Research and development of adaptive beamformers for interference suppression in smart antennas

Functional block diagram of a smart antenna using DOA-based adaptive beamforming algorithms.. Functional block diagram of a smart antenna using training-based adaptive beamforming algori

VIETNAM NATIONAL UNIVESITY, HANOI

UNIVERSITY OF ENGINEERING AND TECHNOLOGY

TONG VAN LUYEN

RESEARCH AND DEVELOPMENT OF ADAPTIVE BEAMFORMERS FOR

INTERFERENCE SUPPRESSION

IN SMART ANTENNAS

Dissertation for the Degree of Doctor of Philosophy

in Communication Engineering

VIETNAM NATIONAL UNIVESITY, HANOI

UNIVERSITY OF ENGINEERING AND TECHNOLOGY

TONG VAN LUYEN

RESEARCH AND DEVELOPMENT OF ADAPTIVE BEAMFORMERS FOR INTERFERENCE SUPPRESSION IN

I confirm that:

- This dissertation represents my own work;

- The contribution of my supervisor and others to the research and to the dissertation was consistent with normal supervisory practice;

- External contributions to the research are acknowledged

Date: September 26th, 2018

Tong Van Luyen

First of all, I would like to express my sincere thanks to my supervisor, Assoc.Prof Dr.-Ing Truong Vu Bang Giang, for his supervision, his support andassessment comments in the work, and what he has done for me at VNU University

of Engineering and Technology He believed me in my scientific ability, challenged

my work, and encouraged me to pursue my ideas during the time we workedtogether

I would like to thank Faculty of Electronic Engineering, Hanoi University ofIndustry, and Faculty of Electronics and Telecommunications, VNU University ofEngineering and Technology for their support for me to do PhD course

My special thanks to M.S Nguyen Minh Tran for his discussions andcomments, and his technical support in our lab to my dissertation

I highly appreciate the help from Dr Hoang Manh Kha, Dr Dao Thanh Hai,and thank them for their helpful discussions in nature-inspired optimization, andtheir kind encourages to the success of this work

I would like to thank M.S Pham Thi Quynh Trang for her kind support at boththe simulation technique in my dissertation and the work in my office

I am grateful to my dear colleagues, Nguyen Viet Tuyen, Duong Thi Hang, BoQuoc Bao, Vu Thi Phuong Quynh, and the other colleagues of HaUI Faculty ofElectronic Engineering, for their practical support during my work

Finally, my beloved thanks and my deepest gratitude to my parents of bothsides, my wife Duyen, my daughter My Quyen, and my son Minh Duc for their loveand encouragement Thanks to your sharing and sacrifice and to you I dedicate thisdissertation

Declaration i

Acknowledgement ii

Contents iii

List of Abbreviations 1

List of the Symbols and Notations 2

List of Figures 3

List of Tables 6

Introduction 7

I Rationale for the Study 7

II Objectives, Subjects, Scope, and Methodology of the Study 10

II.1 Objectives 10

II.2 Subjects, Scope, and Methodology 11

III Significance of the Study 11

IV Dissertation Outline 13

Chapter 1: Overview of Beamforming 14

1.1 Beamforming for Smart Antennas 14

1.2 Mathematic Basis of Smart Antennas 18

1.2.1 Geometric Relations 18

1.2.2 The Model of Smart Antennas with Linear Arrays 20

1.3 Optimal Beamforming Techniques 23

1.3.1 Classical Optimization Techniques 24

1.3.2 Nature-inspired Optimization Techniques 25

1.4 Chapter Conclusions 30

Chapter 2: General Process to Develop BA-based Adaptive Beamformers for Interference Suppression 31

2.1 Problem Determination 31

2.2 Array Factor Building 32

2.3 Pattern Nulling Techniques 33

2.3.1 Amplitude-only Control 33

2.3.2 Phase-only Control 34

2.3.3 Complex-weight Control 34

2.4 Formation of Objective Function 35

2.5 Building of BA-based Adaptive Beamforming Algorithms 37

2.6 Development of Adaptive Beamformers 38

2.7 Proposals of General Process to Build Adaptive Beamformers 40

2.8 Chapter Conclusions 41

Chapter 3: Developments of BA-based Adaptive Beamformers for Interference

Suppression 42

3.1 Common Items of BA-based Adaptive Beamformers 42

3.2 The Beamformer Based on Phase-only Control 45

3.2.1 Diagram of the Beamformer 45

3.2.2 Penalty Parameter in the Objective Function 46

3.2.3 Numerical Results and Discussions 46

3.2.4 Summary 50

3.3 The Beamformer Based on Amplitude-only Control 51

3.3.1 Diagram of the Beamformer 51

3.3.2 Numerical Results and Discussions 51

3.3.3 Summary 56

3.4 The Beamformer Based on Complex-weight Control 57

3.4.1 Diagram of the Beamformer 57

3.4.2 Numerical Results and Discussions 58

3.4.3 Summary 64

3.5 Effect of Mutual Coupling 64

3.6 Summary 67

3.7 Chapter Conclusions 72

Conclusions and Future Works 73

List of Publications 76

Bibliography 77

Appendix 81

A Smart Antennas 81

A.1 Antenna Arrays 81

A.2 Classification of Beamforming 86

A.3 Application Model of Smart Antennas 89

B Classical Optimization Techniques 91

B.1 Optimal Criteria 91

B.2 Adaptive Beamforming Algorithms 92

B.3 Dolph-Chebyshev Weighting Method 95

C Software for Modeling Adaptive Beamforming in Smart Antennas 99

C.1 Application Model 100

C.2 Simulation Results 100

D Supported Simulation Results 105

D.1 Additional Results for Patterns with Single and Multiple Nulls 105

D.2 Some Sets of Weights for the Investigated Scenarios 110

Amplitude-only Control and Bat Algorithm-based AdaptiveBeamformer

Accelerated Particle Swarm OptimizationBat Algorithm

Complex-weight Control and Bat Algorithm-based AdaptiveBeamformer

Digital BeamformingDirection-Of-ArrivalDigital Signal ProcessorFirst-Null BeamwidthGenetic AlgorithmHalf-Power BeamwidthLeast Mean SquareMutual CouplingMinimum Mean Square ErrorMean Square Error

Null Depth LevelPhase-only Control and Bat Algorithm-based Adaptive

BeamformerParticle Swarm OptimizationRadio Frequency

Recursive Least SquareSpace Division Multiple AccessSidelobe Level

Sample Matrix InversionSignal-Not-Of-InterestSignal-Of-InterestUniform Linear Array

List of the Symbols and Notations

I In-phase channel in of binary baseband signals

Q Quadrature-phase channel in of binary baseband signals

Sum

The real vector space (n-dimensional space of the variables)

Subset of or equal to

An element ofElevation angle in the coordinate system for antenna analysisAzimuth angle in the coordinate system for antenna analysisWavelength

Differential value ofWavenumber

Vector and its components

Z, Zij Maxtrix and its components

x* Complex conjugate of x

Transposition of a matrixHermitian transpose of a matixCross correlation of andCovariance of

Real part ofImaginary part ofCosine integralSine integralInfinity

List of Figures

Figure 1.1 Beamforming for smart antnenas 15

Figure 1.2 Applications of beamforming 15

Figure 1.3 Block diagram of analog beamforming in smart antennas 16

Figure 1.4 Block diagram of DBF in smart antennas 16

Figure 1.5 Simple block diagram of adaptive beamformer at the receiving end 18

Figure 1.6 The analyzed linear array 19

Figure 1.7 Linear-array smart antennas at the receiving end 20

Figure 1.8 Radiation pattern of 20-element ULA 22

Figure 1.9 Flowchart of Bat algorithm 29

Figure 2.1 Geometry of ULAs of 2N elements 32

Figure 2.2 Block diagram of adaptive beamformers for interference suppression 38

Figure 2.3 Flowchart of the proposed beamformers 39

Figure 2.4 General process to build adaptive beamformers 41

Figure 3.1 Diagram of PHA_BA_ABF 45

Figure 3.2 NDL and maximum SLL with different in the case of pattern with single null 46

Figure 3.3 Objective function comparisons of BA, PSO, and GA 47

Figure 3.4 Optimized pattern with a single null at 14° 48

Figure 3.5 Optimized pattern with three nulls at -48°, 20°, and 40° 49

Figure 3.6 Optimized pattern with a broad null from 30° to 40° 49

Figure 3.7 Diagram of AMP_BA_ABF 51

Figure 3.8 Objective function comparisons of BA, PSO, and GA 52

Figure 3.9 Optimized pattern with single symmetric null at 14° 53

Figure 3.10 Optimized patterns with three symmetric multiple nulls at 14°, 26°, and 33° 54

Figure 3.11 Optimized patterns with a symmetric broad null from 20° to 50°, unchanged main lobe beamwidth and peak SLL = -18.3 dB 55

Figure 3.12 Optimized pattern with a symmetric broad null from 20° to 50°, broaden main lobe beamwidth and SLL ≤ -30 dB 56

Figure 3.13 Diagram of CW_BA_ABF 57

Figure 3.14 Objective function of BA with different population sizes 59

Figure 3.15 Objective function between BA and APSO 59

Figure 3.16 Optimized patterns with single null at 14° 60

Figure 3.17 Optimized pattern with three nulls at -33°, -26°, and -14° 61

Figure 3.18 Optimized pattern with three nulls at -40°, 20°, and 40° 62

Figure 3.19 Optimized pattern with a broad null from -50° to -20° 62

Figure 3.20 Optimized pattern with a broad null ([-30°, -20°] and [45°, 60°]) 63

Figure 3.21 Optimized pattern with a broad null ([-30°, -20°] and [45°, 60°]) and SLL of -30 dB 64

Figure 3.22 Optimized pattern (nulls: -48°, 20°, 40°) with mutual coupling 65

Figure A.1 Radiation pattern of a twenty-element ULA 82

Figure A.2 Coordinate system for antenna analysis 82

Figure A.3 Different array geometries for smart antennas: (a) uniform linear array, (b) circular array, (c) two-dimensional grid array and (d) three-dimensional grid array 86

Figure A.4 Switched-beam system 87

Figure A.5 Comparison of (a) switched-beam system, and (b) adaptive array system 88

Figure A.6 Relative coverage area comparison among sectorized systems, switched-beam systems, and adaptive array systems in (a) low interference environment, and (b) high interference environment 88

Figure A.7 Functional block diagram of a smart antenna using DOA-based adaptive beamforming algorithms 89

Figure A.8 Radiation pattern of a smart antenna 90

Figure A.9 Functional block diagram of a smart antenna using training-based adaptive beamforming algorithms 90

Figure B.1 Geometry of ULA antennas of 2N elements 99

Figure B.2 Normalized array factor for 20-element Chebyshev arrays with sidelobes at -30 dB 99

Figure C.1 The main lobes of the 8-element ULA have been steered to the desired directions as θ = 49°, -30°, 30°, 60° 101

Figure C.2 Five nulls have been set at elevation angles of -55°, -35°, -15°, 20°, and 45° 101

Figure C.3 The main beam is steered to θ = 30° and 5 nulls are set at θ = -55°, -35°, -15°, 0°,45° at the same time 102

Figure C.4 The optimized pattern with all side lobe levels are suppress to -30dB by Dolph-Chebyshev weighting method 103

Figure C.5 The optimized pattern by applying both LMS algorithm and Dolph-Chebyshev weighting method 103

Figure C.6 The optimized pattern of 1×8 ULA using LMS algorithm 104

Figure C.7 The optimized pattern of 1×8 ULA using both LMS algorithm and Dolph-Chebyshev weighting method 104

Figure D.1 Pattern with a single symmetric null in the range of θ: a) (-90°, 90°); b) (13°, 16°) 105

Figure D.2 Pattern with three symmetric nulls in the range of θ:

List of Tables

Table 3.1 Common parameters for all proposed beamformers 43

Table 3.2 NDL and maximum SLL of the patterns in all scenarios with or withoutmutual coupling 66

Table 3.3 Summary of the proposals 67

Table 3.4 Comparisons between the proposals in this dissertation and the proposalin 71Table B.1 Resulting weights computed by Dolph-Chebyshev weighting method 98

Table D.1 Some sets of weights consisting amplitudes (a n ) and phases (δ n) of thepatterns shown in Figures 3.4-3.6 110

Table D.2 Some sets of weights for the patterns shown in Figures 3.9-3.12 110

Table D.3 Some sets of weights for the patterns shown in Figures 3.16-3.21 111

I Rationale for the Study

Beamforming is a signal processing technique in sensor arrays to directionallytransmit or receive signals in space-time In order to do that, the signalscorresponding to array elements are combined in the interest of boosting the desiredsignals in particular directions and minimizing the undesired signals (interferences)

in the others Beamforming can be applied for both transmitting and receiving ends

in order to achieve spatial selectivity, thus, it is also called spatial filteringtechnique In fact, it can be used for radio or sound waves and has been widelyapplied for various applications such as Radar, Sonar, Wireless communications,Radio Astronomy, Seismology, and Topography [6, 18, 26, 56]

Over the last decades, wireless technology has been developed at a remarkablerate, which has brought new and high-quality services at lower costs This hasresulted in an increase in airtime usage, and in the number of subscribers As aresult, this leads to new challenges for next generations of wireless communicationsnetworks The most practical solution to this problem is to use spatial processing

remains as the most promising, if not the last frontier, in the evolution of multiple access systems” [42] Spatial processing lies at the heart of adaptive antennas or smart antenna

systems that employ beamforming As a result, space division multiple access (SDMA),one of the most complicated applications of smart antenna technology, is indispensable tothe development of cellular radio systems [11]

The advances of beamforming in cellular phone standards and other wirelesscommunication ones over the generations have resulted in the achievement of highdensity cells and higher throughput [1, 4, 14, 16, 23, 38, 45, 52, 63] In fact,

beamforming has been used in all the second, the third, and the fourth generationcellular standards Additionally, beamforming is being deployed in indoor networkssuch as Wi-Fi Even though it is still unsure which frequency band will be utilizedfor 5G technology, beamforming is going to play a major role in the future [16].

As mentioned above, beamforming for smart antennas plays a vital role inwireless communication systems, especially for new generation ones Actually,smart antennas exhibit various benefits in coverage, data rate, spectrum efficiency,interference suppression, which are all the vital factors of wireless communicationsystems [21, 48-50] Therefore, it has received enormous interest worldwide [11].Nowadays, the increasing number of wireless devices causes serious pollution

in the electromagnetic propagation environment In this context, smart antennaswith pattern nulling capabilities emerge as a promising solution for interferencesuppression applications Beamformers offer smart antennas the capability ofinterference suppression by: (i) steering the main lobe to the desired signal; (ii)suppressing sidelobes at directions of interferences; (iii) or placing nulls atdirections of interferences [10, 11, 20, 26, 43, 44] In the cases of (i) and (ii), whenthe desired signal boosted at the main lobe is still weaker than the interferencesreceived at sidelobes, the desired signal is overwhelmed by the interference Inorder to solve this problem, pattern nulling is regarded as one of the best solutionsfor interference suppression, because it allows smart antennas to adaptively placenulls at directions of interferences while maintaining the main lobe at the direction

of desired signal and suppressing sidelobes However, this has resulted in anincrease in the complexity of computation and the requirement of the effectiveoptimization tools [11, 19, 20, 52]

In order to implement the pattern nulling by using adaptive beamformers, twomain aspects including pattern nulling control and optimization techniques havebeen addressed

Firstly, several pattern nulling control techniques such as controlling theamplitude-only, the phase-only, position-only, and the complex-weight (both the

amplitude and the phase) have been widely studied and implemented All thesetechniques have their own advantages and limitations [18, 20, 47] Among those,the complex-weight control has been considered as the most flexible and efficienttechnique because it allows adjusting amplitude and phase simultaneously [13, 20,

27, 64] Nonetheless, it is the most complicated and expensive technique due to thefact that each array element must have a controller, a phase shifter and anattenuator More critically, the computational time will be a considerable issue inlarge array antennas Indeed, the problem for the phase-only and position-onlycontrols is inherently nonlinear [30] The position-only control [3, 12, 29] requires amechanical driving system such as servomotors for adjusting the array elementposition This makes the system more complicated, and causes difficulty in accuracycontrol Phase-only control is less complex and more attractive for the phased arrayssince the required control is available at no extra cost [2, 33, 34, 46] The amplitude-only control is simple compared to the others as it only changes the amplitudeexcited at each array element [5, 30, 37, 54]

Secondly, in recent years, optimization techniques have been widely applied inbeamforming for antenna array pattern synthesis including pattern nulling Theclassical optimization techniques used for the array pattern synthesis are likely to bestuck in local minima if the initial guesses are not reasonably close to the finalsolution Most of the classical optimization techniques and analytical approachesalso suffer from the lack of flexible solutions for a given antenna pattern synthesisproblem To overcome these issues, various nature-inspired optimization algorithmsbased on computational intelligence approaches have been developed Thesealgorithms such as ant colony optimization [13], bacterial foraging algorithm [30],differential evolution [54], clonal selection [5], bees algorithm [28], especially thegenetic algorithm (GA) [15, 25, 35, 47, 64] and particle swarm optimization (PSO)[15, 31, 39] have been proved to be better and more flexible than the classical ones.These nature-inspired optimization algorithms have been proposed and imple-mented with their own benefits and limitations in pattern nulling In general, there

are still some challenges for the pattern nulling based on these nature-inspiredalgorithms as: (i) computation speed and performance; (ii) the lack of detailedanalysis about the general process to obtain pattern nulling, which leads to thedifficulty in understanding, applying and developing applications These issues arethe motivation for further research in this field.

Recently, Bat algorithm (BA) is a new nature-inspired computation techniquebased on the bat behavior of using echolocation to detect prey, avoid obstacles, andlocate their roosting crevices in the dark It has been successfully used to solvevarious kinds of engineering problems BA is better than PSO and GA optimization

in terms of convergence, robustness and precision [59, 61] This algorithm wasapplied for the first time for beamforming in 2016 [40] Authors of [40] showed thatthe BA is a promising optimization tool for adaptive beamforming in terms ofcomputation time Nevertheless, this work was still in preliminary phase and thus, itlacked adequate analysis on the application of BA in beamforming

Therefore, the development of adaptive beamformers for interferencesuppression is obviously still a challenge for researchers regarding the improvement

in computational speed and capability of pattern nulling To tackle these challenges,this dissertation will concentrate on proposing a general process to build BA-basedadaptive beamformers to suppress interference for ULAs in smart antennas Thisgeneral process is then implemented to develop three types of BA-based adaptivebeamformers to suppress interference for ULAs using: (i) amplitude-only, (ii)phase-only, and (iii) complex-weight control techniques

II Objectives, Subjects, Scope, and Methodology of the Study

II.1 Objectives

- To research and propose a general process to build BA-based adaptive beamformers to suppress interference for ULAs in smart antennas

- To implement the general process to develop three types of BA-basedadaptive beamformers to suppress interference for ULAs using: (i) amplitude-only, (ii)phase-only, and (iii) complex-weight control techniques.

II.2 Subjects, Scope, and Methodology

This study focuses on:

- Pattern analysis of antenna arrays;

- Adaptive beamforming techniques for antenna arrays;

- Global optimization algorithms (nature-inspired optimizationalgorithms such as genetic algorithm (GA), accelerated particle swarm optimization(APSO), and Bat algorithm (BA));

- Interference suppression using beamformers

- Synthesis and analysis of: antenna array pattern using adaptive

beamforming in smart antennas; and nature-inspired optimization;

- Modeling of proposed beamformers in terms of interference

suppression using smart antennas;

- Simulation and evaluation of the proposals in particular scenarios

The significance of the study in science and in practice is as follows:

Scientific significance:

- Proposal of a general process to build BA-based adaptive beamformers for interference suppression applications in smart antennas;

- Proposals of three high performance BA-based adaptive beamformersfor suppressing interference, which use amplitude-only, phase-only, and complex-weightcontrol techniques, respectively.

- The proposals have been implemented to develop three differentbeamformers for 20-element ULA with isotropic or dipole element Additionally, themutual coupling has also been investigated in the case of dipole element and phase-onlycontrol According to the numerical results, the proposed beamformers have shown theability to suppress sidelobes, to maintain predefined beamwidth, and to place preciselysingle, multiple, and broad nulls at an arbitrary direction of interferences Furthermore,those beamformers are much faster and more effective in terms of null steering and sidelobe suppression in pattern synthesis than GA and APSO-based ones

- These proposals can be applied to design and implement adaptivebeamformers for interference suppression applications in radar and wirelesscommunication networks

(1) Proposal of a general process to build BA-based adaptive beamformers

to suppress interferences for ULAs in smart antennas

(2) Successful implementation of the general process to develop three types

of BA-based adaptive beamformers to suppress interferences for ULAs using only, phase-only, and complex-weight control techniques, respectively

amplitude-12/112

IV Dissertation Outline

The dissertation consists of an introduction, three chapters, and a conclusion,

in which:

- Chapter 1 presents a general review on beamforming: an overview ofbeamforming; beamforming techniques including mathematical basis, optimizationtechniques These are related to the contents of this dissertation

- Chapter 2 presents the first proposal, a general process to build BA-basedadaptive beamformers for pattern nulling of ULAs This process includes six steps fromproblem determination to developments of adaptive beamformers

- Chapter 3 presents the second proposal by applying the process given inChapter 2 This proposal includes three different BA-based adaptive beamformers forpattern nulling of ULAs, of which pattern nulling controls are amplitude-only, phase-only, and complex-weight (both the amplitude and the phase), respectively Thesebeamformers have been successfully implemented and verified in terms of pattern nullingsynthesis

Chapter 1 Overview of Beamforming

This chapter presents an overview of beamforming, its applications for smartantennas in wireless communication systems, technical basis of beamformingincluding application models, mathematical basis, optimization techniques that arerelated to the contents of this dissertation

1.1 Beamforming for Smart Antennas

In smart antennas, beamforming is used along with antenna array to form anequivalent directional antenna system [6, 18, 26, 56] This directional antennasystem (smart antenna systems or shortly written, smart antennas) is able to focus

on the radiation power or spatially receive power in a particular direction in space.This spatial radiation or power reception of smart antennas, also called “beam”, isachieved by electrical control using beamforming, in which the desired signals inparticular directions are boosted and the interferences in the others are minimized.Therefore, beamforming has been widely used in many applications such as radar,sonar, and wireless communication systems In wireless communication system, it isdeployed to enhance the performance by increasing the efficiency of radio spectrumutilization, interference suppression, and power saving [11, 14, 16-18, 23, 24, 26]

In beamforming, the signal corresponding to each element has beencontrolled by a specific principle This control aims to form and steer the beam ofthe array in such a way as: (i) form and steer the main beam to a desired direction;

(ii) suppress the sidelobes; (iii) and set nulls at undesired directions The beam ofthe array has been formed and controlled according to the requirements of the specificapplications [11, 18, 26]

Figure 1.1 Beamforming for smart antnenas [8].

Figure 1.2 Applications of beamforming [18].

In general, common controlling parameters are the amplitude, the phase, orboth the amplitude and the phase of excitations corresponding to the elements.These controlled parameters are also called “weights” Beamformers at thereceiving end apply this set of weights for the signals from elements to gain thecontrolled signals, then, combine all these signals to a desired output

In analog beamforming, the weight ( ) of each array element is controlled inthe analog domain (Radio Frequency) Phase shifters and attenuators are used toadjust the phase ( ) and the amplitude ( ) of each antenna path, respectively Based

on specific rules, these controls, or beamforming techniques are applied to form andsteer the beam of the antenna arrays to meet particular requirements A

simple block diagram of analog beamforming in smart antennas is given in the Figure 1.3 [26, 58].

Figure 1.3 Block diagram of analog beamforming in smart antennas [58].

Digital beamforming (DBF) controls the weight of each array element in thedigital domain DBF is a marriage between antenna and digital technologies It hasbeen used to construct the smart antenna systems as presented in Figure 1.3including three major components: the antenna array, the digital transceivers, andthe digital signal processor (DSP) [11, 26]

As shown in Figure 1.4, the received signals (Radio frequency signals - RF signals) are detected and digitized at the element level Keeping RF information in

16/112

the form of a digital stream gives access to a large domain of signal processingtechniques, as well as algorithms that can be used to extract information from thespatial domain data Particularly, digital beamformers digitize and convert thereceiving signals into two streams of binary baseband signals (i.e., in-phase (I) andquadrature-phase (Q) channels) Included within these baseband signals are theamplitudes and phases of signals received at each elements of the array DBF iscarried out by weighting these digital signals, thereby adjusting their amplitudes andphases such a way that when added together they form the desired beam Thisprocess can be carried out using a special-purpose DSP [26].

Adaptive beamforming is capable of automatically adapting its response todifferent situations It has been applied for adaptive array systems to provide moredegrees of freedom since they have the ability to adapt in real time the radiationpattern to the RF signal environment In other words, they can direct the main beamtoward the pilot signal or Signal-Of-Interest (SOI) while suppressing the antennapattern in the direction of the interferers or Signals-Not-Of-Interest (SNOIs) To put

it simply, adaptive array systems can customize an appropriate radiation pattern foreach individual user This is far superior to the performance of a switched-beamsystem (see Appendix A for more details) [11]

A simple structure of adaptive beamformers (ABF) in the receiving end isdisplayed in Figure 1.5 ABF carries out weighting the receiving signals, therebyadjusting their amplitudes or phases in such a way that when added together theyform desired output They are able to adaptively adjust the value of weights ( ) topoint the beam in any wanted direction and to manipulate its shape to optimize thesystem performance Because of their flexibility, adaptive beamformers have beenutilized in various applications [26]

Additionally, some basic concepts and characteristics have been introduced inAppendix A for more information about smart antennas In order to support thestudy, mathematical basis and optimization technique of adaptive beamforming insmart antenna will be introduced in the next sections

Desired direction (signal)

Undesired directions (interferences)

Figure 1.5 Simple block diagram of adaptive beamformer at the receiving end [26].

1.2 Mathematic Basis of Smart Antennas

In smart antenna, although there are different array geometries, the principle ofsignal processing techniques shares some common points Therefore, for simplicity,only linear arrays will be analyzed in this section

1.2.1 Geometric Relations

Figure 1.6 shows a linear array where N elements are positioned along the α axis with uniform inter-element spacing, d, and the first element (element 0) is at

the origin of the coordinate system The direction of incoming waves has been

defined by elevation angel θ and azimuth angle φ in spherical coordinates [11, 17,

26] To make it simple, we assume that:

- The inter-element spacing is small enough to have no significant difference

of amplitude of incoming waves and therefore amplitudes of receiving signals at differentelements are considered as the same

18/112

- Incoming waves at each element with a particular plane wave is considered

as a radio signal As a result, there are a limited number of radio signals impinging thearray

The distance from element n (the coordinate is (x n, y n, z n)) to the origin of the coordinate system is defined as

where: ⃗ is the unit vector on direction of the incident waves at element 0 at the origin of the coordinate system and is represented by

Consequently, the wavefront arrives at element n sooner than at element 0 and

the differential distance is calculated as:

1.2.2 The Model of Smart Antennas with Linear Arrays

Figure 1.7 presents a basic model of a linear-array smart antenna [11, 17, 26]:

The receiving signal corresponding to each element is multiplied by an

appropriate weight, w n, which is able to be adjusted by both the amplitude and the

phase Then, all these products are summed to make an output signal, y.

For simplicity, the α axis in Figure 1.6 is chosen as the y axis In this case, the coordinate of element n becomes (0, y n , 0), y n = nd, and φ=90°.

Figure 1.7 Linear-array smart antennas at the receiving end [17].

From (1.4), the phase difference between the signal at element n and element

0

Then, the receiving signal at element n is represented as

The combination signal at the output of the smart antenna is

20/112

is the array factor.

If the complex weights are

If 0, a maximum response of AF will result at the angle θ0 That

is, the main beam of the array has been steered towards the wave source at elevation

angle θ0 An example of AF for twenty-element uniform linear array (ULA), in which all weights wn = 1, is shown in Figure 1.8, where the main beam is steered

towards the antenna boresight

Accordingly, AF, which is at a direction of the incident wave and with a specific weight vector, defines a ratio of the received signal at output of the smart

antenna to the received signal at a basic element By adjusting the weights, the

beam of the array will be controlled in space

21/112

Figure 1.8 Radiation pattern of 20-element ULA.

Additionally, if each element is directional and identical, f 0 (θ,φ), the radiation pattern of the array, f(θ,φ), has been calculated by the pattern multiplication principle, which states that the beam pattern of an arrays is the product of element

pattern and the array factor [17, 26]:

The AF can be expressed in terms of vector inner product

(1.12)where

determine the array factor of a complicated array that is composed of simplesubarrays, e.g the array factor of a planar array.

In addition to placing elements along a line to form a linear array, one canposition them on a plane to form a planar array Planar arrays provide additionalvariables which can be used to control and shape the pattern of the array The mainbeam of the array can be formed and steered towards any point in its half space

Additionally, as shown in Figure 1.5, the output at time t, y(t), is given by a linear combination of the data at N elements at time t [11, 26] as

H represents Hermitian transpose,

1.3 Optimal Beamforming Techniques

If the signal environment is stationary, weights are easily computed by solvingthe normal equations However, in practice, the signal environment is dynamic ortime varying; therefore, the weights need to be computed with adaptive methods As

a result, optimal beamforming techniques based on these adaptive methods play animportant role in adaptive beamforming [11]

In optimal beamforming techniques, a weight vector that minimizes a costfunction is determined Typically, this cost function, related with a performancemeasure, is inversely associated with the quality of the signal at the array output, so

when the cost function is minimized, the quality of the signal is maximized at thearray output [24] In terms of optimization methods, there are various types ofobjective functions to be made [11, 26] This leads to various types of adaptivebeamforming algorithm for adaptive beamforming Two types of the optimalbeamforming methods have been studied in this dissertation: classical optimizationtechniques and nature-inspired optimization techniques.

1.3.1 Classical Optimization Techniques

The most commonly used optimal criteria are the Minimum Mean SquareError (MMSE), Maximum Signal-to-Noise Ratio, and Minimum (noise) Variance[11, 17, 26] Among those, MMSE is a popular performance measures in computingthe optimum weights by minimizing the MSE objective function (cost or fitnessfunction) [11, 17, 26] The solution of this function leads to a special class of

optimum filters called Wiener filters [7, 51] (see details in Appendix B) whereby the

optimum weights are yielded:

−1

(1.17)

the covariance, respectively

Equation (1.17) is the so-called Wiener solution, which is the optimal antenna

array weight vector, , in the MMSE sense Based on this solution, there are someconventional adaptive beamforming algorithms such as Sample Matrix Inversion(SMI), Least Mean Square (LMS), and Recursive Least Square (RLS) [7, 11, 22,

measure to obtain weights for uniformly spaced linear arrays steered to broadside (θ

= 0°) This is a popular weighting method because the sidelobe level (SLL) can bespecified, and the minimum possible first-null beamwidth is obtained (see details inAppendix B)

Applying the fundamentals of adaptive beamforming mentioned above, a basicsoftware code has been built to model adaptive beamforming in smart antennas,which is oriented to investigate the basis of adaptive beamforming and to supportfurther studies (See Appendix C)

1.3.2 Nature-inspired Optimization Techniques

1.3.2.1 Nature-inspired Optimization Approach

The classical gradient-based optimization methods applied for adaptivebeamforming still have some limitations due to the following reasons: (i) highsensitivity for starting points when the number of solution variables and hence thesize of the solution space increase; (ii) frequent convergence to the local optimumsolution or divergence or revisiting the same suboptimal solution; (iii) requirement

of continuous and differentiable objective function (gradient search methods); (iv)requirement of the piecewise linear cost approximation (linear programming); and

(v) problem of convergence and algorithm complexity (nonlinear programming) Toovercome this, various nature-inspired optimization methods have been employed for theoptimal design of adaptive beamforming with better parameter performance [53, 61, 62]

Nature-inspired optimization provides promising and effective globaloptimization approaches for problem solving in machine intelligence, data miningand resource management Nature has evolved over millions of years under avariety of challenging environments and can thus provide a rich source ofinspiration for designing algorithms and approaches to tackle challenging problems

in real-world applications The success of these algorithms in applications hasincreased their popularity in recent years, and active research has also led to the

significant increase in the number of algorithms It is estimated that about 140different types of algorithms now exist in the literature, and this number is certainlygradually increasing Researchers have tried to find inspiration from various sources

in nature, such as ants, bees, bats, fish, birds, mammals, plants, physical andchemical systems such as gravity, river systems, waves and pheromone This leads

to a diverse of range of algorithms with different capabilities and different levels ofperformance [53, 61, 62]

A combination of nature-inspired optimization algorithms (global optimizationalgorithms), computational electromagnetics, and computer-processing is apromising tool for solving challenges of smart antennas in wireless communication[44, 53]

1.3.2.2 Bat Algorithm

Bat algorithm is a new nature-inspired optimization approach developed byXin-She Yang in 2010 [59], in which the fundamental principle is inspired by thesocial behavior of bats and the phenomenon of echolocation to sense distance It hasbeen successfully applied to solve various kinds of engineering problems BA isbetter than PSO and GA optimization in terms of convergence, robustness and

precision [59, 61]

In BA [59, 61], each bat (i) at time step t is defined by its position , velocity

, frequency , loudness, and the emission pulse rate in a d-dimensional

search space The new solutions 1 and velocities 1 are given by

is the current global best location (solution) which is located after comparing

all the solutions among all n bats Frequency range is defined by and ,

which are chosen depending on the domain size of the problem of interest Initially,each bat is randomly given a frequency which is drawn uniformly from [ , ] For the

local search part, once a solution is selected among the current best

solutions, a new solution for each bat is generated locally using random walk as

(1.21)where [ ] is a random number, while is the average loudness of all the

bats at time step t.

Furthermore, in consecutive iterations, the loudness and the rate ofemission pulse can be updated by [60]

1

(1.22)

where α and γ are constants; and 0 < α < 1; 0 < γ

Bat algorithm has the advantage of simplicity and flexibility BA is easy toimplement, and such a simple algorithm can be very flexible to solve a wide range

of problems BA is better than particle swarm optimization (PSO) and geneticalgorithm (GA) in terms of convergence, robustness and precision There are manyreasons for the success of bat-based algorithms By analyzing the key features andupdating equations, we can summarize the following three key points/features [60]:

- Frequency tuning: BA uses echolocation and frequency tuning for problem

solving Though echolocation is not directly used to mimic the true function in reality,frequency variations are used This capability can provide some functionality that may besimilar to the key feature used in particle swarm optimization and harmony search.Therefore, BA possesses the advantages of other swarm-intelligence-based algorithms

- Automatic zooming: BA has a distinct advantage over other metaheuristic

algorithms To be specific, BA has the ability of automatically zooming into a regionwhere promising solutions have been found This zooming is

accompanied by the automatic switch from explorative moves to localintensive exploitation As a result, BA has a quick convergence rate, atleast at early stages of the iterations, compared with other algorithms.

- Parameter control: Many metaheuristic algorithms used fixed parameters

by using some, pre-tuned algorithm-dependent parameters In contrast, BA

uses parameter control, which can vary the values of parameters ( and ) asthe iterations proceed This provides a way to automatically switch fromexploration to exploitation when the optimal solution is approaching Thisgives other advantages of BA over other metaheuristic algorithms

BA can be expressed in a flowchart (Figure 1.9) [59-61] as follows:

Initialization (IS)

- Each bat(i) in the population has been initialized by its parameters

including , , , , and at time step t, in which is being mapped to a solution

for the problem to be solved

- The current global best solution is selected after comparing all thesolutions among all bats in the population based on an objective function The currentglobal best solution is the one that has the best value of the objective function, e.g thesmallest one

- Bats are moved in space using equations (1.18) – (1.20) This leads to new locations of bats, which correspond to new solutions

- Some random bats are moved to new locations around the current global best solution using equation (1.21)

- New solutions are then evaluated based on the objective function Bat(i)

for example, if its new solution (location) is better than the old one, its location will beupdated

- Bats are ranked and a new current global best solution is updated

- If the termination condition is satisfied, the process will end and the currentglobal best solution will become the final global best one to the

problem Otherwise, the process continues with New solutions generation

step for one more iteration, and repeats it.

YES

NS: Generating new local

solutions around the current

global best solution by (1.21)

YES

NS: Accepting the new

solutions; updating rᵢ and Aᵢ

by (1.22)-(1.23)

Start

IS: Initializing population: frequency (fᵢ ), velocity (vᵢ ), pulse

emission rate (rᵢ ), loudness (Aᵢ ), and location/solution (xᵢ )

IS: Finding the current global best solution

based on objective function

NS: Generating new solutions by adjusting fᵢ , updating vᵢ

and xᵢ using (1.18) – (1.20) NS: rand >rᵢ

US: Updating new current global best solution

FS: Is the termination condition satisfied?

1.4 Chapter Conclusions

In this chapter, the fundamental of beamforming has been presented Firstly,basic model of beamforming for smart antenna has been given Then, themathematical basis of beamforming for ULAs has been presented for the arraypattern synthesis Finally, the optimization techniques for beamforming have beenintroduced and focused on the advantages and potential of nature-inspiredoptimization, specifically Bat algorithm These contents will be applied as thefundamental for proposals presented in the next chapters

Chapter 2 General Process to Develop BA-based Adaptive

Beamformers for Interference Suppression

In this chapter, a general process will be developed to build BA-basedadaptive beamformers for pattern nulling of ULAs from problem determination todevelopments of adaptive beamformers steps

2.1 Problem Determination

As mentioned in section I, the increasing number of wireless devices andcrowded frequency spectrum cause serious pollution in the electromagneticpropagation environment In this context, the null-steering adaptive beamformersemerge as a promising solution for interference suppression in wirelesscommunications and radar applications

The BA-based adaptive beamformers for interference suppression applicationwill be developed in following manners:

- Based on the principle presented in chapter 1, in which beamformers areequipped with Direction-Of-Arrival (DOA) estimators (see Appendix A for moredetails);

- Applied for pattern nulling of ULAs including a single null, multiple nulls, and a broad null at directions of interferences;

- Able to maintain the direction of the main lobe and the beamwidth while suppressing the sidelobes