 10 -5 10.1 2

0 15.3 4

5 17.6 4

)(dBIth )(dBD

-5 0 55

10

15

20

Ith

(dB)

D(d

B)

Algorithm 1, M = 5

Algorithm 1, M = 10

Algorithm 2, M = 5

Algorithm 2, M = 10

45

4.4 PROBLEM FORMULATION II

The second problem of SNR maximization imposes individual transmit power constraint on each

individual potential relay, rather than defining sum transmit power constraint for the whole relay

network. The transmission power mP of each relay is limited not only by battery capacity due to

regulations specifying the maximum power that each node is allowed transmit, but also by the

interference threshold of the PU.

Thus, the mathematical formulation of the multiple relay selection and power allocation problem

can be expressed as,

SS m mm

mm

m

mD21

21

1max

..ts

:1C maxPPm

:2C thm

m

m IfP

S

2

)2.4(

where, maxP is the maximum power that can be transmitted by any selected relay and thI is the

maximum tolerable interference for the PU. 1C defines the individual relay’s transmit power

constraint, whereas 2C is the interference constraint. Both constraints must be satisfied by the

relay network to enable the secondary communication.

4.5 THE PROPOSED ALGORITHMS

The algorithms proposed to solve the SNR maximization problem formulated above assume the

availability of perfect CSI at each relay. The algorithms are explained one by one in this section.

4.5.1 PROPOSED ALGORITHM I

The optimization problem is again solved using ABC. The pseudocode of the proposed algorithm

is explained in table 4.6.

46

Table 4.6: Psuedocode For Algorithm I

Let },......,2,1{ Minitial be the initial set of relays. The proposed multiple relay selection and

power allocation algorithm is a two-phase algorithm. In the first phase, total interference power

due to initial towards the PU is computed and S is updated by excluding those relays offering



mNgPAP mSmm );||( 0

22

1// Csatisfying

m

th

m

mm IfPmIwhile )||)(( 2 // Constraint 2C

1&)max(;0 NIforP Smm

computeRe m

mI )(

];|[ mmm hff Smfor

{})( if

][ mf Smfor

endif

))(:1( sizeifor

;mmm PP );min(for

min// previousmm

;, max mm PPwhere

)2( satisfiednotCif

satisfiedisIuntiltakenPlastdecrement thm

exitfor

endif

endfor

;1 21

21

SS m mm

mm

m

mD

SDOutputs ,:

endwhile

47

relatively high interference until thI is satisfied. In the second phase, transmit power of relays in

S is increased to improve SNR at secondary destination while satisfying the constraints. This is

achieved by increasing transmit power of those relays in S which exhibit good channel gains

towards the secondary destination as compared to the corresponding channel gains towards the

PU assuring that constraints C1 and C2 are still satisfied. If none of the relays is able to satisfy

mm hf , then transmit power of relays in S are increased by selecting relays in the ascending

order of mf .


All parameter settings are done according to table 4.3 for fair comparison. Fig. 4.4 illustrates the

performance of the proposed algorithm for different levels of interference threshold thI and

different number of potential relays M . The number of selected relays are provided in table 4.7.

Hence the proposed algorithm exploits spatial diversity and multiple relays participate in

improving secondary network’s performance ensuring reliability of primary communication.


-5 -3 -1 1 3 58

10

12

14

16

18

Ith(dB)

D(d

B)

M = 5

M = 10

48

Furthermore, high SNR is achieved at the destination as compared to the proposed algorithm

presented in section 4.2 above, since each selected relay gets an increment in its transmit power

depending on its corresponding channel coefficients towards the destination and the PU. Thus,

the relays which exhibit mm fh are preferably allowed to transmit at maxP . It is also observed

that an increase in the value of thI allows more relays to take part in communication to enhance

SNR D achieved at the destination, because the higher the value of thI , the more is the freedom

given to the relays to transmit at maximum available power maxP .

Table 4.7: No. of Selected Relays Obtained From Proposed Algorithm 1

4.5.1.2 Concluding Remarks

A CSI-based multiple relay selection with adjustable power allocation scheme is proposed for

underlay CRNs aiming to maximize the SNR achieved at the destination while adhering to

interference power constraint towards the PU. The proposed scheme achieved satisfactory results

at different interference threshold levels and outsmarts previously proposed algorithm in section

4.2 at all interference threshold levels.

4.5.2 PROPSOED ALGORITHM II

This algorithm is the same power set algorithm proposed in section 4.2.2 to solve the SNR

maximization problem I, and it works as follows. Let },,2,1{ Minitial denotes the initial set

of potential relays. The proposed algorithm initializes the transmit power of each relay while

satisfying individual relay’s transmit power constraint and creates all possible non-trivial subsets

M N

5

-5 11.3 2

0 13.7 3

5 18.2 3

10

-5 11.7 2

0 14.3 3

5 18.7 3

)(dBIth)(dBD

49

of transmit power vector ],,,[ 21 MPPPP , where the number of non-trivial subset in is

given by,

M

n n

MJ

1

. Given J subsets in , the relay subset selection algorithm selects JL

subsets denoted by Lll 1

, such that, each element in the thl subset l satisfies the condition

mm hf . This relay selection criteria aims to select those relays which are less harmful to the

primary communication. Next, the interference offered by each thl subset in is computed,

where interference offered by each thm relay is defined as2|| mmm fPI . Thus, the better is the

channel coefficient mf towards the PU, the higher is the interference offered by that relay.

Finally, that subset is declared as the selected subset out of L selected subsets which maximizes

the SNR D received at the destination, keeping in view the constraints. The pseudo code of the

proposed algorithm is given in table 4.8 below.

Table 4.8: Pseudocode For Algorithm II



)||( 0

22 NgPAP mSmm m

1// Csatisfyingwhile

PJ

j 1][ PofsubsetstrivialnonAll //

L

ll 1 0..// mmm fhsatisfiesinelementeachts

Llfor :1 l

th subsetleachfor //

if (interference constraint is satisfied by l )

l mm

mm

ml

DR

l

SR

l

DR

l

SRl

D

1

else

ifend

forend

imumistoingcorrespondts sD

L

lls max..1

initialS N SofycardinalitthedenotesN //

SDOutputs ,:

50


The effectiveness of the proposed CSI based relay subset selection scheme is proved in this

section. Fig. 4.5 shows the obtained results, where the number of potential relays and range of

interference thresholds is kept the same for fair comparison with other proposed algorithms. It is

strongly observed that the proposed “Power Set Algorithm” outperforms the proposed algorithm

1, due to the same reason highlighted with the simulation results of power set algorithm of

section 4.3.1. The improved performance is due to the freedom of selecting that subset of relays

from the whole relay network which exhibit good channel conditions towards the destination.


Table 4.9: No. Of Selected Relays From Proposed Algorithm 2

-5 -3 -1 1 3 510

12

14

16

18

20

Ith(dB)

D(d

B)

M = 5

M = 10

M N

5

-5 12.8 2

0 16.5 3

5 19.1 3

10

-5 13.2 2

0 16.8 4

5 19.5 4

)(dBIth)(dBD

51

Whereas, if full diversity is achieved by allowing all relays to forward the source’s data, the

transmit power of the source and/or each relay need(s) to be suppressed keeping in view the

interference constraint of the PU, which in turn causes negative effect on the SNR received at the

destination. Again the larger relay network increases the probability of selecting those relays

which exhibit favorable channel conditions towards destination as compared to PU thus showing

higher improvement. Finally, as mentioned earlier, relaxing thI allows the relays to transmit at

high power thus enhancing secondary SNR for both cases.


A power set algorithm is proposed to solve multiple relay selection problem in AF based

Underlay CRNs. The proposed scheme outperforms the previously proposed multiple relay

selection scheme due to the CSI-based exhaustive search involved in finding the best subset of

relays from all non-trivial subsets of the potential relay network which is able to achieve the

objective while satisfying the constraints.

4.5.3 PROPOSED ALGORITHM III

This algorithm is based on a strong observation about the behavior of m . In literature, tight upper

bound for end-to-end SNR m given in 1.4 for thm relay link, exist in for a comprehensive

performance analysis [123]. In terms of these bounds, m is expressed as,ub

mm

lb

m , where,

),min(2

121 mm

lb

m and, ),min( 21 mm

ub

m . The lower bound clearly indicates, the minimum

value of m occurs when mm 12 , and, for this case,2

1m

m

, and, if m2 is increased further by

performing high amplification at the corresponding relay, the upper bound is approached. Thus

any attempt to improve end-to-end SNR m of thm relay link through power allocation to those

relays with mm 21 will not produce any significant change in m , as m in this case is entirely

dependent on the SNR of first hop m1 , i.e. mm 1 .Thus, if m2 is increased by performing high

52

amplification at the corresponding relay, the upper bound is approached. On the basis of this

strong observation, our relay selection and power allocation algorithm works as follows.

Let },,.........2,1{ Minitial be the initial set of potential relays. The proposed algorithm is a two-

phase algorithm and it performs decision making on the basis of the SNR m1 of first hop, the

SNR m2 of the second hop and the interference mI offered by each relay to the PU.

Phase-1 performs relay selection to satisfy interference constraint towards the PU by calculating

total interference power I offered by the relays in initial . In the underlay model, the relay with

good SNR may be destructive for primary network by generating high interference to the

neighboring PU and hence may not be chosen to participate in communication. Thus, if I

exceeds thI , the relay having best channel gain mf towards the PU is excluded from the set and

thI is recomputed after updating S . If S still fail to satisfy thI , then the algorithm builds up 1

considering those relays in S having mm 12 and updates S by deselecting relays in the

decreasing order of elements of mm 121 until interference constraint is satisfied.

Phase-2 works on S and SNR maximization is performed by creating 2 such that 2 is a

complement of 1 and increasing the transmit power of relays in 2 by selecting relays in the

increasing order of elements of mm 212 while satisfying constraints C1 and C2.

The pseudo code for the proposed algorithm is provided in table 4.10.

53

Table 4.10: The Pseudo Code For The Proposed Algorithm


mfhgNNMNMIPPInputs mmmSinitialths ,,,,,,,,,: 0max

)||( 0

22 NgPAP mSmm 1// Csatisfying m

1// Phase

))(( thIIsumif 2int// Cconstra

)(max;0 mm fforP 1 NS

endif

))(( thIIsumif

]|[ 12121 mmmm Sm

))(:1( 1 lengthjfor

;0mP )max( 1for

1 NS

)( Isumrecompute

)2( satisfiedisCif

;break

endif

endfor

endif

NS where MN

2// Phase

]|[ 21212 mmmm Sm

))(:1( 2 lengthkfor

;mmm PP for )min( 2

..,, max tsPPwhere mm 2min previousofanyformm

2int// CconstrasatisfytoassoPtostepssmallinfedis mm

endfor

m m mm

mmmD

21

21

1

Sm

SDOutputs ,:

54


In this section, performance analysis of the proposed algorithm is carried out through

simulations. The performance of cognitive relay network is again evaluated for different

interference threshold levels thI , and different number of potential relays M .

In Fig. 4.6, again it is observed that D increases with an increase in the value of thI as expected.

However, there is a strong observation that the proposed algorithm does not perform well under

low interference thresholds, because the relay selection criteria does not take into account the

interference offered to the PU, rather a selfish approach is adopted in which the decisions are

taken on the basis of SNRs achieved on the source-relay and relay-destination links. Thus, a

selected relay might not be allowed to transmit at high power because of being harmful to the

PU. On the other hand, when interference threshold is relaxed, the algorithm outsmarts the power

set algorithm, because it strictly considers the bounds on the SNR of selected relay link, and

increases the transmit power of those selected relays in the second phase that have SNR of relay-

destination m2 close to SNR of source-relay m1 .


-5 -3 -1 1 3 55

10

15

20

25

Ith(dB)

D(d

B)

M = 5

M = 10

55

Table 4.11: No. Of Selected Relays From Proposed Algorithm 3


A CSI based multiple relay selection algorithm is proposed for underlay CRNs which maximizes

SNR at the intended destination under PU’s interference constraint and individual relay’s

transmit power constraint. The performance of the proposed algorithm is very-well studied

against different levels of interference threshold and number of potential relays. The low

computational complexity of the proposed algorithm makes it a suitable candidate for designing

underlay systems.

4.6 COMPARISON OF THE PROPOSED ALGORITHMS

The comparison of the above proposed algorithms is shown in the Fig. 4.7 alongwith table 4.12,

showing the number of selected relays for the three algorithms at different interference threshold

levels and different number of potential relays. Similar trends are observed as with the case of

algorithms proposed to solve problem 1. However, Algorithm 2 (Power Set Algorithm)

outsmarts algorithm 1 and Algorithm 3 at all interference threshold levels, because it performs

exhaustive search and considers each and every possible subset of potential relay subset by

excluding the relays which exhibit bad channel coefficients towards destination as compared to

the corresponding coefficients towards the PU. However, at high interference threshold levels,

algorithm 3 gives satisfactory results.

M N

5

-5 8.9 1

0 10.9 2

5 20.3 3

10

-5 9.2 2

0 11.3 2

5 20.8 3

)(dBIth)(dBD

56

Fig. 4.7: Comparison Of The Proposed Algorithms

Table 4.12: Comparison Of The Proposed Algorithms

M N

Algorithm 1

5

-5 11.3 2

0 13.7 3

5 18.2 3

10

-5 11.7 2

0 14.3 3

5 18.7 3

Algorithm 2

5

-5 12.8 2

0 16.5 3

5 19.1 3

10

-5 13.2 2

0 16.8 4

5 19.5 4

Algorithm 3

5

-5 8.9 1

0 10.9 2

5 20.3 3

10

-5 9.2 2

0 11.3 2

5 20.8 3

-5 -3 -1 1 3 58

12

16

20

24

Ith(dB)

D(d

B)

Algorithm 1, M = 5

Algorithm 1, M = 10

Algorithm 2, M = 5

Algorithm 2, M = 10

Algorithm 3, M = 5

Algorithm 3, M = 10

)(dBIth)(dBD

57


In this chapter, several multiple relay selection schemes are proposed in an underlay scenario

under interference constraints and transmit power constraints. For each algorithm, SNR achieved

at the destination is strictly observed. All the algorithms provide high SNR at high interference

threshold levels and for large potential relay network. However the power set algorithm

outsmarts all the algorithms in both cases of transmit power thresholds, i.e. for individual relay’s

transmit power constraint and for total transmit power constraint of the whole relay network

outsmarts a. Another important observation is that, increase in the number of available choices of

relays increases SNR at the destination, because a system which assigns more relays to the users

gives more opportunities to select those relays which effectively participate in enhancing

secondary communication performance without degrading the quality of primary communication

taking place in parallel. Furthermore, convergence behavior of ABC is also showed for one of

the proposed algorithms. ABC proved its ease of implementation, fast convergence and

effectiveness in solving multi-constrained optimization problem.

58

Chapter 5

OUTAGE ANALYSIS OF MULTIPLE

RELAY SELECTION

Objective

In this chapter, we investigate the outage behavior of multiple relay selection over Rayleigh flat-

fading channels. In particular, we derive closed-form expressions for outage probability and bit-

error rate of underlay relay-assisted CRN. For this purpose, a dual-hop CRN operating in AF

mode is considered, and a multiple relay selection scheme is proposed, while ignoring the line-

of-sight path between source-destination pair. Finally, simulation results are presented to verify

the derived results.

5.1 PROBLEM FORMULATION

Refer to chapter 1 for the system model and basic assumptions that apply here also. The total

instantaneous end-to-end SNR D at the secondary destination due to M relaying links

expressed in 1.5 is again expressed here,

M

m mm

mmM

m

mD

1 21

21

11

)1.5(

In underlay networks, sophisticated signal processing techniques are employed to mitigate the

interference offered to the PU. However, due to the inherent simplicity of AF protocol, such

computationally complex techniques may not be supported at the AF relays. Thus, enabling the

secondary communication exploiting the services of all relays in the potential relay set may not

59

be a viable idea in terms of total interference offered by the relay network to the nearby PU. For

simultaneous primary and secondary transmissions in such power-constrained environment, the

total interference power experienced by the PU by the potential relay set must satisfy the

predefined interference threshold given as,

th

M

m

mm

M

m

m IfPII 1

2

1

|| )2.5(

where thI is the interference threshold set by the PU.

Relay selection stands as a fascinating solution to this problem. The proposed relay subset

selection problem as follows. Let P represents the transmit power vector of the potential relays

in the network, i.e. ],......,,,[ 321 MPPPPP . The number of all non-trivial subsets P is

given by

M

k k

MS

1

. The thl subset is denoted as l where }).,,.........2,1{( Ml . The

cardinality of thl subset is lM . Next is to compute total interference power due to each

thl subset

of relays towards the PU, where interference offered by each thm relay in any subset is defined

as2|| mmm fPI .

Let J be the number of subsets out of S , denoted asJ

jjL 1}{ , which satisfy the interference

threshold thI towards the PU. The interference constraint forthj such subset can be given as,

th

Lm Lm

mimj IfPII

j j

2|| Jj .,,.........2,1 )3.5(

The mathematical formulation of this optimization problem is as follows:

jj

ij Lm mm

mm

Li

RDL 21

21

1max

th

j IIts .. )4.5(

60

In order to investigate the performance of the overall system in terms of outage

probability and average probability of error, we need to know the distribution of D which is not

mathematically tractable. To overcome this problem, tight upper and lower bounds for end-to-

end SNR m , given in 5.1, exist in literature for a comprehensive performance analysis, as

explained in section 4.4.4.

Keeping the behavior of m under consideration, we aim to maximize m2 through controlled

transmit power allocation to each relay so that m of each relay link tends to approach its upper

bound causing an overall favorable impact on ,

j

DLm

m while keeping the sum interference

constraint satisfied. Thus, the relay subset selection algorithm aims to pick up that subset of

relays, which maximizes combined SNR of relay-links, j , where, for each thj subset jL , j is

defined as

jLm

mj 2 .

Thus, the mathematically tractable form of our optimization problem is given as,

jjLm

mj

L

2max

s.t. thj II )5.5(

5.2 THE PROPOSED ALGORITHM

Let },,2,1{ Minitial be the initial set of potential relays. The proposed algorithm works as

follows. Transmit power of each relay is initialized, followed by selecting all possible subsets of

relays which are able to satisfy sum interference power threshold thI set by the PU. For all such

subsets, combined SNR of relay-destination links is computed and finally that subset is declared

as the selected subset which maximizes the SNR. The pseudo code of the proposed algorithm is

61

table 5.1 below. For more clarity, the flowchart of the proposed algorithm is also presented in

Fig. 5.1.

Table 5.1: Pseudocode For The Proposed Relay Subset Selection Algorithm

:INPUTS initialmmminitials mfhgNMP ,,,},,......,2,1{,, 0

Mmfor .,,.........2,1

)||()( 0

22 NgPAP msmm ionamplificatimumuse min//

2|| mmm fPI

forend

NP 1][ Pofsubsetstrivialnon//

j jLm

thmm

Lm

i

j

j

th IfPIILsubsetjfortsofoutsubsetsAllL 2||,..

Jjfor :1 Linsubsetsselectedofnumber//

th

j IIwhile

mm PP

m

m

f

hfor max

j

thth Lsubsetselectedjofelementmdenotesm//

endwhile

jj Lm

mm

Lm

mj hP 2

2 || Jj ,,1

forend

imumisLtoingcorrespondtsLL jjj max..

jjLOUTPUTS ,:

62

Fig. 5.1: Flowchart Of The Proposed Algorithm

?Jj

Yes

No

Initialize transmit power of each potential relay

Obtain S non-trivial subsets of potential relay set

Extract J

jjL1

subsets

which satisfy interference

constraint, where, SJ

1j

mmm PP for

m

m

f

hmax , jLm

m is adjusted to ensure that interference

constraint remains satisfied

1 jj

Pick the subset jL that offers maximum SNR

on the relay-destination links

Outputs: jjL ,

Start

End

63

5.3 PERFORMANCE ANALYSIS

In this section, the performance of the proposed multiple relay selection scheme is investigated

and has been compared with the best relay selection scheme. The criterion for relay selection is

kept the same for both schemes for fair comparison. Performance evaluation is carried out in

terms of outage probability and average probability of error. We consider both cases separately

as follows.

5.3.1 MULTIPLE RELAY SELECTION

Rayleigh distributed channel coefficients are assumed with their squared amplitudes being

exponential random variables. Therefore, the PDFs of m2 and mI being independent and

exponentially distributed are given by,

mm

epm

1

)(2

and mm

x

mI exp

1

)( J

jjLm 1][ )6.5( a

and the corresponding CDFs are given by,

mm

eP

1)(2

and mm

x

I exP

1)( J

jjLm 1][ )6.5( b

Where m denotes the average second hop SNR for thm relaying link, and m is the average

strength of interference channel from the thm relay and PU.

Given J subsets for selection, the conditional PDF of the SNR of the finally chosen subset sel

according to the proposed relay subset selection scheme where jsel L is given as,

]Pr[]Pr[)(]Pr[]Pr[)()|( 212121 21 JJ ppJpsel

]Pr[]Pr[)( 11 JJJJp )7.5(

64

In order to simplify further analysis of above PDF, we assume that, during a hop transmission,

instantaneous SNRs have same average values for all relays. Hence, M21 .

Using this assumption, 5.7 is rewritten as,

1]Pr[)()|(

Jjjjsel

JpJp

1]Pr[)(

Jjjj

Jp jj )8.5(

In the above equation, first part )( jp

is the PDF of combined SNR of final selected subset

being evaluated at . Since each element in the selected subset jL is exponentially distributed,

thus the PDF of the combined SNR j being the sum of exponential random variables with same

mean will be Erlang distributed and is given by,

1/

)(

1)(

j

j

N

jN

ep

)9.5(

Where jN is the cardinality of selected set jL .

The second part of 5.8 i.e. ]Pr[ jj is the CDF of SNR of thj subset j being evaluated at j .

As mentioned above, SNR of each thj subset follows Erlang distribution, thus the CDF of j will

be expressed as,

nj

N

n

jjj

jj

j neP

1

0

/

!

11]Pr[)( )10.5(

Therefore, the PDF of selected relay subset in 5.8 will take the form as,

11

0

/1/

!

11

)()|(

Jn

jN

nj

N

N jj

j

j

sel ne

N

JeJp

)11.5(

65

An important consideration is that the PDF given above is conditioned over J .i.e. the number of

subsets which are able to satisfy the interference constraints. The value of J may vary from 0 to

S . If ,0J communication between secondary source-destination pair is not possible. This

situation occurs if sum interference threshold thI imposed by PU is too low that no subset of

relays is able to meet the requirement without amplification, thus making secondary

communication impossible. But this is not the case in our scenario as the relay network is

assumed to be far away from the PU. Thus J takes the values between 1 to S . If ,1J there

would be no relay subset selection and if ,2J the destination will decide which relay subset is

the one satisfying the proposed criteria. The interference constraint thI can be satisfied by each

subset in with a probabilitythIP , where, )Pr( th

j

I IIPth

dictates the Erlang distribution

following the same assumption for interference strengths of relayed links i.e.

M21 . Thus, the PDF can be obtained in the same way as,

1/

)(

1)(

j

jN

j

I

II

N

eIp

)12.5(

And the corresponding CDF is written as,

n

thN

n

thj

I

I

neIIP

j

th

1

0

/

!

11]Pr[ )13.5(

Thus, the probability of availability of J subsets out of S subsets which satisfy interference

threshold follows binomial distribution,

JM

I

J

IIJ thththPP

J

SPSJp

)1()(),;( )14.5(

The unconditional PDF of SNR at the destination sel due to selected subset can be found by

using 5.13 and 5.14 in 5.11 as,

66

1/

11

0

/

1 )(!

11)1()();(

j

j

jj

ththsel

N

j

N

Jn

jN

n

S

J

JSI

JIth

j

N

e

nePPJ

J

SIIp

)15.5(

The corresponding CDF can be obtained by integrating the above PDF w.r.t. and using [124,

Eq. (3.381.1)], thus,

,~

!

11)1()(

)(

1)(

11

0

/

1

j

Jn

jN

n

S

J

JSI

JI

j

Nn

ePPJJ

S

NP

jj

ththsel )16.5(

),(~ xa denotes the incomplete gamma function given in [124, Eq. (8.354.1)] as,

kN

k j

k j

kNkxa

0)(!

)1(),(~

The outage event occurs in a communication system if the SNR received at the destination falls

below a set threshold th . The probability of this event can be directly obtained from the CDF of

the received SNR given in 5.16 evaluated at th i.e. )( thout selPP .

Average bit error probability is usually evaluated using the probability of error conditioned over

a given SNR in AWGN. This conditional probability of error is defined in terms of standard Q

function and its average is taken over the PDF of received SNR. Therefore,

0

)()/( dpPPselselee )17.5(

where selsele QP )/( and is a constant and its selection depends on the modulation

scheme employed. Refer to the technique in [125], the above eq. takes the form,

0

22

2

2

1dte

tPP

t

e sel )18.5(

67

Solving the above equation using [124, Eq. 3.461.2], we obtain,

0)(

11

0

/

1 )(

!)!1)(2(

!

)1(

!

11)1()(

)(2

1

kkN

j

j

kJ

nj

N

n

S

J

JSI

JI

je

j

jj

thth

kN

kNknePPJ

J

S

NP

)19.5(

5.3.2 BEST RELAY SELECTION

In order to verify the effectiveness of the proposed multiple relay selection scheme, similar

derivation has been carried out for best relay selection. Based on the same criteria for multiple

relay selection, the relay which is able to maximize the SNR of relay-destination link, while

satisfying the primary interference threshold, is declared the best relay by the destination. Thus,

the optimization problem formulated in 5.5 can be expressed as,

m

m

2max

s.t. thm II )20.5(

Where mdenotes the index of the best relay selected for communication.

In order to investigate the system performance for best relay selection, we follow the same

assumptions for channel conditions as stated in the above section. Thus, the PDFs and CDFs of

m2 and mI will be exponentially distributed as given in 5.6 for each candidate relay satisfying

interference threshold.

Given K relays for selection out of M potential relays, such that, the interference offered by

each thk relay is below the interference level thI set by the PU, the conditional PDF of sel i.e.

the SNR of the final selected relay, where Ksel , is given according to the proposed relay

subset selection scheme as,

]Pr[]Pr[)(]Pr[]Pr[)()|( 22221222212221 2221 KK ppKpsel

]Pr[]Pr[)( )1(222122 KKKKp )21.5(

68

Assuming same average values of instantaneous SNRs for all relays to simplify further analysis,

5.21 can be rewritten as,

122 ]Pr[)()|(

2

Kkkksel

KpKp

1

22 ]Pr[)(2

Kkkk

Kp kk )22.5(

In the above equation, first part )(2

kp

is the PDF of SNR of best chosen relay being evaluated

at .

The second part of the equation i.e. ]Pr[ 22 kk is the CDF of SNR of thk relay being

evaluated at k 2 . Since the SNR of each thk relay follows exponential distribution as mentioned

above, thus the conditional PDF of selected relay using 5.6 will take the form as given by,

1// 21)|(

Kk

selee

KKp

)23.5(

An important consideration is that the PDF obtained in the above eq. is conditioned over K .i.e.

the number of relays which satisfy the interference constraints. The value of K may take any

value from 0 to M . If ,0K communication between secondary source-destination pair is not

possible. This situation occurs if relay network experiences too high interference threshold set by

the PU which is not satisfied by even a single relay, thus making secondary communication

impossible. But this is not the case in our scenario as the relay network is assumed to be far away

from the PU. Thus K takes the values between 1 to M . If ,1K there would be no relay

selection and if ,2K the destination picks up the best relay satisfying the proposed criteria.

Each member of the potential relay set can satisfy the interference constraint thI with a

probabilitythIP ,where, )Pr( thkI IIP

th dictates the exponential distribution and

/1 th

th

II eP

following the same assumption for interference strengths of relayed links i.e.

K21 .

69

Thus, the probability of availability of K relays out of M relays which satisfy interference

threshold follows binomial distribution,

KMI

KIIK ththth

PPK

MPMKp

)1()(),;( )24.5(

The unconditional PDF of SNR sel due to the best selected relay can be found by using 5.6 and

5.24 in 5.23 as,

1/

1

/21)1()();(

KM

K

KNI

KIthk

kththsel

ePPKK

MeIIp

)25.5(

The corresponding CDF can be obtained by integrating the above PDF w.r.t. . The resulting

CDF is,

1/

1

/ 21)1()(1)(

KM

K

KMI

KI

kththsel

ePPKK

MeP

)26.5(

Outage probability can be directly obtained from the CDF of the received SNR given in 5.26

evaluated at th i.e. )( thout selPP .

Average bit error probability using the same technique as employed for multiple relay selection

and using [124, Eq. 3.321.3 ] is given by,

1/

1

21)1()(2/12

1 KM

K

KMI

KIe

kthth

ePPKK

MP

)27.5(

In the next section, the results derived for the single best relay selection and multiple relay

selection have been investigated for well-defined range of certain parameters for the primary and

secondary networks.

70

5.4 SIMULATION RESULTS

This section verifies the effectiveness of the proposed scheme for selecting the subset of relays.

Zero mean unit variance AWGN is assumed for each link. Furthermore, for the relay subset

selection algorithm, M and jN represent the number of potential relays and selected relays

respectively. BPSK with 2 is the modulation scheme employed. The interfering channels

towards the PU are generated by setting 9.0 . Parameter settings are kept the same as

provided in table 4.3 for 0,,, Nfhg mmm and SP for fair analysis.

In Fig. 5.2, the performance of best relay selection, multiple relay selection and all relays has

been compared. SNR achieved at the relay-destination links against different levels of

interference threshold thI . The figure shows that the multiple relay selection algorithm

outperforms both best relay selection and all relay techniques due to freedom of selecting the

best subset of relays which can maximize secondary system performance through controlled

transmit power allocation to the relay network keeping in view the privilege of PUs. Whereas, in

order to allow all relays to participate in transmission, source transmit power needs to be

suppressed keeping to satisfy the interference constraint, which in turn produces negative effect

on the power received at the relay network, eventually decreasing the SNR received at the

destination. Furthermore, a single best relay is also unable to maximize the secondary

performance.

The corresponding total number of selected relays for best and multiple relay selection schemes

is provided in table 5.2. The greater is the number of candidate relays in the potential relay

network, the higher is the flexibility added to the system to allow more relays to participate in the

communication, which are favorable for secondary communication and not harmful for primary

communication at the same time. Thus, the multiple relay selection scheme is the optimal choice

as it provides the optimal combination of relays to maximize secondary performance as

compared to best relay selection and all relay participation.

71

Fig. 5.2: Performance Analysis Of Different Schemes

Table 5.2: Performance Analysis Of Best Relay, Multiple Relay and All Relays Participation

Schemes

M N

Best Relay

Selection

5

-5 9.06 1

0 16.8 1

5 20.5 1

10

-5 9.8 1

0 18.2 1

5 23.3 1

Multiple Relay

Selection

5

-5 15.7 2

0 28.15 3

5 34.12 3

10

-5 17.76 3

0 29.54 4

5 36.23 4

All Relays

5

-5 12.8 5

0 21.5 5

5 28.23 5

10

-5 14.36 10

0 23.75 10

5 31.32 10

-10 -5 0 5 100

10

20

30

40

Ith

(dB)

D(d

B)

Best Relay Selection, M =5

Best Relay Selection, M = 10

All Relays, M = 10

All Relays, M = 5

Multiple Relay Selection, M =5

Multiple Relay Selection, M = 10

)(dBIth )(dBD

72

In Fig. 5.3 and Fig. 5.4, outage probability and bit error rate of the best and multiple relay

selection schemes are investigated by varying the average SNR per hop for different number of

potential relays N , thI and th are both set to 1 respectively [62],[126].

Fig. 5.3: Outage Behavior Of Best And Multiple Relay Selection Schemes

As obvious from Fig. 5.3, the outage probability is maximum for the single best relay selection

and significantly decreases in the case of proposed relay subset selection due to the fact that

spatial diversity enhances system performance by improving SNR received at the destination. An

important observation is the improved system performance in the case of proposed multiple-relay

selection scheme because in order to design an underlay network with full cooperative diversity,

transmit power of the source needs to be suppressed even if the relays just forward the received

signal without any further amplification. On the other hand, in multiple relay selection,

increasing the number of potential relays generates more subsets which are able to satisfy

interference threshold set by the PU, thus giving more freedom to choose the optimal

combination of relays which exhibit good channel conditions towards the destination.

Furthermore, relay selection gives priority to those relays that exhibit good channel conditions

0 4 8 12 16 2010

-6

10-4

10-2

100

(dB)

Po

ut



Multiple Relay Selection, M = 10, Nj' = 5


Best Relay Selection

Relay Subset Selection

73

towards secondary destination and allow them to transmit at high power to improve secondary

throughput. Similar trends are observed in Fig. 5.4 due to the same reasons.

Fig. 5.4: BER Of Best And Proposed Multiple Relay Selection


The major contribution of this chapter is the derivation of the outage probability and bit error rate

for multiple relay selection scheme. For this purpose, a relay subset selection algorithm is

proposed for CRNs operating in an underlay environment near a PU. In this scenario, we select

the optimal combination of relays from the potential relay aiming to maximize the SNR received

at the destination, keeping in view the interference threshold of the primary network. The

proposed scheme proves the effectiveness of multiple relay selection in energy-constrained

CRNs. Finally, the outage probability and average probability of error have been derived in

closed forms through CDF of the received SNR at secondary destination, which has not been

done in literature so far for multiple relay selection. Performance evaluation shows that multiple

relay selection outperforms best relay and all relay techniques. Simulation results recommend

different operating points for the entire system under different levels of interference threshold

and number of potential relays.

0 4 10 12 16 2010

-6

10-4

10-2

100

(dB)

BE

R





Best Relay Selection

Relay Subset Selection

74

Chapter 6

TRANSMIT POWER MINIMIZATION IN

UNDERLAY CRNS

Objective

In this chapter, a relay subset selection algorithm is proposed to select optimal combination of

relays from a potential relay set aiming to ensure minimum QoS requirements at primary and

secondary networks. The proposed scheme considers AF relaying and declares that subset as the

optimal choice after exhaustive search which minimizes the total transmit power at the relay

network while satisfying interference and SNR thresholds of the primary and secondary

networks. Simulation results are provided to prove the effectiveness of relay selection for

underlay CRNs.

6.1 PROBLEM FORMULATION

This chapter addresses the deficiency found in literature in the area of AF-based cognitive radio

networks. All the best and multiple relay selection algorithms found in literature based on

assumptions highlighted in section 2.4.1.3 solve the SNR maximization problem under transmit

power and interference constraints, but no one has solved the problem of relay selection aiming

to minimize the total transmit power at the relay network while guaranteeing to satisfy the

minimum QoS of both primary and secondary networks in underlay CRNs. Thus, the objective

behind the proposed multiple relay selection is to choose the optimal subset of relays which is

able to achieve secondary target SNR theshold th utilizing minimum sum transmit power at the

relay network, while satisfying primary interference threshold thI . Thus the relay subset selection

problem is mathematically expressed as,

75

minimize

m

mP

s.t.

th

m mm

mmD

n

C

21

21

1:1

)1.6(

nm

thm IIIC :2

where, m denotes any relay in the thn subset n that satisfies constraints C1 and C2.

6.2 THE PROPOSED ALGORITHM

Let },,2,1{ Minitial be the initial set of potential relays and P denotes the transmit power

vector of the relay network, i.e. ],,,[ 21 MPPPP . The number of all non-trivial subsets

P is given by

M

m l

MR

1

. The thn subset is denoted as n , where })1.,,.........2,1{( Rn .

The cardinality of subset || n is denoted as M where MM .

Relay subset selection works as follows. First interference power I due to eachthn subset of

relays towards the PU is computed where interference power mI due to thm relay in any subset is

given by, 2|| mmm fPI . Let K be the number of subsets out of R, such that for every subset

Kkk 1

, interference threshold thI towards the PU is satisfied. The interference constraint forthk

such subset can be given as,

th

m m

mmik IfPII

k k

2|| Kk .,,.........2,1 )2.6(

In a similar way, P number of subsets, given by Ppp 1

are selected out of R which are

able to satisfy SNR threshold thI at the secondary destination. The SNR constraint for thp such

subset can be given as,

76

th

mpm

pm

pm

pmp

D

p

21

21

1 Pp .,,.........2,1 )3.6(

After having two types of subsets, i.e. Pp 1 which achieves SNR threshold th and Kk 1 which

satisfies interference threshold thI , the intersection of PpK

k 11 is performed to extract all

matching subsets having same number of relays and same index position of each relay in the

matching subsets. Finally, that subset is declared as the selected subset from the matching

subsets which consumes minimum sum transmit power at the relay network according to the

proposed criterion.

Summarizing, the proposed algorithm initializes the transmit power of each relay in the potential

relay set followed by performing exhaustive search on

M

m l

M

1

combinations of relays. The aim

is to sorting out all possible subsets of relays which satisfy interference threshold thI and target

SNR threshold th at the same time. Finally, the selected subset consumes the minimum transmit

power out of all subsets. The pseudocode of the proposed algorithm is shown in table 6.1.

Table 6.1: The Proposed Algorithm

:INPUTS initialmmmsinitialththS mfhgNNMIP ,,,,,,,, 0

)||()( 0

22 NgPAP mSmm max0 PPm initialm

P PofsubsetstrivialnonAll //

th

m m

mmm

k IfPIIk k

2||

K

kkwhere1

,

th

mpm

pm

pm

pmp

D

p

21

21

1

P

ppwhere1

,

PpK

km 11 indicessameatPhavingsubsetsmatchingallcontain mm //

imumisinPts sel

m

mmsel

sel

min..

initials N relaysofsubsetselectedthedenotesS//

Selm

ms POUTPUTS ,:

77

6.3 SIMULATION RESULTS

In this section, we will prove the effectiveness of our proposed multiple relay selection scheme

to achieve cooperative diversity while satisfying minimum QoS requirements of both the primary

and secondary networks. The channel coefficients are kept the same as provided in table 4.3.

Furthermore, M and N represent the number of potential relays and selected relays

respectively. In Fig. 6.1, the behavior of the relay network is evaluated in terms of total transmit

power required for different levels of SNR threshold and different number of available relays.

SP and thI are set to dB10 and dB0 respectively. The figure shows that total transmit power

required at the relay network increases significantly by increasing the SNR threshold th , since

more power is required to satisfy high SNR threshold. Furthermore, the higher is the number of

available relays, the lower is the total transmit power required by the selected relay subset. This

is due to the fact that the total number of non-trivial subsets R increases by increasing the

number of potential relays M , which in turn increases the probability of selecting those relays

which exhibit good channel conditions towards the destination than the PU, thus reducing the

total transmit power required at the relay network to meet th while satisfying .thI

Fig. 6.1: Transmit Power Allocation to Relay Network keeping dBI th 0

0 2 4 62

3

4

5

6

th(dB)

Pt(d

B)

M = 10

M =5

78

Table 6.2 provides total transmit power required at different interference threshold levels. The

relaxation in the value of thI not only generates more matching subsets from PpK

k 11 but

also gives more freedom to the relays to communicate at high transmit power.

Table 6.2: Total Transmit Power required For Different Values of

M

N N

0 5 2 3.06 2 3.56

10 2 2.98 2 3.87

2 5 2 3.89 2 4.25

10 2 3.64 2 4.92

4 5 2 4.79 2 5.39

10 2 4.54 3 5.12

6 5 3 5.95 3 6.65

10 2 5.72 3 6.95

Fig. 6.2 demonstrates the behavior of the proposed algorithm under different levels of source

transmit power with corresponding no. of selected relays provided in table 6.3. At low values of

source transmit power, higher amplification is required at the relay network to satisfy the SNR

threshold of the destination. As the transmit power of source increases, transmit power of the

selected subset of relays decreases, since less amplification is required at the relay to satisfy the

SNR threshold. However, at very high values of source transmit power and low interference

threshold level, relay selection problem becomes critical, as the higher is the power received at

the potential relay subset, the higher is the power transmitted by the relays with even less

amplification performed, due to which the interference power experienced by the PU increases

significantly.

dBI th 5)(dBPt

dBI th 0)(dBth )(dBPt

MandI thth ,,

79

Fig. 6.2: Transmit Power Allocation To Relay Network Considering

Table 6.3: Transmit Power Allocation To Relay Network For

M N

5

0 3.77 3

5 3.28 2

10 3.06 2

15 2.65 1

10

0 3.69 2

5 3.14 2

10 2.98 2

15 2.54 1


A multiple relay selection algorithm is proposed for CRNs operating in underlay environment in

the vicinity of a PU, aiming to select an optimal combination of relays from the potential relay

set, which consumes minimum transmit power while satisfying interference threshold and SNR

threshold of the primary and secondary network respectively.

0 5 10 152.5

3

3.5

4

PS(dB)

Pt (

dB

)

M = 5

M = 10

dBdBI thth 0,0

dBdBI thth 0,0

dBdBI thth 0,0

)(dBPS )(dBPt

80

Chapter 7

PERFORMANCE ENHANCEMENT OF

CRNS USING FUZZY RULE BASED

SYSTEM

Objective

In this chapter, Fuzzy Logic has been used to solve the problem of performance enhancement of

secondary communication in underlay spectrum sharing environment. FRBS assisted relay

selection and transmit power allocation (RSTPA) schemes exploit the degrees of freedom

involved in fuzzy logic to solve the highlighted problems of performance enhancement in CRNs.

Standard Mamdani fuzzy control is used for this purpose. The proposed schemes for SNR

maximization and transmit power minimization assume the availability of perfect CSI and select

the best combination of relays keeping in view the objectives and constraints. Furthermore, the

proposed schemes have been investigated for a well-defined range of certain parameters.

7.1 SNR MAXIMIZATION

In this section, two algorithms based on FL have been proposed to solve the modified problem of

SNR maximization formulated in chapter 5 for underlay cognitive relay networks. For the sake

of convenience, the multiple relay selection and power allocation problem formulated in section

5.1 is again expressed here as,

Sm

m22max )(1.7 a

s.t. th

m

m III

S

)(1.7 b

81

where, S denotes the selected set of relays.

The algorithms proposed to solve this constrained problem are explained as follows.

7.2 FRBS ASSISTED SYSTEM DESIGN 1

A novel idea of FRBS-assisted RSTPA system in AF-based cognitive relay networks over

Rayleigh flat-fading channels has been proposed in this section. The proposed scheme comprises

two FLSs as shown in Fig. 7.1.

Fig. 7.1: The Proposed System Design 1

82

As shown in the Fig. 7.1, FLS 1 takes mm hP , and mf corresponding to each thm relay as inputs

and the fuzzy inference engine computes two outputs, one is the SNR m2 achieved on the link

from thm relay to destination, and the other is the interference mI offered to the PU at the thm

relay link corresponding to each rule. The consequents of FLS 1 are then fed to FLS 2 to

determine the relay selection factor RSF for each thm relay.

7.2.1 MAMDANI FUZZY CONTROL

The basic structure of FLS is already explained in chapter 3. The fuzzy rule based system to

solve the SNR maximization problem is explained below.

7.2.1.1 Fuzzification

Fuzzification represents the entire range of each antecedent (input variable) and the consequent

(output variable) through distinguishing fuzzy sets, with each fuzzy set assisted by a MF. The

MFs of the antecedents and consequents of FLS 1 and FLS 2 are given in Fig. 7.2.

83

84

Fig. 7.2: MFs Of The Antecedents And The Consequents Of FLS 1 And FLS 2

Triangular MFs are used to represent the antecedent mf , whereas, for all other cases of

antecedents and consequents, trapezoidal MFs are chosen for minimum and maximum levels and

triangular MFs are used to represent medium levels. The fuzzy sets for the final consequent, the

RSF, taken from FLS 3 are explained as follows. The output variables are NS (Not Selected),

WCS (Weak Consideration for Selection), CS (Consider for Selection), SCS (Strong

Consideration for Selection), and, S (Selected). Same MFs for RSF will be used in all algorithms

proposed in this chapter.

85

7.2.1.2 Rule Based Decision

Referring to the MFs shown in Fig. 7.2, three fuzzy sets are used for the first input variable mf ,

four fuzzy sets for the second input variable mh , and six fuzzy sets for the third input variable mP .

Thus, there are 72643 “IF-THEN” rules for FLS 1 of the form,

NSisISisTHENfisfhishPisPIFR llllll

121314161 ,,,,,: Ll ,,2,1

where L denotes the total number of rules for FLS 1. The outputs DRm and mI for each thm

relay from FLS 1 are then fed to FLS 2 to determine the consequent Relay Selection Factor

(RSF) of each thm relay. In FLS 2, there are 2555 “IF-THEN” rules of the form,

SisRSFTHENSisISCSisIFR kkkk

1121 ,,,: Kk ,,2,1

where K represents the total number of rules for FLS 2.

The rule base contains rules for every possible combination of the input space. Standard

Mamdani Inference Engine (MIE) serves the purpose.

7.2.1.3 Defuzzifier

The fuzzy variables produced by FIE are converted into the crisp values by the Defuzzifier to

represent the actual output of the system. Center Average Defuzzifier (CAD) is preferred

because of its computationally simplicity. After applying CAD method, the relationship of the

consequents m2 and mI with the antecedents mP , mh and mI in FLS1 are shown by the rule

surfaces in figure 7.3(a), whereas, figure 7.3(b) shows the rule surface for RSF according to rule-

based decisions taken in FLS2.

86

Fig. 7.3 (a): Rule Surface For FLS 1

Fig.7.3 (b): Rule Surface For FLS 2

Fig. 7.3: The Rule Surface

87

7.2.2 THE PROPOSED ALGORITHM

According to the proposed algorithm, FLS 1 computes m2 and mI based on the CSI available at

each relay, followed by FLS 2 which assigns RSF to each relay on the basis of consequents of

FLS 1. The relays are picked up in the descending order of RSF while ensuring to satisfy the

interference threshold thI . Finally, the fine tuning of transmit power of the relay with highest RSF

is carried out as it exhibits the best channel condition towards the destination and offers the

minimum interference to the nearby PU. The flowchart of the proposed algorithm is shown in

Fig. 7.4.

m = 1

Initialize transmit

power of mth

relay

FLS1 computes

m2 and mI

FLS2 computes RSF of mth

relay based on m2 and

mI

m = m + 1

m=M?

k = 1

Pick the relay with highest

RSF (excluding previous

selection)

interference

constraint

satisfied?

k = k + 1

k = M?

Fine tuning of transmit

power of selected relays

Outputs:

sm

ms 2,

Yes No

Yes

No

No

Start

Yes

Fig. 7.4: Flow Chart of the Proposed Fuzzy

Rule Based RSTPA Design

88

7.2.3 SIMULATION RESULTS

The performance of the proposed schemes is presented in this section. The channel gains mm hg ,

and mf are taken according to table 4.3. Noise variance is assumed to be 1.0 . Fig. 7.5 illustrates

the performance of the proposed scheme in terms of total SNR of relay-destination link 2 given

in 7.1 achieved for different interference thresholds levels. Two different cases of total number

of potential relays are considered which are taken to be }10,5{M .

Fig. 7.5: SNR Performance Of The Proposed Scheme

From Fig. 7.5, it is observed that SNR 2 increases significantly by increasing the interfernce

threshold thI due to the fact that relaxing thI allows the relays to transmit at high power thus

enhancing secondary network’s performance. Furthermore, another strong observation is that, as

the relay network grows, there is considerable improvement in the SNR, as large relay network

adds more flexibility to choose those relays according to FRBS, which exhibit favorable channel

conditions towards the destination and at the same time are not harmful for the primary

communication. Table 7.1 shows the corresponding number of selected relays for different cases

considered in figure 7.5(a). Similar trends are observed here by relaxing interference threshold,

-5 -3 -1 1 3 512

14

16

18

20

22

24

26

Ith

(dB)

2(d

B)

M = 10

M = 5

89

but higher SNR is achieved, due to FRBS assisted relay selection which intelligently assigns a

RSF to each individual relay to make reliable decision about relay selection.

Table 7.1: Corresponding Number Of Selected Relays

In Fig. 7.6, the proposed FRBS is extended to evaluate 2 given in 7.1 against different levels of

source transmit power. Interference threshold thI is set to 0dB. We consider the system’s

response for different number of potential relays }15,10,5{M . The obtained results show the

improvement in received SNR for low and medium levels of source transmit power, but SNR

decreases at high transmit power levels. This is due to the reason that the higher is the power

received at the relay network, the higher is the total interference offered by the relay network to

the PU, which makes the relay selection problem difficult and reduces the number of selected

relays, thus resulting in performance degradation.

Fig. 7.6: SNR Performance For Different Source Transmit Power Levels

M N

5

-5 11.3 2

0 13.7 3

5 18.2 3

10

-5 11.7 2

0 14.3 3

5 18.7 4

)(dBIth )(dBD

0 5 10 15 204

6

8

10

12

14

PS (dB)

2 (dB

)

M = 5, Ith

= -5dB

M = 10,Ith

= -5dB

M = 15, Ith

= -5dB

90

Thus, in energy-constrained relay-assisted networks, where the interference threshold of the

primary network limits the transmit power of the relays, increasing the transmit power of the

source does not enhance the communication performance.

In Fig. 7.7, the performance of the proposed scheme is validated by comparison with the greedy

scheme [65] for multiple relay selection. The comparison has been carried out in terms of total

end-to-end SNR D . All parameter settings are done according to [65]. The figure shows

obvious performance difference and prove that the proposed scheme is able to achieve high total

end-to-end SNR as compared to the greedy scheme. This improvement is achieved due to two

strong factors. First, using FRBS assisted RSTPA design of the secondary system which checks

every possible combination of relay-destination and relay-PU channel conditions and picks the

best combination of relays to enhance secondary performance while operating in underlay mode.

Second, the flexibility in the transmit power allocation to the selected relays with the priority

given to the relay with the highest RSF obtained from FRBS keeping in view the interference

threshold. On the other hand, in the greedy scheme, either the relay does not participate in the

communication or transmits at fixed powerSm PP , thus, there is no flexible transmit power

allocation for the selected relays to for enhancement of secondary communication. However, in

high power areas, the SNR decreases in both cases due to the same reason mentioned in the

comments given for Fig.7.6.

Fig. 7.7: Comparison of the Proposed Scheme and the Greedy Scheme

0 5 10 15 20 252

4

6

8

10

12

PS (dB)

D (

dB

)

Proposed Scheme, M = 5, Ith

= 10

Greedy Scheme, M = 5, Ith

= 10

91


A novel fuzzy rule based RSTPA scheme is proposed to enable the secondary communication in

the frequency band of PUs. Once the CSI based knowledge is pumped into the FRBS, it provides

the optimum RSF upon offering the three parameters that are power received at each relay, relay

to PU interference channel gain and relay to destination channel gain. The RSF obtained from

the FRBS, is used as a priority factor that highlights the relays having ability to enhance

secondary system performance and are less harmful to the PUs. It is proved that the proposed FL

based relay selection scheme outsmarts an existing scheme in literature for multiple relay

selection.

7.3 FRBS ASSISTED SYSTEM DESIGN 2

The second FRBS-assisted RSTPA technique to solve the SNR maximization problem works as

follows. The proposed scheme uses three-stage FL system as shown in Fig. 7.8. This FRBS

assisted design is more complex than the design 1, since it contains three FLS module in cascade

and makes decisions not only on the bases of m2 and mI , but also includes the effect of m1 to

assign relay selection factor.

Fig. 7.8: Proposed System Design 2

As the figure shows, the proposed design, being a three-stage FLS is a bit computationally

complex, however, decisions taken in FLS1 are based on single antecedent (input), thus reducing

the overall computational cost involved. The three Mamdani fuzzy control modules FLS1, FLS2

and FLS3 are shown in Fig. 7.9.

FLS 3 hm

fm

Pm

γ1m

γ2m

Im

FLS 1 FLS 2 gm

RSFm

92

Fig. 7.9: Proposed FLS Modules

93

FLS1 takes the thm source-relay channel coefficient mg as input, and computes power mP

received at the corresponding relay and SNR of the first hop (source-relay link) m1 . FLS 2

follows FLS1 and takes three inputs for each thm relay; mP obtained from FLS1, relay-

destination channel coefficient mh and relay-PU channel coefficient mf . The outcomes of FLS2

are SNR of second hop (relay-destination link) m2 and the interference power mI of each thm

relay. Finally, FLS3 assigns RSF to each thm relay based on the consequents of FLS1 and FLS2.

The RSF is assigned to each relay to make a decision about its selection, aiming to give the

highest priority to those relays in the relay selection procedure, which exhibit the ability to

maximize SNR on the corresponding link with minimum amplification, and at the same time are

less harmful to the PU.

Thus, the selection of a particular thm relay depends on the following factors:

i. SNR achieved at the source-relay link m1

ii. SNR achieved at the relay-destination link m2

iii. Interference offered to the PU mI


FRBS to solve the SNR maximization problem is explained below.


Fuzzification represents the entire range of each antecedent (input variable) and the consequent

(output variable) through distinguishing fuzzy sets, with each fuzzy set assisted by a MF. The

entire range of each linguistic variable is covered with sufficient number of fuzzy sets for

accuracy and reliability of the system as shown in Fig. 7.10. Triangular MFs represent each

fuzzy set.

For FLS1, twenty three fuzzy sets are used for the antecedent mg and eight fuzzy sets each are

used for the consequents mP and m1 as shown in Fig. 7.10(a).

94

Fig. 7.10(a): MFs For Antecedents And Consequents of FLS1

95

The fuzzy sets for the inputs to FLS 2 are shown in Fig. 7.10(b). Fuzzy sets for the input mP are

shown earlier, whereas, six fuzzy sets cover the entire range of the input mh and three fuzzy sets

represent the input mf . For the outputs, eight fuzzy sets are used for mI , and seven fuzzy sets are

used for m2 .

96

Fig. 7.10(b): MFs For Antecedents And Consequents of FLS2

Since FLS 3 is fed from the consequents of FLS 1 and FLS 2, the fuzzy sets for the antecedents

mmI 1, and m2 of FLS 3 are shown in Fig. 7.10(a) and Fig. 7.10(b), where the fuzzy sets for the

final output RSF are the same as shown in Fig. 7.2.


Since there is a single antecedent for FLS1, there are 23 “IF-THEN” rules with each thl rule of

the form,

1111 ,,,,: BisandPisPTHENGisgIFR l

m

l

m

l

m

l Ll ,,2,1

where, L denotes the total number of rules for FLS 1.

There are 144386 “IF-THEN” rules for FLS2, with eachthn rule of the form,

424624 ,,,,,,: SisandIisITHENPisPFisfHishIFR n

m

n

m

n

m

n

m

n

m

n Nn ,,2,1

where, N represents the total number of rules for FLS 2.

For FLS 3, there are 448788 “IF-THEN” rules for FLS3, with each thp rule of the form,

SisRSFTHENIisISisBisIFR p

m

p

m

p

m

p

m

p ,,,,: 17261 Pp ,,2,1

where, P represents the total number of rules for FLS3.

97

7.3.1.3 Defuzzifier

After applying CAD method, the relationship of the consequents of each FLS with corresponding

antecedent is shown by the rule surfaces in Fig. 7.11(a) and 7.11(b) respectively, whereas, Fig.

7.11(c) shows the rule surface for RSF according to rule-based decisions taken in FLS2.

Fig. 7.11(a): Rule Surface For FLS 1

98

Fig. 7.11(b): Rule Surface For FLS 2

Fig. 7.11(c): Rule Surface For FLS3


Fig. 7.11 (c) clearly indicates that, the selection factor of that relay is the highest, which exhibit

mm 12 , thus requiring less amplification to approach upper bound and so less harmful for the

99

primary communication. RSF goes on decreasing for other values of m1 and m2 , and RSF for a

particular relay goes to zero for mm 21 , since very high amplification is required in this case

to achieve even lower bound of SNR, thus resulting in high interference towards the PU.


The algorithm proposed for secondary performance enhancement based on FL is explained with

the help of flow chart in Fig. 7.12 below.

Fig. 7.12: Flow Chart Of The Proposed Algorithm

m = 1

FLS1 computes Pm

and γ1m based on gm

FLS3 computes RSFm

based on γ1m, γ2m and Im

FLS2 computes γ2m and Im

based on Pm, hm and fm

m = m + 1

m = M ?

j= 1

Pick the relay with highest

RSF (excluding previous

selection)

interference

constraint

satisfied?

j = j + 1

j= M ?

Fine tuning of transmit

power of the selected relays

Outputs:

selj

msel 2,

Yes No

Yes

No

No

Yes

Start

100


The performance of the proposed FRBS assisted relay selection technique is evaluated to select

the best possible combination of relays, which maximizes secondary communication

performance, in such a way that minimum amplification is required at the intermediate relay

network and at the same time, the primary interference constraint is not violated. Same parameter

settings provided in section 7.2.3 are maintained for the purpose of comparison of both schemes.

Fig. 7.13 illustrates the SNR performance of the fuzzy rule based relay selection algorithm

against different levels of interference threshold thI . The system response is observed for three

different sizes of the intermediate relay network, i.e. 10,5M . It is observed from that figure

that the total SNR 2 achieved at the selected relay-destination links significantly increases as the

interference threshold is relaxed, thus allowing the relays to transmit at high power. Furthermore,

adding more relays to the network considerably increases 2 , because more options are available

for FRBS to select those relays which can efficiently increase SNR by increasing signal

diversity, thus enhancing the overall system performance.

Fig. 7.13: Performance Of The Proposed Scheme Vs Interference Threshold thI

-5 -3 -1 1 3 512

14

16

18

20

22

24

26

28

Ith

(dB)

2(d

B)

M = 10

M = 5

101

Table 7.2: Total No. of Selected Relays

Table 7.2 shows the corresponding number of selected relays. The nu1mber of selected relays

not only increases by relaxing the interference threshold, but also by increasing the number of

candidate relays, since more candidates present more choices for FRBS to pick the best possible

combination. Algorithm 2 outsmarts algorithm 1 because it takes into account SNR of source-

relay and relay-destination links alongwith interference offered to the PU by each relay and

assigns RSF to each individual relay.

7.4 COMPARISONS OF THE PROPOSED ALGORITHMS

Now we compare the performance of the multiple relay selection problems proposed in chapter 4

and in this chapter. For this purpose we compare the SNR achieved by the relay selection

schemes proposed in sec. 4.5.1 and 4.5.3 with the algorithms proposed in sec. 7.2 and 7.3

respectively. The observed results are provided in table 7.3.

Table 7.3: Comparisons of Proposed Algorithms in Ch. 4 and Ch. 7

M N

5

-5 11.3 2

0 13.7 3

5 18.2 3

10

-5 11.7 2

0 14.3 3

5 18.7 4

M = 10

Interference

Threshold

No. of Selected Relays

Algorithm 1 Algorithm 2 Algorithm 1 Algorithm 2

without

FL

with FL without

FL

with FL without

FL

with FL without

FL

with FL

11.7 12.1 9.2 12.8 2 2 2 2

14.3 15.1 11.3 16.2 3 3 2 3

18.7 20.3 20.8 24.1 3 4 3 4

)(dBIth )(dBD

dBIth 5

dBIth 0

)(dBD

dBIth 5

102

As the table shows, Algorithm 1 solved with fuzzy logic (sec. 7.2) outsmarts the algorithm

proposed in sec 4.5.1. Similarly, algorithm 2 proposed in sec. 7.3 outsmarts the algorithm

proposed in sec. 4.5.3.


The proposed FRBS assisted design 1 and 2 takes the information about incoming and outgoing

channel coefficients for each relay into account, and allocates a priority factor or RSF to each

relay. The RSF indicates the preference in which the relays are selected, aiming to maximize the

secondary performance in an underlay spectrum sharing environment. The effectiveness of the

proposed scheme is highlighted through the simulation results, which shows that the design 2

achieves high SNR as compared to design 1, although more computational complexity is

involved in design 2. The better results are obtained because the priority factor is assigned to

each thm relay on the bases of SNR of corresponding source-relay and relay-destination links and

the interference offered by the thm relay, which results in more sophisticated decision making.

Furthermore, it is proved that FL based relay selection outsmarts ABC optimization.

7.5 TRANSMIT POWER MINIMIZATION

This section solves the problem of transmit power minimization in underlay cognitive relay

networks using FL. For this purpose, a FRBS is proposed for intelligent relay selection such that

sum transmit power at the relay network is minimized, while achieving the desired SNR at the

destination and keeping the primary communication undisturbed. For the sake of inconvenience,

the mathematical formulation of the relay subset selection problem is again expressed as,

..ts

thm

m

m hP

S

22 ||

thm

m

m IfPIS

2|| 2.7

Sm

mPmin

103

where, th and thI denotes the SNR threshold and the interference threshold respectively. s

denotes the selected set of relays, and the cardinality of the selected subset S is MN .

7.5.1 THE PROPOSED FLS DESIGN

The FLS proposed to solve the constrained problem is explained as under. FRBS is employed to

choose the best combination of relays to enable the secondary communication in an underlay

environment while consuming minimum overall transmit power. The proposed Mamdani-based

FLS is shown in Fig. 7.14.

Fig. 7.14: Proposed FRBS assisted Design

As Fig. 7.14 shows, the power mP transmitted by the thm relay in the relay network and the ratio

of relay-destination channel coefficient relative to the relay-PU channel coefficient act as the

antecedents and the fuzzy inference engine computes a RSF for each candidate relay based on

the fuzzy rules.


The phases involved in the Mamdani fuzzy control are explained as under.

104


The fuzzy sets for each input and output variable is shown in Fig. 7.15. To make the whole

design sophisticated and accurate, the whole range of the inputs and the output are covered with

sufficient number of fuzzy sets. Fifteen fuzzy sets are used for the both antecedents mP and m

m

f

h

and five fuzzy sets for the consequent mRSF , with trapezoidal MFs for the minimum and

maximum levels and triangular MFs for the intermediate levels. Refer to Fig. 7.2 for the MFs of

RSF.

Fig. 7.15: Fuzzy Sets for the Antecedents and the Consequent

105


There are 2251515 “IF-THEN” rules with each thk rule of the form,

SisRSFTHENGisf

hPisPIFR k

k

m

k

mkk

115131 ,,,: Kk ,,2,1

where, K represents the total number of rules.

7.5.2.3 Defuzzifier

The rule surface specifying the relationship between each input and the output is shown in Fig.

7.16 below.



FRBS assisted relay selection scheme works as follows. The inputs to the FRBS are the transmit

power mP of each relay without amplification, and the ratio of coefficients mh and mf , and RSF is

assigned to each thm candidate relay. Let ],,,[ 21 MRSFRSFRSFR be the RSF vector of the

potential relay network obtained from FRBS. The relays are then picked up in the descending

order of the RSF in R . The flowchart of the proposed technique is shown in Fig. 7.17.

106

Fig. 7.17: The Flow Chart of the Proposed Algorithm


Fig. 7.18 illustrates the behavior of the proposed scheme for different levels of interference

threshold thI . SNR threshold th is set to 1. Source transmit power is set to 10. We consider three

1m

Initialize transmit

power of thm relay

FRBS computes RSF of thm

relay based on

m

m

f

hand mP

1mm

?Mm

Outputs:

s

m

m

DRs ,

Yes No

Start

Pick the relays in the descending

order of RSF and perform fine

tuning of transmit power of each

selected relay to satisfy

constraints

107

different sizes of the potential relay network, i.e. 15,10,5M . As Fig. 7.18(a) shows, overall

transmit power of the relay network increases by relaxing the interference threshold thI , because

high levels of thI enable the relays to freely transmit at high power to meet the QoS requirements

of the secondary network without interfering the primary signals. However there is another

strong observation that as the potential relay network grows in size, the total transmit power

decreases. This is due to the reason that a large relay network provides more opportunities to

FRBS to intelligently pick those relays, which are capable of enabling the secondary

communication performing minimum amplification, at the same time not harmful for the primary

communication. Fig. 7.18(b) shows the total number of selected relays corresponding to different

cases of Fig. 7.18(a). As the figure shows, full cooperative diversity is observed for the

interference threshold level dBI th 10 .On the other hand, setting the interference threshold too

low makes the relay selection problem very difficult and a single best relay is selected to

participate in the communication. Thus, relay selection will become impossible below this level

of interference threshold.

Fig. 7.18(a): Total Transmit Power vs Interference Threshold thI for 1th

-10 0 10 200

5

10

15

20

Ith

(dB)

To

tal T

ran

sm

it P

ow

er

(W)

M = 15

M = 10

M = 5

108

Fig. 7.18(b): Corresponding Number Of Selected Relays

In Fig. 7.19, the proposed scheme is analyzed by setting different SNR thresholds for the

secondary network while the interference threshold is set to dB0 . Transmit power of the source

is kept the same as in the case of Fig. 7.18. Again the performance is evaluated considering

three different sizes of potential relay networks, i.e. 15,10,5M . As the figure shows, the

transmit power of the relay network increases when SNR threshold is increased, because in order

to satisfy high SNR threshold, high transmit power is required at the intermediate relay network.

Fig. 7.19: Total Transmit Power Vs SNR Threshold th

-10 0 10 200

4

8

12

16

Ith

(dB)

Nu

mb

er

of S

ele

cte

d R

ela

ys

M = 5

M = 10

M = 15

0.5 1 1.5 27

8

9

10

th

To

tal T

ran

sm

it P

ow

er

(W)

M = 5

M = 10

M = 15

109

Furthermore, only slight variations in the SNR threshold were possible in order to study the

system response, owing to the fact that underlay environments does not allow the SUs to transmit

at high power keeping in view the privilege of the PUs. Thus, setting too high SNR threshold

eventually makes the secondary communication impossible. Finally, for large potential relay

networks, total transmit power significantly decreases, because a large network provides more

opportunities to the FRBS to intelligently pick the best combination of relays to satisfy the QoS

requirements of both primary and secondary networks while saving the resources available at

each node.

Fig. 7.20 extends the proposed FRBS assisted relay selection system design to study the behavior

of the relay network for different cases of source transmit power. thI and th are set to dB10 and1

respectively. As observed from the figure, the total power transmitted by the relay network

significantly decreases by increasing the transmit power of the source because the higher is the

power received at the intermediate relay, the lower is the amplification required at each selected

relay to satisfy the SNR threshold.

Fig. 7.20: Total Transmit Power Of For Different Source Transmit Power Levels Keeping

dBI th 10 And 1th

0 5 10 15 20 2510

12

14

16

18

20

PS (dB)

To

tal T

ran

sm

it P

ow

er

(W)

M = 15

M = 10

M = 5

110


A FRBS is proposed to solve multiple relay selection problem for underlay cognitive relay

networks. Proposed FRBS takes the CSI of each candidate relay in the potential relay network as

an input and assigns RSF to each relay, aiming to enable the coexistence of the primary and the

secondary networks with minimum QoS requirements, while utilizing the minimum transmit

power at the relay network. The simulation results confirm the effectiveness of the FRBS-

assisted relay assignment scheme.

111

Chapter 8

DETECTION AND ESTIMATION OF

MULTIPLE FAR-FIELD PRIMARY

USERS USING SENSOR ARRAY

Objective

In this chapter, an efficient, reliable and low-complexity spectrum sensing scheme is proposed

for CRNs which not only detects the number of active PUs but also estimates their parameters

such as frequency, power strength and DOA. The proposed scheme is based on GA as global

optimizer hybridized with PS as local optimizer. The system model used for this purpose

constitutes a uniform linear array of sensors. Fitness function is derived from Maximum

Likelihood (ML) principle and defines the MSE between actual and estimated signals. The

effectiveness of the proposed scheme is proved under low SNR conditions. Far-field

approximation is assumed and the signals are detected in the frequency band of 80MHz-

108MHz. The snapshots are available to us after 10-15 seconds.

8.1 BACKGROUND

As explained earlier, spectrum sensing [127] is a process conducted to become aware of the

status of the spectrum usage which involves detection of active signals then estimation of the

signal parameters, followed by decision but it has revamped as a very active area of research

with the advent of cognitive radio technology [128]. In CR, spectrum sensing is a decision

making technique in which SUs are required to dynamically detect spectrum holes to become

aware of the presence of the PUs which have high priority being the licensed users. Being the

core component of CRN, spectrum sensing faces many challenges [129] in terms of hardware

112

requirements, hidden terminal problem, detection of spread spectrum primary users,

data/decision fusion in scenarios of cooperative sensing, multipath fading, noise power

uncertainty, implementation complexity, security etc. In order to meet these challenges

efficiently, spectrum sensing requires innovative techniques for not only detecting the number of

PUs but also estimating their amplitudes, DOAs and frequencies to avoid interference between

primary and secondary transmissions.

8.2 SPECTRUM SENSING METHODS

A number of spectrum sensing methods to detect spectrum holes in CRs have been proposed in

literature which have been broadly categorized into three main classes: Non-cooperative

spectrum sensing [130], cooperative spectrum sensing [131]-[132] and interference based

spectrum sensing [133].

8.2.1 NON-COOPRATIVE SPECTRUM SENSING

Non-cooperative spectrum sensing also known as transmitter detection is further classified into

Energy Detection (ED), Matched Filter Detection (MFD) and Cyclostationary based Detection

(CBD).

8.2.1.1 Energy Detector

Energy Detector [134] is the most widely studied spectrum sensing technique because of its less

complexity and no requirement of prior knowledge of PU signal, but it is accompanied by a

number of shortcomings which include noise power uncertainty, poor performance under low

SNR and inefficient to detect spread spectrum signals.

8.2.1.2 Matched Filter Detector

MFD [135] is considered as the optimum method of signal detection when perfect knowledge of

PU is available otherwise it performs poorly. Implementation complexity of MF is impractically

large because it demands CR to have dedicated receivers for all signal types.

113

8.2.1.3 Cyclostationary Detector

CBD [136] relies on the prior knowledge of PU signals and exploits cyclostationary features of

the received signals, hence it is capable of differentiating PU signals and noise. Its

implementation complexity lies between energy detector and matched filter.

8.2.2 INTERFERENCE BASED SPECTRUM SENSING

The focus of interference-based spectrum sensing is to design the CRNs to operate in underlay

spectrum sharing environment. In this method, SUs do not perform spectrum sensing to find

spectrum opportunities rather they identify spectrum occupancy status of PUs and an interference

power threshold is set up for SUs towards PUs for a particular frequency band and location.

8.2.3 COOPERATIVE SPECTRUM SENSING

In cooperative spectrum sensing, SUs collaborate and share sensing information to solve

problems like hidden terminal problem, receiver uncertainty and multipath fading at the cost of

increased detection delay and high implementation complexity due to requirement of control

channels efficient information sharing algorithms.

8.3 SOURCE LOCALIZATION

Source localization by means of sensor arrays has been one of the fundamental and effective

ways to estimate amplitude, frequency, DOA and range estimation of both far and near field

sources upto high accuracy in many systems including radar, navigation and wireless

communication systems. In order to achieve optimum performance of a sensor array [137], array

geometry, the number of sensors and the physical separation between the sensors are critical

design parameters in addition to the number of other factors including signal-to-noise ratio.

Many algorithms have already been proposed in array signal processing for source localization

which can be categorized into far-field source localization and near-field source localization on

the basis of range between the radiating source and the array of sensors.

114

Far-field source localization algorithms make assumption that sources are located in the far-field

region of the array. Thus each signal arriving at the array has planar wavefront. ESPRIT

algorithm [138] and MUSIC algorithm [139] are among the widely studied far-field source

localization algorithms. However, the far-field assumption is no longer valid when the sources

are located close to the array and are described by spherical wavefront assumption, thus range

parameter is also included in addition to amplitude, DOA and frequency to characterize the

sources. A number of techniques have been proposed in this area such as 2-D MUSIC [140],

Linear Prediction method [141], higher order ESPRIT-method [142] etc but most of these

algorithms are computationally complex.

8.4 CONTIBUTION OF THE CHAPTER

This dissertation addresses the problem of detecting the number of active PU signals and then

estimating their signal parameters to ensure interference free communication in CRNs. Most of

the existing techniques to determine the number of sources are based on the Singular Value

Decomposition SVD of the covariance matrix of the snapshots which yields M distinct

eigenvalues, where M is the number of signals present and the remaining eigenvalues are either

zero or non-zero repeated eigenvalues [143] or non-zero eigenvalues less than threshold.

However, SVD has high uncertainty in terms of decisions about setting of the threshold and so

different schemes [144] have been proposed for threshold setting to detect the presence of

signals. These include Maximum Eigenvalue Detection (MED), Maximum Minimum Eigenvalue

(MME), Maximum Eigenvalue to Trace (MET) etc. Unfortunately, most of the existing methods

are either problem specific or computationally complex due to exhaustive comparisons of test

hypothesis involved to achieve high accuracy. In [145], a technique is proposed to detect number

of signals in order to solve problem of DOA.

In this dissertation, a generalized spectrum sensing method is proposed to first detect the number

of possibly active PUs located in the far field region of the array and then estimate their

amplitudes, DOAs and frequencies. The proposed spectrum sensing scheme is not application

specific. It can be used for cooperative as well as non-cooperative spectrum sensing. Mean

Square Error (MSE) is used as a fitness function which defines an error between actual and

115

estimated signals at different sensors of the ULA and is derived from ML Principle. MSE is one

of the easy and optimum fitness functions to be minimized using array of sensors and fairly good

results are obtained even in the scenario of low SNR. We employed heuristic optimization

techniques to minimize the error in which GA being one of the most popular evolutionary

algorithm because of its reliability, efficiency and robustness is used as global optimizer

hybridized with PS as local optimizer. This simple and elegant technique simply demands a

passive sensor array whose snapshots should be readily available to us for calculation after every

10-20 seconds.

8.5 SYSTEM MODEL AND PROBLEM FORMULATION

We have an array of sensors that is sensing the signals from different base stations of primary

users. If the array is almost at the same height as that of the base station transmitters, we do not

have to detect the elevation angle . So, consider a uniform linear array as shown in Fig. 8.1

consisting of L omnidirectional sensors observing M far-field primary signals radiating with

different unknown carrier frequencies. The distance d between two consecutive elements is kept

one-quarter of the minimum wavelength of received signals i.e. 4

min.

Fig. 8.1: The System Model

116

The composite signal ix received by the thi sensor is expressed as,

i

M

m

idjk

mi zeax mm

1

sin)1( Li 1 )1.8(

where ma and m represent the amplitude and DOA of the thm source impinging on the array, mk

is the propagation constant and cfk mmm /2/2 with mf representing the frequency of

the thm signal incident on the array and iz is the AWGN added to the output of thi sensor. Thus

the parameters to be estimated for M incident sources are expresses in a vector as,

where M is the number of active PUs and is also unknown and has to be detected first.

The received signal vector X at the L-element ULA is expressed as,

T

LLi xxxxxX ],.,,..................,,[ 121

where superscript T denotes the transpose.

Thus the problem in hand is to develop a novel technique for two purposes, first detecting the

number of active PU signals impinging on ULA and second, performing joint estimation of

amplitude, DOA and frequency of the detected sources considering the sensor array as reference.

We also consider the effect of variation in SNR on the detection and estimation results. The

fitness function can be expressed mathematically as,

2

,,,

ˆmin XXfaM g

)2.8(

where X defines the estimated signal vector at the sensor array and is given as,

T

LLi xxxxxX ]ˆ,ˆ.,,.........ˆ.........,ˆ,ˆ[ˆ121

ix is the estimated output at the ith

sensor and is expressed as,

g

mm

M

m

idkj

mi eax1

ˆsin)1(ˆ

'

''

'ˆˆ

Li 1 )3.8(

],......,,,......,,,......,[ 111 MMM ffaa

117

where gM is the number of sources randomly selected to detect the possibly active PUs.

Thus the elements of the estimated vector obtained through optimization algorithm are given

by,

8.6 PROPOSED ALGORITHM FOR DETECTION OF PUS

In this section, we give an overview of the procedural steps carried out in GA optimization,

parameter settings for GA and hybrid scheme PS, and pseudo code for the proposed algorithm.

We solve the problem of detection first. To achieve this purpose, we randomly select gM

number of sources in the estimated signal vector and calculate mean square error MSE given

in 8.2. The value of gM is then increased or decreased aiming to decrease the MSE in each

selection. This process is repeated until minimum mean square error MMSE is obtained with

gM corresponding to MMSE indicating the number of active PUs. After detection of the number

of sources is done, we perform joint estimation of amplitude, DOA and frequency of the detected

signals by further refining the MMSE. The optimization problem given in 8.2 is solved through

GA hybridized with PS.

GA has been widely used to solve optimization problems in communication and array signal

processing because of being simple in concept, reliable, ease in implementation and with less

probability of getting stuck in local minima [146]. Efficiency, accuracy and reliability of GA can

be considerably improved by hybridization with any other competent computational technique

such as Interior Point Algorithm (IPA), PS etc. In [147], performance of GA, PS and Simulated

Annealing (SA) is compared with GA-PS and SA-PS in the joint estimation of amplitude and

DOA of multiple far-field sources incident on L-type array considering MSE as fitness function.

The flowchart for GA optimization has been provided in chapter 3. The steps followed in GA-PS

optimization are summarized below.

]ˆ,......,ˆ,ˆ,......,ˆ,ˆ,......,ˆ[ˆ 111 ggg MMM ffaa

118

_________________________________________________________________________________________

Algorithm: GA hybridized with PS _______________________________________________________________________________________________

Step (i): Initialization

Randomly generate P number of chromosomes (potential solutions). Lower and upper

bounds are specified for the genes.

Step (ii): Fitness Function Evaluation

Fitness of each chromosome in the population is computed using mean square error MSE

derived from ML Principle as fitness function and is given in 8.2. The chromosomes are

sorted on the basis of their fitness values.

Step (iii): Termination Criteria

The algorithm terminates if any of the following two criteria are met, i.e. reaching the

maximum number of cycles or achieving the predefined fitness value.

Step (iv): Create New Generations

Select the best chromosome depending on the value of its fitness and create next

generation by employing mutation and crossover.

Step (v): Fine-Tuning via Local Search

The PS algorithm takes the best chromosome obtained from GA as a starting point for

further improvement and refinement of results.

Step (vi): Storage:

Store the global best and to achieve better results repeat the steps 2 to 5 for sufficient

numbers of iterations for better statistical analysis.

_____________________________________________________________________________

MATLAB optimization toolbox is used for this purpose and parameter settings for GA and PS

are shown in table 8.1. Pseudo code of the proposed algorithm to solve the detection and

estimation problem is provided in table 8.2.

119

Table 8.1: Parameter Settings For GA-PS

8.7 SIMULATION RESULTS AND DISCUSSIONS

In this section, the performance of the proposed technique is evaluated in terms of accuracy for

two purposes, first, to detect of number of far-field sources impinging on ULA, and second, for

joint estimation of amplitude, DOA and frequency of the detected signals. Inter-element spacing

in the array is kept4

min. We perform spectrum sensing in the frequency band of 81MHz –

108MHz. The signals received at the array were polluted by AWGN. Different cases are

discussed on the basis of different number of sources M impinging on ULA, different number of

sensors L, and for different SNR levels, with SNR to be as high as 35dB and as low as 15dB. All

the values of DOA and SNR are taken in degrees and dB respectively.

GA PS

Parameters Settings Parameters Settings

Population size 300 Start point Optimal values

from GA

No. of generations 2000 Poll method GPS positive

basis 2N

Selection Stochastic uniform Polling order Consecutive

Mutation function Adaptive feasible Maximum

iterations

1000

Crossover function Heuristic Maximum

function

evaluation

10000

Crossover Fraction 0.2 Function

Tolerance

1e-18

Hybridization PS

No. of generations 3000 Expansion

Factor

2.0

Function Tolerance 1e-15 Contraction

Factor

2.5

Migration Direction Both Way

Penalty Factor

100 Scaling Function Rank

Elite Count 8

120

Table 8.2: Pseudocode Of The Proposed Algorithm For Detection Of Number Of Sources

.1,,,: MmwherefadInputs mmm

MforguessaasMChoose g

.1

M

m

idjk

mimmeax

1

sin)1(

Li 1

.2

g

mm

M

m

idkj

mi eax1

ˆsin)1(ˆ

'

''

'ˆˆ

Li 1

.3 compute2

0ˆ

gMM XXE

ErrorSquareMean//

.4 let 1 g

new

g MM

// sourceoneadd

.5 compute

newg

mm

M

m

idkj

mi eax1

ˆsin)1(ˆ

'

''

'ˆˆ

Li 1

.6 compute

2'0

ˆnewgMM XXE

.7 if )( 0'0 EE

.i MofvaluepossibleaasMsaveandEE new

g

'

00

.ii newgMupdate updatelastportingMM new

gnewg sup1//

.iii newgMlastofrecordkeepingwhiletostepsrepeat 75

gincreastartsMSEuntilacquired sin

else

.i 1 g

new

g MM

.ii 75 tostepsrepeat

ifend

directionsbothinEaroundEofvaluesthreeatleastObservemin00.8

new

gg MasEinincreaseensuretoMthatgconsiderin 00

min0EtocorrespondwhichMarounddecreasesorincreases new

g

min0..Re: EeiMMSEtocorrespondthatMturnOutput new

g

Fig. 8.2 illustrates the performance of GA for two incoming sources i.e. M = 2 under different

SNR conditions. A ULA with L = 20 sensors is employed for this purpose. The amplitude A ,

DOA and frequency f of the incoming signals are taken as ,5.4,0.3 21 aa

121

,145,75 21

oo MHzfMHzf 100,85 21 where 111 ,, fa correspond to the first PU and

222 ,, fa correspond to the second PU. The obtained results are averaged over 20 snapshots.

Fig. 8.2(a) illustrates the detection of two sources with gM ranging from 1 to 6. Minimum

Mean square error (MMSE) is plotted against the number of sources gM in the estimated signal

vector which clearly gives the minimum value when gM coincides with M. The figure also

indicates that increase in error is less significant in the case when gM > M as compared to the

case when gM < M which represents an under-determined system i.e. number of solutions are

less than the number of unknowns. After the detection of active sources, table 8.3 provides the

estimates of amplitudes, DOAs and frequencies of both PUs for different values of SNR. Fig.

8.2(b) and Fig. 8.2(c) plot error in DOA and frequency of the incident signals versus SNR

respectively and it is obvious from the figures that estimation accuracy increases to 99.87% in

DOA and 99.77% in frequency as the SNR increases from 15dB to 35dB.

Fig. 8.2(a): Detection of M = 2 PUs

1 2 3 4 5 610

-4

10-3

10-2

10-1

100

101

102

Mg

Me

an

Sq

ua

re E

rro

r

SNR = 30dB

SNR = 25dB

SNR = 20dB

122

Fig. 8.2(b): Error in DOA vs SNR for M = 2, L = 20

Fig. 8.2(c): Error in frequency vs SNR for M = 2, L = 20

15 20 25 30 350.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1

0.11

0.12

0.13

SNR (dB)

Err

or

in

(D

eg

ree

s)

delta

1

delta 2

15 20 25 30 350.1

0.12

0.14

0.16

0.18

0.2

0.22

SNR (dB)

Err

or

in f(

MH

z)

delta f

1

delta f2

123

Table 8.3: Amplitude, DOA And Frequency Estimation For Different SNR Levels With M = 2,

L = 20

SNR

35dB 3.00 4.50 75.04 144.97 84.89 100.10

30dB 3.00 4.50 74.95 144.96 85.13 100.12

25dB 2.99 4.51 75.07 145.06 84.84 99.85

20dB 2.98 4.52 74.91 145.08 84.82 100.19

15dB 3.02 4.48 74.90 144.89 85.19 100.21

In Fig. 8.3 illustrates the performance of GA-PS with M = 4 primary users. ULA with L = 25

sensors is used for this purpose. The values of amplitude, DOA and frequency of the sources are

taken as },81,60,2{ MHzo },88,90,5.2{ MHzo },95,135,3{ MHzo

and }.105,160,5.3{ MHzo

Figure 8.3(a)

plots MMSE versus gM to detect the number of active sources by setting gM in the range of 1

to 7 and it is obvious from the figure that error is minimum when MM g giving a clear

indication of 4 active PUs. Fig. 8.3(b) and 8.3(c) plot error in DOA and frequency estimates of

the detected users versus different SNR levels with SNR raised from 15dB to 35dB. The values

estimated by GA are tabulated in table 8.4. The results are averaged over 20 snapshots. Table 8.4

provides the amplitude, DOA and frequency estimates obtained. Fig. 8.3 proves the validity of

the proposed technique when the number of signals incident on the array increases and it can still

simultaneously estimate amplitudes, DOAs and frequencies with high estimation accuracy.

Figure 8.3(a): Detection of M = 4 PUs

1 2 3 4 5 6 710

-4

10-3

10-2

10-1

100

101

102

Mg

Me

an

Sq

ua

re E

rro

r

SNR = 30dB

SNR = 25dB

SNR = 20dB

1a 2a1 2 )(ˆ

1 MHzf )(ˆ2 MHzf

124

Fig. 8.3(b): Error in DOA vs SNR for M = 4, L = 25

Fig. 8.3(c): Error in frequency vs SNR for M = 4, L = 25

15 20 25 30 35

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

SNR (dB)

Err

or

in

(Degre

es)

delta fi

1

delta fi2

delta fi3

delta fi4

15 20 25 30 350.2

0.25

0.3

0.35

0.4

0.45

SNR (dB)

Err

or

in f (

MH

z)

delta f

1

delta f2

delta f3

delta f4

125

Table 8.4: Amplitude, DOA and frequency estimation for different SNR levels with M = 4, L =

25

In Fig. 8.4 we evaluate the performance of our proposed scheme with different number of

sensors L in the array as the SNR is raised from 15dB to 35dB. Number of PUs and the values

of amplitudes, DOAs and frequencies of the PUs are kept the same as in the case of figure 8.2.

The values estimated by GA-PS are tabulated in table 8.5. The results are averaged over 20

snapshots. It is obvious from Fig. 8.4(a) and Fig. 8.4(b) that the greater the number of sensors in

the array, the higher is the accuracy in the estimated values with further improvement achieved at

high SNR levels.

Fig. 8.4(a): Error in DOA estimation for different SNR levels and different number of sensors in

the array considering M = 2

SNR

35dB 2.00 2.50 3.00 3.50 60.20 90.18 135.19 159.84 81.23 88.24 94.79 104.77

30dB 1.99 2.51 3.00 3.48 59.78 90.24 134.75 159.79 80.71 88.28 94.74 104.72

25dB 1.98 2.41 3.01 3.47 59.72 90.31 134.73 160.28 80.68 87.67 95.33 104.69

20dB 2.02 2.52 2.99 3.53 59.62 89.66 135.37 160.34 80.63 88.35 95.39 105.38

15dB 2.03 2.47 2.98 3.54 60.40 90.34 134.59 159.63 81.45 87.39 95.43 105.42

6 8 10 12 14 16 18 20 220.06

0.08

0.1

0.12

0.14

0.16

0.18

L

Err

or

in

(De

gre

es)

SNR = 30dB

SNR = 25dB

SNR = 20dB

SNR = 15dB

1a 2a 3a14a 2 3 4 )(ˆ

1 MHzf )(ˆ2 MHzf )(ˆ

3 MHzf )(ˆ4 MHzf

126

Fig. 8.4(b): Error in frequency estimation for different SNR levels and different number of

sensors in the array considering M = 2

Table 8.5: Amplitude, DOA and frequency estimation for different SNR levels and different

number of sensors in the array with M = 2

SNR

L

30dB

6 3.04 4.54 75.12 144.88 85.22 100.23

10 3.03 4.53 75.11 144.90 85.21 100.22

14 3.02 4.53 74.91 144.93 84.83 100.19

18 2.98 4.49 75.07 145.05 85.16 99.82

22 3.00 4.50 74.94 144.96 85.15 100.16

25dB

6 2.96 4.55 75.13 145.12 85.23 99.77

10 2.95 4.46 75.12 145.11 85.22 100.22

14 2.95 4.48 74.91 145.09 84.81 100.18

18 3.02 4.48 74.91 145.07 84.82 100.17

22 2.99 4.51 74.92 144.93 85.17 99.83

20dB

6 3.06 4.57 75.15 144.84 85.25 99.76

10 3.05 4.56 75.13 145.15 84.77 99.75

14 3.03 4.53 75.12 145.12 84.80 100.23

18 2.98 4.51 74.90 145.09 84.82 99.80

22 2.99 4.49 74.91 144.92 85.17 100.18

15dB

6 3.07 4.42 75.17 144.82 85.27 100.26

10 2.96 4.57 74.83 144.84 84.78 99.75

14 3.03 4.45 74.86 145.12 85.79 100.23

18 2.98 4.46 74.88 145.08 85.81 99.79

22 2.98 4.47 75.10 145.09 84.82 99.80

6 8 10 12 14 16 18 20 22

0.16

0.18

0.2

0.22

0.24

0.26

0.28

L

Err

or

in f

1(M

Hz)

SNR = 30dB

SNR = 25dB

SNR = 20dB

SNR = 15dB

1a 2a 1 2 )(ˆ1 MHzf )(ˆ

2 MHzf

127


In this chapter, a novel idea based on GA hybridized with PS is presented for detecting the

number of active PUs and estimation of joint amplitudes, DOAs and frequencies of the detected

users for CRNs. The proposed method is not application specific, the signal parameters are

paired automatically and estimated with high accuracy. Moreover, the proposed algorithm has

less computation burden and offers satisfactory results even when number of users increases. The

simulation results verify the effectiveness of the proposed algorithm in AWGN environment.

128

Chapter 9

CONCLUSIONS AND FUTURE WORK

9.1 CONCLUSIONS

In this dissertation, we studied AF based relay-assisted cognitive radio networks in an underlay

spectrum sharing environment. The underlay networks impose strict interference constraints

towards the secondary users which limit their transmit power and allows only short-range

communication. Thus, performance enhancement of secondary communication in the frequency

bands allocated to the PUs is a major design challenge faced by the underlay RCRNs. It requires

relay selection along with the fine tuning and adjustment of the transmit power of the secondary

relays. The main contributions of the dissertation in this area are summarized as follows.

Various “Multiple Relay Selection” schemes are proposed to enable secondary communication

in an underlay spectrum sharing scenario. Rayleigh flat-fading is considered for this purpose,

assuming the availability of perfect instantaneous channel state information (CSI). Relay

selection aims to select the best combination of relays aiming to maximize the signal-to-noise

ratio (SNR) achieved at the destination while adhering to the interference constraint of the

primary network.

Another critical issue in the underlay networks is to optimize the transmit power consumption at

the cognitive relay network while ensuring the minimum quality-of-service requirements of the

primary and secondary networks. This dissertation formulates a transmit power minimization

problem and proposes a relay subset selection algorithm aiming to select the optimal

combination of relays that minimizes the total transmit power utilized at the relay network while

satisfying the minimum interference threshold of the primary network and the SNR threshold of

the secondary network simultaneously. The problem of transmit power minimization of the

129

cognitive relay network operating in an underlay scenario has not been solved by relay selection

so far to the best of our knowledge.

This dissertation also highlights the deficiency in the performance analysis of multiple relay

selection in AF-based underlay RCRNs. For this purpose, a relay subset selection algorithm is

proposed aiming to maximize the SNR received at the destination, keeping in view the

interference threshold of the primary network. The closed form expressions for the outage

probability and average probability of error have been derived through the CDF of the received

SNR at secondary destination, which has not been done in the literature so far, for multiple relay

selection in AF-based underlay RCRNs.

Moving one step further and acknowledging the increasing popularity of the famous Artificial

Intelligence tool, the fuzzy logic, FRBS assisted relay selection and transmit power allocation

(RSTPA) schemes are proposed for intelligent relay selection to solve the highlighted problems

of SNR maximization and transmit power minimization in CRNs.

The effects of variations in the instantaneous CSI, transmit power of source and relays,

interference threshold of the primary network, SNR threshold of the secondary network and size

of the potential relay network on multiple relay selection in underlay RCRNs are the main issues

that are analyzed in depth for all the schemes proposed in this research.

The last topic that is studied in this dissertation is spectrum sensing in Cognitive Radios. In order

to preserve the PUs’ rights of interference-free operation, the SUs are required to sense the

licensed bands at regular intervals, and reliably detect the primary signals. For this purpose, a

novel idea of uniform linear array (ULA) based spectrum sensing is proposed alongwith a hybrid

GA based algorithm. It not only detects the number of active PUs, but also provides the estimates

of amplitude, frequency and DOA of the active users upto high accuracy. The effectiveness and

reliability of the proposed scheme is proved under low SNR conditions.

9.2 FUTURE WORK

The work presented in this dissertation can be further extended in many directions.

130

It is assumed in all schemes that the perfect knowledge of CSI is available to perform

relay selection and power allocation to the selected relays. However, the accurate

information about the incoming and outgoing channels might not always be available,

which may distract the relay selection procedure. In future, the relay selection procedures

in the scenarios of imperfect CSI will be studied.

The line-of-sight path between source-destination pair and the phenomena of frequency

selectivity are neglected throughout the study, which significantly affect relay selection.

In future, these dominant factors will be incorporated in formulating relay selection

problem and performing relay selection and power allocation.

The spectrum sensing scheme proposed in this dissertation is based on the novel idea of

employing uniform linear array of sensors. Since, the relay network behaves like a virtual

array of distributed antennas in space, the proposed spectrum sensing method will be

extended to the relay network. The idea is to perform cooperative spectrum sensing

through potential relay network to detect the number of primary signals and the signal

parameters. Furthermore, relay selection and/or beamforming techniques will be applied

to enhance secondary performance, while guaranteeing to satisfy the interference

threshold of primary network.

The focus of the whole research is to enable secondary communication under interference

constraints, i.e. underlay mode of spectrum sharing. The problem of performance

enhancement of secondary communication in overlay and underlay mode when primary

signal is present and the interweave mode when primary signal is absent can be jointly

formulated to make secondary users smart enough to respond to the changing network

conditions and adapt themselves according to the operating environment in which they

operate.

131

REFERENCES

[1]. P. Lin, J. Jia, Q. Zhang, M. Hamdi, “Dynamic Spectrum Sharing With Multiple Primary

and Secondary Users”, IEEE Transactions on Vehicular Technology, vol. 60, no. 4, pp. 1756–

1765, 2011.

[2]. L. E. Doyle, “Essentials of Cognitive Radios”, Cambridge Wireless Essentials Series, Cambridge

University, Press, 2009.

[3]. Cognitive Wireless Communication Networks, Springer Science + Business Media, LLC, 2007.

[4]. J. Mitola, G. Q. Maguire, “Cognitive Radio: Making Software Radios ore Personal”, IEEE

Personal Communications, vol. 6, no. 4, pp. 13-18, 1999.

[5]. J.Lee,H. Wang, J.G.Andrews, and D. Hong, “Outage Probability of Cognitive Relay Networks with

Interference Constraints”, IEEE Trans. Wireless Commun., vol. 10, no. 2, pp. 390-395, 2011.

[6]. P. Steenkiste, D. Sicker, G. Minden, and D. Raychaudhuri, “Future Directions in Cognitive Radio

Network Research”, NSF Workshop Report, 2009.

[7]. J. Mietzner, L. Lampe, and R. Schober, “Distributed Transmit Power Allocation for Multihop

Cognitive-Radio Systems," IEEE Trans. Wireless Commun., vol. 8, no. 10, pp. 5187-5201, 2009.

[8]. O. Simeone, U. Spagnolini, and Y. Bar-Ness, “Stable Throughput of Cognitive Radios with and

without Relaying Capability," IEEE Trans. Commun., vol. 55, no. 12, pp. 2351-2360, 2007.

[9]. C. Stevenson, G. Chouinard, Z. Lei, W. Hu, S. Shellhammer, and W. Caldwell, “IEEE 802.22: The

First Cognitive Radio Wireless Regional Area Networks (WRANs) Standards”, IEEE

Communications Magazine, vol. 47, no. 1, pp. 130–138, 2009.

[10]. A.H. Bastami, A. Olfat, “Optimal SNR-based Selection Relaying Scheme with M

Ary modulations in Cooperative Wireless Networks”, IET Commun., vol. 5, no. 6, pp. 797-810,

2011.

[11]. D. Chen, H. Ji, X. Li, “Optimal Distributed Relay Selection in Underlay Cognitive Radio

Networks: An Energy-Efficient Design Approach”, IEEE WCNC, 2011, pp. 1203-1207.

[12]. F. GABRY, N. LI, N. Schrammar, M. Girnyk, E. Karipidis, R. Thobaben, L. K. Rasmussen, E.

G. Larsson and M. Skoglundi, “Secure Broadcasting in Cooperative Cognitive Radio Networks”,

Future Network & Mobile Summit (FutureNetw), 2012, pp. 1-8.

[13]. M. Chen, Ted C.-K. Liu, and X. Dong, “Opportunistic Multiple Relay Selection with Outdated

Channel State Information”, IEEE Trans. Vehicular Technology, vol. 61, no. 3, pp. 1333-1345,

2012.

[14]. T. Nechiporenko, P. Kalansuriya, and C. Tellambura, “Performance of Optimum Switching

Adaptive M-QAM for Amplify-and-Forward Relays”, IEEE Trans. Vehicular Technology, vol. 58,

no. 5, pp. 2258-2268, 2009.

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Peng%20Lin.QT.&searchWithin=p_Author_Ids:37536676300&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Juncheng%20Jia.QT.&searchWithin=p_Author_Ids:37289805100&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Qian%20Zhang.QT.&searchWithin=p_Author_Ids:37289940400&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Hamdi,%20M..QT.&searchWithin=p_Author_Ids:37272053300&newsearch=true

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=5738695&queryText%3DDynamic+Spectrum+Sharing+With+Multiple+Primary+and+Secondary+Users

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=5738695&queryText%3DDynamic+Spectrum+Sharing+With+Multiple+Primary+and+Secondary+Users

http://ieeexplore.ieee.org/xpl/RecentIssue.jsp?punumber=25

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Maguire,%20G.Q.,%20Jr..QT.&searchWithin=p_Author_Ids:37331827700&newsearch=true

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=788210&queryText%3DCognitive+Radio%3A+Making+Software+Radios+More+Personal+JOSEPH+MITOLA+111+AND+GERALD+Q.+MAGUIRE%2C+JR.+ROYAL+INSTITUTE+OF+TECHNOLOGY



http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Jemin%20Lee.QT.&searchWithin=p_Author_Ids:37406236800&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Hano%20Wang.QT.&searchWithin=p_Author_Ids:37405989900&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Daesik%20Hong.QT.&searchWithin=p_Author_Ids:37274629700&newsearch=true

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=5659866&queryText%3DOutage+Probability+of+Cognitive+Relay+Networks+with+Interference+Constraints+Jemin+Lee%2C+Hano+Wang%2C+Jeffrey+G.+Andrews%2C+Senior+Member%2C+IEEE%2C+and+Daesik+Hong%2C+Senior+Member%2C+IEEE

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=5659866&queryText%3DOutage+Probability+of+Cognitive+Relay+Networks+with+Interference+Constraints+Jemin+Lee%2C+Hano+Wang%2C+Jeffrey+G.+Andrews%2C+Senior+Member%2C+IEEE%2C+and+Daesik+Hong%2C+Senior+Member%2C+IEEE


http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=5783363&queryText%3DOptimal+SNR-based+selection+relaying+scheme+with+M-ary+modulations+in+cooperative+wireless+networks+A.H.+Bastami+A.+Olfat

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=5783363&queryText%3DOptimal+SNR-based+selection+relaying+scheme+with+M-ary+modulations+in+cooperative+wireless+networks+A.H.+Bastami+A.+Olfat

http://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=6279685



132

[15]. M. Naeem, U. Pareek, and D. C. Lee, “Power Allocation for Non-Regenerative Relaying in

Cognitive Radio System”, WiMob IEEE, 2010, pp. 720-725.

[16]. K-C Chen, R. Prasad, “Cognitive Radio Networks”, Willey Press, 2009.

[17]. E. Hossain, D. Niyato, Z. Han, “Dynamic Spectrum Access and Management in Cognitive Radio

Networks”, Cambridge University Press, 2009.

[18]. C.Bazelon,“Licensed or Unlicensed: The Economic Considerations

in Incremental Spectrum locations”, IEEE Commun. Magazine, vol. 47, no. 3, pp. 110-116, 2009.

[19]. Y. Xiao, G. Bi, and D. Niyato, “Game Theoretic Analysis for Spectrum Sharing with Multi-Hop

Relaying” , IEEE Trans. Wireless Commun., vol. 10, no. 5, pp. 1527-1537, 2011.

[20]. M. Gandetto, C. Regazzoni, “Spectrum Sensing: A Distributed Approach for Cognitive

Terminals”, IEEE J. Selected Areas in Communications, vol. 25, no. 3, pp. 546-557, 2007.

[21]. A. G. Fragkiadakis, E. Z. Tragos, I. G. Askoxylakis, “A Survey on Security Threats and

Detection Techniques in Cognitive Radio Networks Communications”, IEEE Surveys & Tutorials,

vol. 15, no. 1, pp. 428-445, 2013.

[22]. C. Ma, G. Yu, J. Zhang, “A Node Grouping Algorithm for Joint Relay Selection and Resource

Allocation in Cooperative Cognitive Radio Networks” , Int. Conf. Wireless Commun. and Signal

Processing (WCSP), 2011, pp. 1-5.

[23]. A. M. Wyglinski, M. Nekovee, T. Hou, “Cognitive Radio Communications and Networks,

Principles and Practice”, Elsevier, 2010.

[24]. G. J. Minden, J. B. Evans, L. S. Searl, D. DePardo, R. Rajbanshi, J. Guffey, Q. Chen, T. R.

Newman, V. R. Petty, F. Weidling, M. Peck, B. Cordill, D. Datla, B. Barker, and A. Agah,

“Cognitive Radios for Dynamic Spectrum Access- An Agile Radio for Wireless Innovation”, IEEE

Commun. Magazine, vol. 45, no. 5, pp. 113-121, 2007.

[25]. B. Razavi, “Challenges in the Design of Cognitive Radios”, IEEE 1 Custom Integrated Circuits

Conference, 2009, pp. 391-398.

[26]. T. Yucek and H. Arslan, “A Survey of Spectrum Sensing Algorithms for Cognitive Radio

Applications”, IEEE Commun. Surveys & Tutorials, vol. 11, no. 1, pp. 116-130, 2009.

[27]. H. Sadeghi, P. Azmi and H. Arezumand, “Cyclostationarity-based Soft Cooperative Spectrum

Sensing for Cognitive Radio Networks”, IET Commun., vol. 6, no. 1, pp. 29-38., 2012.

[28]. A. A. Khan, S. M. I. Rahman, M. Ahmed, “Research Challenges of Cognitive Radios”, Int. J.

Engg. Research Tech, vol 1, no. 3, pp. 1-4., 2012.

[29]. J. Marinho, E. Monteiro, “Cognitive Radio: Technology Survey and Future Research Directions”,

ANACOM 2013, pp. 1-9.

[30]. K. C. Chen, Y. J. Peng, N. Prasad, Y. C. Liang, and S. Sun, “Cognitive Radio Network

Architecture: Part I - General Structure”, ACM Int. Conf. Ubiquitous Information Management

Commun. (ICUIMC), 2008, pp. 114-119.

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=4804395&queryText%3DLicensed+or+Unlicensed%3A+The+Economic+Considerations+in+Incremental+Spectrum+Allocations

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=4804395&queryText%3DLicensed+or+Unlicensed%3A+The+Economic+Considerations+in+Incremental+Spectrum+Allocations








http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=4197723&queryText%3DAn+Agile+Radio+for+Wireless+Innovation





133

[31]. A. G. Fragkiadakis, E. Z. Tragos, I. G. Askoxylakis, “A Survey on Security Threats and

Detection Techniques in Cognitive Radio Networks”, IEEE Coomun. Survey and Tutorials, vol. 15,

no. 1, pp. 428-445, 2013.

[32]. S. Srinivasa and S. A. Jafar, “The Throughput Potential of Cognitive Radio: A

Theoretical Perspective”, IEEE Commun. Magazine, vol. 45, no. 5, pp. 73-79., 2007.

[33]. M. Choi, J. Park, and S. Choi, “Simplified Power Allocation Scheme for Cognitive Multi-Node

Relay Networks”, IEEE Trans. Wireless Commun., vol. 11, no. 6, pp. 2008-2012, 2012.

[34]. D. W. Bliss, K.W. Forsythe, A. M. Chan, “MIMO Wireless Communication”, Lincoln Laboratory

Journal, vol. 15, no. 1, pp. 97-126, 2005.

[35]. T. Y. Wang, and J. Y. Wu, “Cooperative Communications Using Reliability-Forwarding

Relays”, IEEE Trans. Commun., vol. 61, no. 5, pp. 1776-1785, 2013.

[36]. D. Liu W. Wang, and W. Guo, “ ‘Green’ Cooperative Spectrum Sharing Communication”, IEEE

Commun. Letters, vol. 17, no. 3, pp. 459-462, 2013.

[37]. P. Anghel and M. Kaveh, “Exact Symbol Error Probability of a Cooperative Network in a

Rayleigh-Fading Environment,” IEEE Trans. Wireless Commun., vol. 3 (5), 2004, pp. 1416–1422.

[38]. S. Ikki and M. Ahmed, “Performance Analysis of Cooperative Diversity Wireless Networks over

Nakagami Fading Channel,” IEEE Commun. Letters, vol. 11, no. 4, pp. 334–336, 2007.

[39]. J. Mietzner, R. Schober, L. Lampe, W. H. Gerstacker, and P. A. Hoeher, “Multiple-Antenna

Techniques for Wireless Communications – A Comprehensive Literature Survey”, IEEE

Commun. Surveys & Tutorials, vol. 11, no. 2, pp. 87-105, 2009.

[40]. B. Nazari, A. Jamalipour, “Cooperative Communication with Asymmetric Channel State

Information: A Contract Theoretic Modeling Approach” IEEE China Commun., vol. 10, no. 1, pp.

31-43, 2013.

[41]. H. A. Suraweera, T. A. Tsiftsis, G. K. Karagiannidis, and A. Nallanathan, “Effect of Feedback

Delay on Amplify-and-Forward Relay Networks With Beamforming”, IEEE Trans. Vehicular

Tech., vol. 60, no. 3, pp. 1265-1271, 2011.

[42]. Y. Zou, J. Zhu, B. Zheng, and Y. D. Yao, “An Adaptive Cooperation Diversity Scheme With

Best-Relay Selection in Cognitive Radio Networks”, IEEE Trans. Signal Processing, vol. 58, no.

10, pp. 5438-5445, 2010.

[43]. D. Chen, H. Ji, and V.C.M. Leung, “Distributed Best-Relay Selection for Improving TCP

Performance Over Cognitive Radio Networks: A Cross-Layer Design Approach”, IEEE J. Selected

Areas Commun., vol. 30, no. 2, pp. 315-322, 2012.

[44]. D. Raychaudhuri, and N. B. Mandayam, “Frontiers of Wireless and Mobile Communications”,

Proceedings of the IEEE, vol. 100, no. 4, pp. 824-840, 2012.

[45]. Q. Li, R. Q. HU, Y. Qian, and G. Wu, “Cooperative Communications for Wireless Networks:

Techniques and Applications in LTE-Advanced Systems”, IEEE Wireless Commun., vol. 19, no. 2,

pp. 22-29, 2012.

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=4197718&queryText%3DThe+Throughput+Potential+of+Cognitive+Radio%3A+A+Theoretical+Perspective

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=4197718&queryText%3DThe+Throughput+Potential+of+Cognitive+Radio%3A+A+Theoretical+Perspective














134

[46]. R. Mo, Y. H. Chew, and C. Yuen, “Information Rate and Relay Precoder Design for Amplify-

and-Forward MIMO Relay Networks With Imperfect Channel State Information”, IEEE Trans.

Vehicular Technology, vol. 61, no. 9, pp. 3958-3968, 2012.

[47]. Q. Liu, W. Zhang, X. Ma, G. T. Zhou, “Designing Peak Power Constrained Amplify-and-

Forward Relay Networks with Cooperative Diversity” IEEE Trans. Wireless Communications, vol.

11,no. 5, pp. 1733-1743, 2012.

[48]. A. N. D’Andrea, V. Lottici, and R. Reggiannini, “RF Power Amplifier Linearization Through

Amplitude and Phase Predistortion,” IEEE Trans. Commun., vol. 44 (11), 1996, pp. 1477–1484.

[49]. A. Riberio, X. Cai, and G. Giannakis, “Symbol error probabilities for general cooperative links,”

IEEE Trans. Wireless Commun., vol. 4, no. 3, pp. 1264–1273, 2005.

[50]. P. Merkey and E. C. Posner, “Optimum Cyclic Redundancy Codes for Noise Channels,” IEEE

Trans. Inf. Theory, vol. 30, no. 6, pp. 865–867, 1984.

[51]. A. Chakrabarti, A. Sabharwal, B. Aazhang, “Cooperative Communications: Fundamental Limits

and Practical Implementation”, Rice University.

[52]. J. Si, Z. Li, X. Chen, B. Hao, and Z. Liu, “On the Performance of Cognitive Relay Networks

Under Primary User’s Outage Constraint”, IEEE Commun. Letters, vol. 15, no. 4, pp. 422-424,

2011.

[53]. I. Krikidis, J. N. Laneman, J. S. Thompson, and S. McLaughlin, “Protocol Design and

Throughput Analysis for Multi-User Cognitive Cooperative Systems”, IEEE Trans. Wireless

Commun., vol. 8, no. 9, pp. 4740-4751, 2009.

[54]. L. Li, X. Zhou, H. Xu, G. Y. Li, D. Wang, and A. Soong, “Simplified Relay Selection and Power

Allocation in Cooperative Cognitive Radio Systems”, IEEE Trans. Wireless Communications, vol.

10, no. 1, pp. 33-36, 2011.

[55]. S. Sengupta, K. P. Subbalakshmi, “Open Research Issues in Multi-Hop Cognitive Radio

Networks, IEEE Commun. Magazine, vol. 51, no. 4, pp. 168-176, 2013.

[56]. M. Xia and S. Aissa, “Cooperative AF Relaying in Spectrum-Sharing Systems: Performance

Analysis under Average Interference Power Constraints and Nakagami-m Fading”, IEEE Trans.

Commun., vol. 60, no. 6, pp. 1523-1533, 2012.

[57]. Y. Chen, G. Yu, Z. Zhang, H. Chen, and P. Qiu, “On Cognitive Radio Networks with

Opportunistic Power Control Strategies in Fading Channels”, IEEE Trans. Wireless Commun., vol.

7, no. 7, pp. 2752-2761, 2008.

[58]. M. Naeem, D.C. Lee, U. Pareek, “An Efficient Multiple Relay Selection Scheme for Cognitive

Radio Systems”, IEEE ICC 2010, pp. 1-5.

[59]. K. B. Fredj, and S. Aissa, “Performance of Amplify-and-Forward Systems with Partial Relay

Selection under Spectrum-Sharing Constraints”, IEEE Trans.

Wireless Communications, vol. 11, no. 2, pp. 500 –504, 2012.














135

[60]. D. Li, “Outage Probability of Cognitive Radio Networks with Relay Selection”, IET Commun.,

vol. 5, no. 18, pp. 2730-2735, 2011.

[61]. M. Seyfi, S. Muhaidat, J. Liang, “Relay Selection in Cognitive Radio Networks with Interference

Constraints”, IET Commun., vol. 7, no. 10, pp. 922-930, 2013.

[62]. S. I. Hussain, M. S. Alouini, K. Qaraqe and M. Hasna, “Reactive Relay Selection in Underlay

Cognitive Networks with Fixed Gain Relays”, IEEE Int. Conf. Commun., 2012, pp. 1784-1788.

[63]. V. N. Q. Bao, T. Q. Duong, D. B. Costa, G. C. Alexandropoulos, and A. Nallanathan, “Cognitive

Amplify-and-Forward Relaying with Best Relay Selection in Non-Identical Rayleigh Fading, IEEE

Commun. Letters, vol. 17, no. 3, pp. 475-478, 2013.

[64]. H. Yu, W. Tang, and S. Li, “Outage Probability and SER of Amplify-and-Forward Cognitive

Relay Networks, IEEE Wireless Commun. Letters, vol. 2, no.2, pp. 219-222, 2013.

[65]. J. Xu, H. Zhang, D. Yuan, Q. Jin, and C. X. Wang, “Novel Multiple Relay Selection Schemes in

Two-Hop Cognitive Relay Networks”, IEEE INt. Conf. Commun. and Mobile Computing, 2011,

pp. 307-310.

[66]. M. Naeem, D. C. Lee, and U. Pareek, “An Efficient Multiple Relay Selection Scheme For

Cognitive Radio Systems”, IEEE Int. Conf. Commun., 2010, pp.1-5.

[67]. A. Ponsich, A. L. Jaimes, and C. A. Coello, “A Survey on Multiobjective Evolutionary

Algorithms for the Solution of the Portfolio Optimization Problem and Other Finance and

Economics Applications”, IEEE Trans Evolutionary Computation, vol. 17, no. 3, pp. 321-344,

2013.

[68]. T. Blickle, “Theory of Evolutionary Algorithms and Application to System Synthesis”, Ph.D.

Dissertation, 1967.

[69]. W. F. Gao, S. Y. Liu, and L. L. Huang, “A Novel Artificial Bee Colony Algorithm Based on

Modified Search Equation and Orthogonal Learning”, IEEE Trans. Cybernetics, vol. 43, no. 3, pp.

1011-1024, 2013.

[70]. W. Peng, Y. Zeng, H. Gong, Y. Q. Leng, Y. H. Yan, and W. Hu, “Evolutionary Algorithm and

Parameters Extraction for Dye-Sensitised Solar Cells One-Diode Equivalent Circuit Model”, Micro

and Nano Letters, vol 8, no. 2, pp. 86-89, 2013.

[71]. A. Sharaqa, N. Dib, “Position-Only Side Lobe Reduction of a Uniformly Excited Elliptical

Antenna Array using Evolutionary Algorithms”, IET Microw. Antennas Propag., vol. 7, no. 6, pp.

452-457, 2013.

[72]. W. R. A. Ibrahim and M. M. Morcos, “Artificial Intelligence and Advanced Mathematical Tools

for Power Quality Applications: A Survey”, IEEE Trans. Power Delivery, vol. 17, no. 2, pp. 668 –

673, 2002.

[73]. S. K. Pal, V. Talwar, and P. Mitra, “Web Mining in Soft Computing Framework: Relevance,

State of the Art and Future Directions”, IEEE Trams. Neural Networks, vol. 13, no. 5, pp. 1163-

1177, 2002.









136

[74]. R. A. Vural, T. Yildirim, T. Kadioglu, and A. Basargan, “Performance Evaluation of

Evolutionary Algorithms for Optimal Filter Design”, IEEE Trans. Evolutionary Computation”, vol.

16, no.1, pp. 135-147, 2012.

[75]. D. Karaboga and B. Basturk, “On the performance of Artificial Bee Colony (ABC) Algorithm,”

Appl. Soft Comput., vol. 8, no.1, pp. 687–697, 2008.

[76]. R. Wang, R. C. Purshouse, and P. J. Fleming, “Preference-Inspired Coevolutionary Algorithms

for Many-Objective Optimization”, IEEE Trans Evolutionary Computation, vol. 17, no. 4, pp. 474-

494, 2013.

[77]. L. D. S. Coelho and P. Alotto, “Gaussian Artificial Bee Colony Algorithm Approach Applied to

Loney’s Solenoid Benchmark Problem”, IEEE Conf. Electromagnetic Field Computation, 2010,

pp.1.

[78]. L. Costa, I.A.C.P.E. Santo, R. Denysiuk and E.M.G.P. Fernandes, “Hybridization of a Genetic

Algorithm with a Pattern Search Augmented Lagrangian Method”, Int. Conf. Engineering

Optimization, 2010, pp.1-10.

[79]. Z. He, P.F. Zhu, “A Hybrid Genetic Algorithm for Multiresponse Parameter Optimization within

Desirability Function Framework”, Int. Conf Industrial Engineering and Engineering Management,

2009, pp. 612-617.

[80]. Y.G. Petalas, K.E. Parsopoulos, M.N. Vrahatis, “Memetic Particle Swarm Optimization”, Ann

Oper Res, 156, 2007, pp. 99-127.

[81]. L. D. S. Coelho and V. C. Mariani, "Combining of Chaotic Differential Evolution and Quadratic

Programming for Economic Dispatch Optimization with Valve-Point Effect," IEEE Trans. Power

Systems, vol. 21, pp. 989-996, 2006.

[82]. F. S. Abu-Mouti, and M. E. El-Hawary, “Optimal Distributed Generation Allocation and Sizing

in Distribution Systems via Artificial Bee Colony Algorithm”, IEEE Trans. Power Delivery, vol.

26, no. 4, pp. 2090-2101, 2011.

[83]. D. Karaboga and B. Basturk, “A Powerful and Efficient Algorithm for Numerical Function

Optimization: Artificial Bee colony (ABC) Algorithm”, J. Global Optimiz., vol. 39, pp. 459–471,

2007.

[84]. D. Karaboga, B. Akay, “A Comparative Study of Artificial Bee Colony Algorithm”, Appl. Math.

Comput., vol. 214, no. 1, pp. 108–132, 2009.

[85]. D. Jeya Mala, V. Mohan, M. Kamalapriya, “Automated Software Test Optimization Framework –

an Artificial Bee Colony Optimization-Based Approach”, IET Software, 2010.

[86]. X. Li, M. Yin, “Hybrid Differential Evolution with Artificial Bee Colony and its Application for

Design of a Reconfigurable Antenna Array with Discrete Phase Shifters”, IET Microwaves,

Antenna and Propagation, vol. 6, no. 14, pp. 1573-1582, 2012.






137

[87]. N. Taspinar, D. Karaboga, M. Yildirim, and B. Akay, “Partial Transmit Sequences based on

Artificial Bee Colony Algorithm for Peak-to-Average Power Ratio Reduction in Multicarrier Code

Division Multiple Access Systems”, IET Commun., vol. 5, no. 8, pp. 1155-1162, 2011.

[88]. Y. Wang, W. Chen, and C. Tellambura, “A PAPR Reduction Method Based on Artificial Bee

Colony Algorithm for OFDM Signals”, IEEE Trans. Wireless Commun., vol. 9, no. 10, pp. 2994-

2999, 2010.

[89]. N. Todorovic and S. Petrovic, “Bee Colony Optimization Algorithm for Nurse Rostering”, IEEE

Trans. Systems, Man, and Cybernetics, vol. 43, no. 2, pp. 467-473, 2013.

[90]. X. Zhang , X. Zhang , S. Yin Yuen , S. L. Ho , and W. N. Fu, “An Improved Artificial Bee

Colony Algorithm for Optimal Design of Electromagnetic Devices”, IEEE Trans. On Magnetics,

vol. 49, no. 8, pp. 4811-4816, 2013.

[91]. D. S. Weile and E. Michielssen, “Genetic Algorithm Optimization Applied to Electromagnetics:

A Review”, IEEE Trans. Antenna and Propagation, vol. 45, no. 3, pp. 343-353, 1997.

[92]. M. Sanna and M. Murroni, “Optimization of Non-Convex Multiband Cooperative Sensing With

Genetic Algorithms”, IEEE J. Selected Topics Signal Process., vol. 5, no. 1, pp. 87-96, 2011.

[93]. L. Khalid and A. Anpalagan, “Cooperative Sensing With Correlated Local Decisions in Cognitive

Radio Networks”, IEEE Trans. Vehicular Tech., vol. 61, no. 2, pp. 843-849, 2012.

[94]. A. Balieiro, P. Yoshioka, K. Dias, C. Cordeiro and D. Cavalcanti, “Adaptive Spectrum Sensing

for Cognitive Radio based on Multi-Objective Genetic Optimization”, IEEE Electronics Letters,

vol. 49, no. 17, pp. 1099-1101, 2013.

[95]. M. Canales and J.R. Gallego, “Potential Game for Joint Channel and Power Allocation in

Cognitive Radio Networks”, IEEE Electronics Letters, vol. 46, no. 24, pp. 1632-1634, 2010.

[96]. A. Alsharoa, F. Bader, and M.S. Alouini, “Relay Selection and Resource Allocation for Two-

Way DF-AF Cognitive Radio Networks”, IEEE Wireless Commun. Letters, vol. 2, no. 4, pp. 427-

430, 2013.

[97]. S. L. Blond, R. Aggarwal, “A Review of Artificial Intelligence Techniques as Applied to

Adaptive Autoreclosure, with Particular Reference to Deployment with Wind generation”,

Universities Power Engg. Conference, 2009, pp. 1-5.

[98]. D. G. Schwartz, G. J. Klir, H. W. Lewis, and Y. Ezawa, “Applications of Fuzzy Sets and

Approximate Reasoning”, Proceedings of IEEE, vol. 82, no. 4, pp. 482-498, 1994.

[99]. L.A. Zadeh. “A rationale for fuzzy control,” J. Dynamic Syst., Meas., Contr., vol. 94, no. 1, pp.

3-4, 1972.

[100]. -, “Outline of a new approach to the analysis of complex systems and decision processes,” IEEE

Trans. Syst., Man, Cyhern., vol. SMC-3, no. 1, pp. 28-44, 1973.

[101]. K. P. Philip, E.L. Dove, D.D. McPherson, N.L. Gotteiner,W. Stanford, K.B. Chandran,

“The Fuzzy Hough Transform-Feature Extraction in Medical Images”, IEEE Trans. Medical

Imaging, vol. 13, no. 2, pp. 235-240, 1994.

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Philip,%20K.P..QT.&searchWithin=p_Author_Ids:37387981500&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Dove,%20E.L..QT.&searchWithin=p_Author_Ids:37367981700&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.McPherson,%20D.D..QT.&searchWithin=p_Author_Ids:37358299000&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Gotteiner,%20N.L..QT.&searchWithin=p_Author_Ids:37376731500&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Stanford,%20W..QT.&searchWithin=p_Author_Ids:37378286100&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Chandran,%20K.B..QT.&searchWithin=p_Author_Ids:38221346400&newsearch=true

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=293916&refinements%3D4291944246%26queryText%3Dfuzzy+set+theory+in+pattern+recognition



138

[102]. J. M. W. Kickert, H. Koppelaar, “Application of Fuzzy Set Theory to Syntactic Pattern

Recognition of Handwritten Capitals”, IEEE Trans. Systems, Man and Cybernetics, vol. SMC-6,

no. 2, pp. 148-151, 1976.

[103]. G. Fang, “A Survey on Analysis and Design of Model-Based Fuzzy Control Systems”, IEEE

Trans. Fuzzy Sys., vol. 14, no. 5, pp. 676-697, 2006.

[104]. A.B. Ajiboye, R.F. Weir, “A Heuristic Fuzzy Logic Approach to EMG Pattern Recognition for

Multifunctional Prosthesis Control”, IEEE Trans. Neural Systems Rehabilitation Engg., vol. 13, no.

3, pp. 280-291, 2005.

[105]. Cui-Yan, J. Y. Tao, W.J. Tao, “Pattern Recognition Based on Fuzzy Three-Valued Logic

Reasoning”, Int. Conf. Information Engg. Computer Science, 2009, pp. 1-4.

[106]. K. L. Wood, “Computations with Imprecise Parameters in Engineering Design: Background and

Theory,” ASME J. Mechanisms, Transmissions, and Automation in Design, vol. 111, no. 4, pp.

616-625, 1989.

[107]. M. Smithson, “Fuzzy Sets Analysis for Behavioral and Social Sciences”, New York Springer-

Verlag, 1987.

[108]. A. Billot, “Economic Theory of Fuzzy Equilibria: An Axiomatic Analysis”, Berlin-New York:

Springer-Verlag, 1992.

[109]. Elkan, C. ; Berenji, H.R. ; Chandrasekaran, B. ;de Silva, C.J.S. ; Attikiouzel, Y. ; Dubois,

D. ; Prade, H. ;Smets, P. ; Freksa, C. ; Garcia, O.N. ; Klir, G.J. ; Bo Yuan ;Mamdani,

E.H. ; Pelletier, F.J. ; Ruspini, E.H. ; Turksen, B. ;Vadiee, N. ; Jamshidi, M. ; Pei-Zhuang

Wang ; Sie-Keng Tan ;Shaohua Tan ; Yager, R.R. ; Zadeh, L.A. , “The Paradoxical Success of

Fuzzy Logic”, IEEE Expert, vol. 9, no. 4, pp. 3-49, 1994.

[110]. J. Kacprzyk and M. Fedrizzy, “Fuzzy Regression Analysis”, Warsaw, Poland: Omnitech Press,

1992.

[111]. N. Baldo, and M. Zorzi, “Cognitive Network Access using Fuzzy Decision Making”, IEEE Trans.

Wireless Commun, vol. 8, no. 7, pp. 3523-3535, 2009.

[112]. K. Ahmed, F. Bashir, Najum-ul-Hassan, M. Ehsan ul Haq, “Comparative study of Centralized

Cooperative Spectrum Sensing in Cognitive Radio Networks”, IEEE ICSPS, 2010, pp. 246-249.

[113]. N. Baldo, M. Zorzi, “Fuzzy Logic for Cross-layer Optimization in Cognitive Radio Networks”,

IEEE Commun. Magazine, vol. 46, no. 4, pp. 64-71, 2008.

[114]. W. Liu, T. Lv, L. Gao, W. Wang, B. Liu, “A Novel Cooperative Spectrum Sensing Scheme

Based on Fuzzy Integral Theory in Cognitive Radio Networks”, IEEE WiCom, 2009, pp. 1-4.

[115]. W. Liu, T. Lv, L. Gao, W. Wang, B. Liu, “A Novel Cooperative Spectrum Sensing Scheme

Based on Fuzzy Integral Theory in Cognitive Radio Networks”, WiCom IEEE, 2009, pp. 1-4.

[116]. H. Zhang, X. Wang, “A Fuzzy Decision Scheme for Cooperative Spectrum Sensing in Cognitive

Radio”, VTC IEEE, 2011, pp. 1-4.

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Kickert,%20Walter%20J.M..QT.&searchWithin=p_Author_Ids:37402142500&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Koppelaar,%20Henk.QT.&searchWithin=p_Author_Ids:37402139900&newsearch=true




http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Elkan,%20C..QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Berenji,%20H.R..QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Chandrasekaran,%20B..QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.de%20Silva,%20C.J.S..QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Attikiouzel,%20Y..QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Dubois,%20D..QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Dubois,%20D..QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Prade,%20H..QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Smets,%20P..QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Freksa,%20C..QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Garcia,%20O.N..QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Klir,%20G.J..QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Bo%20Yuan.QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Mamdani,%20E.H..QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Mamdani,%20E.H..QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Pelletier,%20F.J..QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Ruspini,%20E.H..QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Turksen,%20B..QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Vadiee,%20N..QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Jamshidi,%20M..QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Pei-Zhuang%20Wang.QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Pei-Zhuang%20Wang.QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Sie-Keng%20Tan.QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Shaohua%20Tan.QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Yager,%20R.R..QT.&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Zadeh,%20L.A..QT.&newsearch=true

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=5555652&queryText%3DComparative+Study+of+Centralized+Cooperative+Spectrum+Sensing+in+Cognitive+Radio+Networks

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=5555652&queryText%3DComparative+Study+of+Centralized+Cooperative+Spectrum+Sensing+in+Cognitive+Radio+Networks

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=5304215&queryText%3Da+vovel+cooperative+spectrum+sensing+scheme+based+don+fuzzy+integral

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=5304215&queryText%3Da+vovel+cooperative+spectrum+sensing+scheme+based+don+fuzzy+integral

139

[117]. W. Ejaz, N. ul Hasan, S. Aslam, H. S. Kim, “Fuzzy Logic Based Spectrum Sensing for Cognitive

Radio Networks”, 2011, NGMAST IEEE, pp. 185-189.

[118]. S. R. Aryal, H. Dhungana, K. Paudyal, “Novel Approach for Interference Management in

Cognitive Radio”, 2012, AH-ICI IEEE. pp. 1-5.

[119]. W. Mustafa, J.S. Yu, E. Rakus-Andersson, A. Mohammed, W.J. Kulesza, “Fuzzy-based

Opportunistic Power Control Strategy in Cognitive Radio Networks”, 2010, ISABEL IEEE. pp. 1-

9.

[120]. H.S.T. Le and Q. Liang, “An Efficient Power Control Scheme for Cognitive Radios”, WCNC

IEEE, 2007, pp. 2559-2563.

[121]. Z. Tabakovic, S. Grgic, M. Grgic, “Fuzzy Logic Power Control in Cognitive Radio”, IWSSIP

IEEE, 2009, pp. 1-5.

[122]. A. Dey, S. Biswas, S. Panda, “A New Fuzzy Rule Based Power Management Scheme for

Spectrum Sharing in Cognitive Radio”, ICCIA IEEE, 2011, pp. 1-4.

[123]. G. Amarasuriya, M. Ardakani, and C. Tellambura, “Output-Threshold Multiple-Relay-Selection

Scheme for Cooperative Wireless Networks”, IEEE Trans. Vehicular Tech., vol. 59, no. 6, pp.

3091-3097, 2010.

[124]. I. S. Gradshteyn, I. M. Ryzhik. “Table of Integrals, Series and Products”, 5th

Ed. New

York, USA: Academic Press, 1994.

[125]. Y Zhao, R Adve T. J., Lim, “Symbol Error Rate of Selection Amplify-and-Forward Relay

Systems”, IEEE Commun. Letters, vol. 10, no. 11, pp.757-759, 2006.

[126]. S. I. Hussain, M. M. Abdallah, M. S. Alouini, M. Hasna, K. Qaraqe, “Best Relay Selection using

SNR and Interference Quotient for Underlay Cognitive Networks”, IEEE ICC, 2012, pp.4176-4180.

[127]. S. Haykin, "Cognitive Radio: Brain-Empowered Wireless communications", IEEE J. Sel. Areas

Comm., vol. 23, no. 2, pp. 201-220, 2005.

[128]. J. Mitola, “Cognitive Radio Architecture Evolution,” 2009, Proceedings of the IEEE, pp. 626–

641.

[129]. A. Ghasemi, E. S. Sousa,” Spectrum Sensing in Cognitive Radio Networks: Requirements,

Challenges and Design Trade-offs”, IEEE Comm. Magazine, vol. 46, pp. 32-39, 2008.

[130]. S. Kapoor, G. Singh, “Cooperative Spectrum Sensing: A Hybrid Model Approach”, ICDECom

IEEE, 2011, pp. 1-5.

[131]. S. Hussain, X. Fernando, “Spectrum Sensing in Cognitive Radio Networks: Up-to-date

Techniques and Future Challenges”, TIC-STH IEEE, 2009, pp. 736-741.

[132]. N. Pratas, N. R. Prasad, A. Rodrigues and R. Prasad, “Cooperative Spectrum Sensing: State of the

Art Review”, Wireless VITAE IEEE, 2011, pp.1-6.

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Ejaz,%20W..QT.&searchWithin=p_Author_Ids:37671665700&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.ul%20Hasan,%20N..QT.&searchWithin=p_Author_Ids:38231522900&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Aslam,%20S..QT.&searchWithin=p_Author_Ids:37545282800&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Hyung%20Seok%20Kim.QT.&searchWithin=p_Author_Ids:37696369900&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Mustafa,%20W..QT.&searchWithin=p_Author_Ids:37844714700&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Jhang%20Shih%20Yu.QT.&searchWithin=p_Author_Ids:37856373000&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Rakus-Andersson,%20E..QT.&searchWithin=p_Author_Ids:37297697500&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Mohammed,%20A..QT.&searchWithin=p_Author_Ids:37302062800&newsearch=true

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=p_Authors:.QT.Kulesza,%20W.J..QT.&searchWithin=p_Author_Ids:37267653600&newsearch=true

140

[133]. M. Subhedar and G. Birajdar, “Spectrum Sensing Techniques in Cognitive Radio Networks: A

Survey”, vol. 3, IJNGN.

[134]. A. Chandran, A. Karthik, A. Kumar, R. Naidu, S. Siva, U. S. Iyer, R. Ramanathan, “Evaluation of

Energy Detector based Spectrum Sensing for OFDM Based Cognitive Radio”, INCOCCI, IEEE,

2010, pp. 163-167.

[135]. S. M. Kay, “Fundamentals of Statistical Signal Processing: Detection Theory”, vol. 2, Prentice

Hall, Upper Saddle River, NJ, USA, 1998.

[136]. W. A. Gardner, “Exploitation of Spectral Redundancy in Cyclostationary Signals,” IEEE Signal

Processing Magazine, vol. 8, pp. 14–36,1991.

[137]. L. Gupta, R. P. Singh, “Array Signal Processing: DOA Estimation for Missing Sensors”,

ICPCES IEEE, 2010, pp. 1-4.

[138]. R. Roy, T. Kailath, “ESPRIT-Estimation of Signal Parameters via Rotational Invariance

Techniques,” IEEE Trans. Acoust., Speech, Signal Process., vol. 37, no. 7, pp. 984–995, 1989.

[139]. R. Schmidt, “Multiple Emitter Location and Signal Parameter Estimation,” IEEE Trans. Antennas

Propagation, vol. 34, no. 3, pp. 276–280, 1986.

[140]. Y. D. Huang and M. Barkat, “Near-Field Multiple Source Localization by Passive Sensor Array”.

IEEE Trans. Antennas Propag., vol 39, no. 7, pp. 968–975, 1991.

[141]. K. A. Meraim, Y. Hua and A. Belouchrani, “Second-Order Near-Field Source Localization:

Method and Performance Analysis”, Conference Record of the Thirtieth Asilomar Conference on

Signals, Systems and Computers, vol. 1, pp. 723–727, 1996.

[142]. N. Yuen and B. Friedlander, “Performance Analysis of Higher Order ESPRIT for Localization of

Near-Field Sources,” IEEE Trans. Signal Process., vol. 46, no. 3, pp. 709–719, 1998.

[143]. Q.T. Zhang and K.M. Wong, “A New Information-Theoretic Criterion for Detection of the

Number of Signals in Spatially Correlated Noise with Covariance Matrix”, ICASSP IEEE, 1992,

vol. 5, pp. 381-384.

[144]. U. Konda, Y. Cheng, T. Singh, P. D. Scott, “Source Identification of Puff-Based Dispersion

Models Using Convex Optimization”, IEEE Conf. Computing & Processing

(Hardware/Software) ;Engineered Materials, Dielectrics & Plasmas ; Nuclear Engineering, 2010,

pp. 1-6.

[145]. C. Candan and S. Koc, “Beamspace Approach for Detection of the Number of Coherent

Sources”, Radar Conference IEEE, 2012, pp. 913-918.

[146]. J. Zhang, H. S. Chung, and W. L. Lo, “Clustering-based Adaptive Crossover and Mutation

Probabilities for Genetic Algorithms," IEEE Trans. on Evol. Comput., vol. 11, no. 3, pp. 326-335,

2007.

[147]. F. Zaman, I. M. Qureshi, A. Naveed, J. A. Khan, and M. A. Z. Raja, “Amplitude and Direction

of Arrival Estimation: Comparison between Different Techniques”, Progress In Electromagnetics

Research B, vol. 39, pp. 319-335, 2012.


http://ieeexplore.ieee.org/search/searchresult.jsp?refinements=4294967045




Documents

SENSOR ARRAY ASSISTED SPECTRUM SENSING AND …