• Tiada Hasil Ditemukan

IMPROVED FRUIT-FLY SWARM ALGORITHM FOR BATHYMETRY SURVEY BY USING AUTONOMOUS

N/A
N/A
Protected

Academic year: 2022

Share "IMPROVED FRUIT-FLY SWARM ALGORITHM FOR BATHYMETRY SURVEY BY USING AUTONOMOUS "

Copied!
24
0
0

Tekspenuh

(1)

IMPROVED FRUIT-FLY SWARM ALGORITHM FOR BATHYMETRY SURVEY BY USING AUTONOMOUS

SURFACE VEHICLES

BY

LWINTHEIN NAING

A dissertation submitted in fulfilment of the requirement for the degree of Master of Science (Mechatronics Engineering)

Kulliyyah of Engineering

International Islamic University Malaysia

SEPTEMBER 2018

(2)

ii

ABSTRACT

Developing a powerful robotic system is not the only solution for solving complicated tasks. In fact, many simple swarm robots can be designed to cooperate and be able to achieve similar or even better result. It is undoubtedly more cost and time efficient to develop this simplistic swarming system. The system can be applied in tasks such as exploration, surveillance, and tracking. In this paper, a swarm optimization algorithm is developed to be used in autonomous surface vehicle (ASV) system in order to locate specific location within the waterbody while performing bathymetry survey.

The developed algorithm is based on the existing fruit-fly optimization algorithm (FOA) and Lévy fruit-fly optimization algorithm (LFOA). These existing algorithms have been developed to be used in practical environment. However, there are several limitations that the system cannot achieve. Thus, the newly proposed algorithm is developed to overcome these constraints in order to obtain better results. The proposed algorithm is called improved-LFOA, or IFOA. It was tested and benchmarked against other several optimization algorithms such as artificial bee colony, particle swarm optimization, covariance matrix adaptation evolution strategy (CMA-ES), FOA, and LFOA. In summary, IFOA’s accuracy is comparable to CMA-ES which is one of the most powerful algorithms when it comes to high-dimensional optimization.

Furthermore, it also performs exceptionally well in terms of convergence rate and accuracy of the results against FOA and LFOA. In another benchmarking against LFOA in 20 restricted virtual environmental conditions, IFOA shows better convergence rate on 19 different conditions. This suggest that IFOA is a suitable algorithm for being used in ASV system for exploration task during bathymetry survey. However, the proposed algorithm still has several limitations that still need to be improved further as it still fails to operate efficiently in certain terrains.

(3)

iii

ثحبلا ةصلاخ

هنإف ،ًةقيقح .ةدقعملا ماهملا لحل ديحولا لحلا ربتعي لا يتوبور ماظن ريوطت نإ م

ملا ن مصت نكم مي

وأ ةلثامم جئاتن قيقحت ىلع نواعتتل ةطيسبلا تاتوبورلا دوشح نم ديدعلا اهنم لضفأ

امم . كش لا

ي .تقولاو ةفلكتلا ثيح نم رثكأ ًلااّعف ربتعي طيسبلا يدشحلا ماظنلا هذه نأ هيف قبط

ه ظنلا اذ ما

زراوخ ريوطت مت ،ةساردلا هذه يف .عبتتلاو ةبقارملاو فاشكتسلااك ماهملا ضعب ىلع ل ةيم

دشح

( ةلقتسملا ةيحطسلا تابكرَملا ماظن يف اهمادختسلا لثمأ

autonomous surface vehicle

) يك

لا ةيمزراوخلا .قامعلأا سايق ةمهم ءادأ عم يئاملا مسجلا يف نيعملا عقوملا ددحت ر ّوطم

دمتعت ة

( ىلث ملا ةهكافلا ةبابذ ةيمزراوخ ىلع

fruit-fly optimization algorithm

يمزراوخ و ) يفيل ة

( ىلثملا ةهكافلا ةبابذل

Lévy fruit-fly optimization algorithm

وخلا هذه ميمصت مت .) تايمزرا

ذل .ماظنلا اذه يف بويعلا نم ديدعلا كانهف ،كلذ عمو .ةيلمع ةئيب يف اهمادختسلا ف ،كل

مت هنإ

مزراوخلا .لضفأ جئاتن ىلع لوصحلاو تاقوعملا ىلع بلغتلل ةحرتق م ةيمزراوخ ميمصت ةي

ةلَّدع ملا ىلثملا ةهكافلا ةبابذل يفيل ةيمزراوخ ىمست ةحرتقملا (

بتخا مت .)

IFOA

اهتنراقمو اهرا

ارتساو تاميسجلل لثملأا دشحلا ،ةيعانصلا لحنلا ةرمعتسمك ىرخأ ىلث م تايمزراوخب ةيجيت

( ةفوفصملا ر ياغت فُّيكت ريوطت

covariance matrix adaptation evolution strategy

)

ملا ةهكافلا ةبابذل يفيل ةيمزراوخو ىلثملا ةهكافلا ةبابذ ةيمزراوخو اصتخاب .ىلث

ف ،ر ةقِد نإ

ُّيكت ريوطت ةيجيتارتساب نراق ت ةلَّدع ملا ىلثملا ةهكافلا ةبابذل يفيل ةيمزراوخ ا ر ياغ ت ف

ةفوفصمل

(

CMA-ES

ثملأا يلاعلا دع بلا صخي اميف تايمزيراوخلا ىوقأ نم ةدحاو ربتعت يتلاو )

.ل و ،كلذك

رياغتلا ةبسن صخي اميف ةيئانثتسا ةروصب ديج اهئادأ نإف خب ةنراقم جئاتنلا ةقدو

مزراو بابذ ةي ة

عم تانراقملا ىدحإ يف .ىلثملا ةهكافلا ةبابذل يفيل ةيمزراوخ و ىلث ملا ةهكافلا راوخ

يل ةيمز يف

يف ىلثملا ةهكافلا ةبابذل 20

كافلا ةبابذل يفيل ةيمزراوخ ترهظأ ،ةيضارتفا ةئيب ملا ىلثملا ةه

ةلَّدع

يف لضفأ رياغت ةبسن 19

ةفلتخم ةلاح لا ةبابذل يفيل ةيمزراوخ نأ ىلإ ريشي اذهو .

ثملا ةهكاف ىل

ةلقتسملا ةيحطسلا تابكرَملا ماظن يف مادختسلال ةبسانم ةيمزراوخ يه ةلَّدع ملا ل

مامهم كتسلاا فاش

لا ضعب نم يناعت تلازام ةحرتقملا ةيمزراوخلا نإف ،كلذ عمو .قامعلأا سايق ءانثأ دويق

ا يتل

ت اهنوكل نيسحت ىلإ جاتحت .ةنيع ملا يضارلأا ضعب يف ءادلأا يف لشف

(4)

iv

APPROVAL PAGE

I certify that I have supervised and read this study and that in my opinion, it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Master of Science (Mechatronics Engineering).

………..

Zulkifli Zainal Abidin Supervisor

………..

Abd. Halim bin Embong Co-Supervisor

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Master of Science (Mechatronics Engineering).

………..

Siti Fauziah binti Toha Internal Examiner

………..

Md. Mozasser Rahman Internal Examiner

This dissertation was submitted to the Department of Mechatronics Engineering and is accepted as a fulfilment of the requirement for the degree of Master of Science (Mechatronics Engineering).

………..

Syamsul Bahrin bin Abdul Hamid Head, Department of

Mechatronics Engineering

This dissertation was submitted to the Kulliyyah of Engineering and is accepted as a fulfilment of the requirement for the degree of Master of Science (Mechatronics Engineering).

………..

Erry Yulian Triblas Adesta Dean, Kulliyyah of Engineering

(5)

v

DECLARATION

I hereby declare that this dissertation is the result of my own investigations, except where otherwise stated. I also declare that it has not been previously or concurrently submitted as a whole for any other degrees at IIUM or other institutions.

Lwinthein Naing

Signature ... Date ...

(6)

vi HT PAGE

INTERNATIONAL ISLAMIC UNIVERSITY MALAYSIA

DECLARATION OF COPYRIGHT AND AFFIRMATION OF FAIR USE OF UNPUBLISHED RESEARCH

IMPROVED FRUIT-FLY SWARM ALGORITHM FOR BATHYMETRY SURVEY BY USING AUTONOMOUS SURFACE

VEHICLES

I declare that the copyright holders of this dissertation are jointly owned by the student and IIUM.

Copyright © 2018 Lwinthein Naing and International Islamic University Malaysia. All rights reserved.

No part of this unpublished research may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise without prior written permission of the copyright holder except as provided below

1. Any material contained in or derived from this unpublished research may be used by others in their writing with due acknowledgement.

2. IIUM or its library will have the right to make and transmit copies (print or electronic) for institutional and academic purposes.

3. The IIUM library will have the right to make, store in a retrieved system and supply copies of this unpublished research if requested by other universities and research libraries.

By signing this form, I acknowledged that I have read and understand the IIUM Intellectual Property Right and Commercialization policy.

Affirmed by Lwinthein Naing

……..……….. ………..

Signature Date

(7)

vii

ACKNOWLEDGEMENTS

All praise due to Allah S.W.T. Firstly, it is my utmost pleasure to dedicate this work to my dear parents and my family, who granted me the gift of their unwavering belief in my ability to accomplish this goal: thank you for your support and patience.

I wish to express my appreciation and thanks to those who provided their time, effort and support for this project. To the members of my dissertation committee, thank you for sticking with me.

Finally, a special thanks to Assistant Professor Zulkifli Zainal Abidin and Assistant Professor Abd Halim bin Embong for their continuous support, encouragement and leadership, and for that, I will be forever grateful.

(8)

viii

TABLE OF CONTENTS

Abstract ...ii

Abstract in Arabic...iii

Approval Page ...iv

Declaration ...v

Copyright Page ...vi

Acknowledgements ...vii

List of Tables...x

List of Figures ...xi

List of Abbreviation...x

List of Symbols ...xi

CHAPTER ONE: INTRODUCTION ...1

1.1 Overview ...1

1.1.1 Bathymetry Survey and its Automatization ...1

1.1.2 Swarm Intelligence and Optimization ...2

1.2 Statement of the Problem ...4

1.3 Research Objectives ...4

1.4 Research Methodology ...5

1.5 Scope of Research ...6

1.6 Dissertation Outline ...7

CHAPTER TWO: LITERATURE REVIEW ...8

2.1 Introduction ...8

2.2 Swarm Behavior in Biology ...8

2.2.1 Decentralised Decision Making in Insects ...9

2.2.2 Coordinated Movement in Animals ...10

2.3 Type of Swarm Intelligence ...10

2.3.1 Swarm Intelligence in Optimization ...11

2.3.2 Swarm Robotics...12

2.4 Fruit-fly Optimization Algorithm ...15

2.4.1 Fruit-fly Optimization Algorithm 1 ...16

2.4.2 Fruit-fly Optimization Algorithm 2 ...18

2.4.3 Difference Between the Two Fruit-fly Optimization Algorithms ...19

2.5 Lévy-flight Fruit-fly Optimization Algorithm ...20

2.5.1 The Inspirational Behaviour of Fruit-fly...20

2.5.2 Lévy Distribution and Lévy Random Flight ...21

2.5.3 Realistic Model for Communication Range ...25

2.5.4 Mathematical Description, Flowchart, and Pseudo Code ...26

2.6 Chapter Summary ...30

CHAPTER THREE: IMPROVED LÉVY FRUIT-FLY OPTIMIZATION ALGORITHM ...31

3.1 Introduction ...31

3.2 Improved Agent’s Local Best System ...32

(9)

ix

3.3 Lambda Factor for Dynamic Step-size Reduction...33

3.4 Self-adaptive Ability ...35

3.5 Mathematical Description ...36

3.6 Chapter Summary ...40

CHAPTER FOUR: ANALYSIS AND DISCUSSION ...41

4.1 Introduction ...41

4.2 Test Functions ...41

4.2.1 Equation and Parameter ...44

4.2.2 Landscape Map ...44

4.3 Benchmark Result in Normal Condition ...50

4.3.1 IFOA Against Other Types of Optimization Algorithm ...50

4.3.2 IFOA Against Other FOA Series ...51

4.4 Benchmark Result in Restricted Conditions Imitating Actual Environment ...63

4.5 Chapter Summary ...65

CHAPTER FIVE: CONCLUSION AND RECOMMENDATIONS ...66

5.1 Conclusion ...66

5.2 Recommendations ...67

REFERENCES ...69

LIST OF PUBLICATIONS ...73

APPENDIX ...74

(10)

x

LIST OF TABLES

Table 4.1 List of test functions 42

Table 4.2 Average RMSE of algorithms at different test functions 51 Table 4.3 Summary of the restricted conditions 63 Table 4.4 Average RMSE of LFOA and IFOA in restricted conditions 64

(11)

xi

LIST OF FIGURES

Figure 1.1 Screenshot of waypoint navigation system used in ASV 2

Figure 2.1 Drosophila Melanogaster 15

Figure 2.2 FOA1 technique in searching for food 16

Figure 2.3 DrosoBots and FOA2’s flowchart 18

Figure 2.4 Comparison between Cauchy, Lévy, and Gaussian

distribution 21

Figure 2.5 The simulated path of fruit-fly calculated from LRF with 𝛽 =

1 24

Figure 2.6 The simulated path of fruit-fly calculated from LRF with 𝛽 =

2 24

Figure 2.7 Representation of agent’s local best in LFOA 26

Figure 2.8 Flowchart for LFOA 28

Figure 2.9 Pseudo code for LFOA 29

Figure 3.1 Representation of improved agent’s local best system

through mesh tropology 33

Figure 3.2 Behaviour of lambda factor at different minimal value 34 Figure 3.3 Representation of how the probability of convergence is

below half 35

Figure 3.4 Flowchart for IFOA 38

Figure 3.5 Pseudo code for IFOA 39

Figure 4.1 Landscape maps of two-dimensional test functions 45 Figure 4.2 Graphs of RMSE of FOA, LFOA, and IFOA in 1000

iterations 52

(12)

xii

LIST OF ABBREVIATIONS

ABC Artificial Bee Colony ACO Ant Colony Optimization AI artificial intelligence

ASV autonomous surface vehicle / autonomous surface vessel CMA-ES Covariance Matrix Adaptation Evolution Strategy DE Differential Evolution

FOA Fruit-fly Optimization Algorithm FOA1 Pan’s Fruit-fly Optimization Algorithm

FOA2 Abidin et al.’s Fruit-fly Optimization Algorithm GRNN general regression neural network

IFOA Improved Lévy Fruit-fly Optimization Algorithm LFOA Lévy-flight Fruit-fly Optimization Algorithm LRF Lévy random flight

PDF probability density function PSO Particle Swarm Optimization

RC remote controller

RMSE root mean square error

SI swarm intelligence

(13)

xiii

LIST OF SYMBOLS

𝑑𝑖𝑠𝑡𝑖𝑗 cartesian distance between 𝑖 and 𝑗

𝑟 communication radius

ℱ Fourier transform

𝑓 function

𝑔𝑙𝑜𝑏𝐴𝑡𝑡 global attractiveness

𝑃0 initial self-adaptive probability

∩ intersection

𝜆 lambda factor

𝜆𝑚𝑎𝑥 lambda maximum 𝜆𝑚𝑖𝑛 lambda minimum 𝐿𝑒́𝑣𝑦𝐴𝑔𝑔 Lévy aggressiveness

𝐼𝑡𝑒𝑟𝑚𝑎𝑥 maximum number of iteration 𝜇 mean / shift parameter

𝜋 pi

𝑃 probability

𝑃𝑖𝑛𝑐 probability increment

𝛽 skewness parameter

𝛼 stability index / Lévy index

𝜎 standard deviation / scale parameter 𝑡𝑎𝑟𝑔𝐴𝑐𝑐 target accuracy

𝑛 total number of agents

𝑋𝐿𝐵 X lower bound

(14)

xiv

𝑋𝑈𝐵 X upper bound

𝑥 x position

𝑌𝐿𝐵 Y lower bound

𝑌𝑈𝐵 Y upper bound

𝑦 y position

(15)

1

CHAPTER ONE INTRODUCTION

1.1 OVERVIEW

1.1.1 Bathymetry Survey and its Automatization

Regularly monitoring waterbody such as lake, dam, or river is essential for conserving environment and preventing flood hazards. Idris et al. (2016) mention that bathymetry, current speed, salinity, oxygen concentration, and other chemical contaminants are important data which need to be collected during monitoring process. These data are significant for evaluating and predicting environmental conditions for water-related disaster prevention or environmental protection purposes.

Normally, bathymetry survey is performed together with other monitoring processes by navigating a boat along the surface of water. The boat is equipped with several sensors for data collection. At least one crew member is needed in order to conduct the operation, thus the process is time consuming, expensive, and inefficient.

With the help of technology advancement, contemporary bathymetry survey can be performed independently without human involvement. An unmanned surface vehicle (USV) or autonomous surface vehicle (ASV) can be used to replace a boat during waterbody monitoring process. Idris et al. (2016) further state that navigation system can be alternated between supervising the ASV from a control room or manually operate it using a remote control (RC).

ASV’s supervision from a control room or ground station is demonstrated in Abidin, Arshad, and Ngah (2010), Hitz et al. (2012), and Capello, Guglieri, and Quagliotti (2013) using an open-access software called waypoint navigation system (WNS). It is used to predetermine ASV’s path so that its routes can be planned in

(16)

2

advance. Fig. 1.1 illustrates an actual planned path on a WNS interface. When the system initiates, the ASV will follow the waypoints from one position to another according to the plan. A satellite images from Google Maps can be integrated into the interface for convenience purposes while planning the path.

Fig 1.1: Screenshot of waypoint navigation system used in ASV (Naing et al., 2017)

Nevertheless, the system still cannot be regarded as fully automatic since the path itself needs to be manually planned. Furthermore, in order for the autonomous bathymetry system to be practically viable for an actual fieldwork, using WNS alone is insufficient. Images from satellite are occasionally inaccurate and most of the time they do not display any obstacles. Thus, it is impossible for the user to plan a perfect route by referring to images from Google Maps alone. Instead of relying solely on the WNS, an object avoidance system is also a necessary component that must be integrated into the system as well. Therefore, designing a fully autonomous bathymetry system is still an area that needs to be further explored.

1.1.2 Swarm Intelligence and Optimization

In nature, some species of living organisms may live on its own while many other species stay together and form societies. They provide for one another and help one

(17)

3

another in order to survive. They have crucial collaborated behaviours such as foraging techniques or predator avoidance strategies which are necessary for their survival. Swarm intelligence (SI) studies these collaborated behaviours and develops them into effective optimization systems. Şahin et al. (2008) reveal that several successful optimization algorithms are initiated from simple swarm behaviour of organisms such as ant, bee, bird, or fish.

There are many great examples of nature-inspired optimization algorithms being continuously developed. For instance, the behaviour of a school of fish beautifully and effectively avoiding predators is developed into fish swarm optimization (Bone, and Moore, 2008; Lobato, and Steffen Jr., 2014); an ant colony optimization originates from a foraging behaviour of ants (Abudi, and Misra, 2014;

Jackson, and Ratnieks, 2006); a flash-light communication technique of fireflies inspire firefly novel optimization (Gandomi et al., 2011; Klempka, and Filipowicz, 2018); and an algorithm based on foraging behaviour of fruit fly known as fruit fly optimization (Abidin et al., 2011; Pan, 2012). These swarm optimization algorithms are just a few examples from the body of knowledge.

Swarm optimization is remarkably demonstrated in robotic system as well. It is a system that collaborates and coordinates multiple robots in order to perform specific tasks. Hoff et al. (2010) mention that there is neither leader nor follower among these robots, each of them shares data that it has been collected with one another and behaves in a way that would benefit the swarm most. Furthermore, Idris (2016) informs that the system is robust, flexible, and scalable. Its reliability is high because a failure in one robot does not affect the whole operation. The system can adapt itself easily while performing its task. In some studies, swarm robotic system has been applied in waterbody monitoring system as well. For instance, several units of ASVs

(18)

4

were deployed to monitor water quality after Deepwater Horizon oil spill in 2010 off the coast of Louisiana (Mukhopadhyay et al., 2014). A well-developed SI system together with WNS can automatically generate paths for ASV (Abidin et al., 2012;

Majid, and Arshad, 2016).

1.2 STATEMENT OF THE PROBLEM

Bathymetry survey system and ASV have been heavily studied in the past (Bakar, M., and Arshad, M., 2017; Haworth et al., 2016; Idris et al., 2016; Naing et al., 2017).

However, most of them focus on navigation and object avoidance aspect of the ASV, while only a few studies attempt on swarm integration. An effective swarm robotic bathymetry survey system is difficult to achieve, but it is not impossible. A successfully developed system would certainly revolutionised bathymetry system.

Fruit fly optimization algorithm (FOA) developed by Abidin et al. (2011) and Lévy-flight fruit fly optimization algorithm (LFOA) developed by Idris (2016) are swarm robotic optimization algorithms aim to be used in search mission during ASV’s waterbody monitoring system. Unlike other optimization algorithms which are used in computational purposes, FOA and LFOA must include several restrictions in their design to represent physical limitations of an actual swarm robotic system. However, these algorithms are not very efficient and it can be further developed in order to improve its effectiveness.

1.3 RESEARCH OBJECTIVES

The study aims to achieve the following objectives:

(19)

5

1- To modify and improve the Lévy-flight Fruit-fly Optimization Algorithm to make it comparable to the original Fruit-fly Optimization Algorithm and suitable to be used in bathymetry survey.

2- To include physical limitations of the autonomous surface vehicle and the environment into account when benchmarking the algorithm.

3- To analyse the performance of the improved Lévy-flight Fruit-fly Optimization Algorithm via simulation in different virtual environment generated from multiple test functions.

1.4 RESEARCH METHODOLOGY

The first step in the research is to extensively review literatures as well as existing and on-going works conducted by other researchers on swarm intelligence, swarm robotics, Fruit-fly Optimization Algorithm, and Lévy-flight Fruit-fly Optimization Algorithm of swarming ASV. This is to obtain the latest information and ideas regarding the topic related to the paper and different kind of technique and improvement done in the past. It would help in forming a better concept idea for the proposed algorithm in this paper.

After analysing FOA and LFOA, the next step is to develop the new version of LFOA. The proposed algorithm for this paper will be called Improved Lévy Fruit-fly Optimization Algorithm (IFOA). Each individual component of the previous LFOA will need to be broken down in detail and improve accordingly. Flowchart and pseudo code would help in explaining how the algorithm performs. Not only that, other techniques used in solving the optimization problem will be critically analysed and included in the proposed algorithm.

(20)

6

The IFOA will then be analysed via simulation. The algorithm will be benchmarked with other well-known algorithms such as Particle Swarm Optimization (PSO), Artificial Bee Colony (ABC), Covariance Matrix Adaptation Evolution Strategy (CMA-ES), as well as the original FOA and LFOA. Several 2-dimensional test functions will be used in the analysis process to determine the performance of IFOA.

Afterward, IFOA will be compared to LFOA in the test that involves physical limitations of ASV. Only these two algorithms are selected to be tested in this step because they are designed and developed to be used in the actual physical environment, whereas other algorithms are developed for solving computation and mathematical problems only.

This research would end with a conclusion and recommendation for further analysis and improvement that can be made on the algorithm.

1.5 SCOPE OF RESEARCH

This research focuses on improving the existing LFOA into a more efficient version that can be compared with other optimization algorithms. There are still several problems faced by LFOA that cannot be overcome, such as its ineffective data sharing system and its inability to search accurately for the desired locations. The research will highly focus on these aspects while maintaining the design of the algorithm that is suitable for real environment. However, the algorithm does not include several physical aspects such as the volume of the ASV agents or its momentum while turning in the simulation.

The test functions used for benchmarking are simple 2D test functions so that they can represent the physical coordinates of the real environment. It would be

(21)

7

challenging to develop a powerful enough algorithm to be compared to others, such as PSO, ABC, or CMA-ES, that can even perform in more than 50-dimension test functions.

1.6 DISSERTATION OUTLINE

Chapter One discusses about the introduction of the dissertation along with the overall research conducted in the past. It also includes statement of the problem, objectives, research methodology and scope of the research.

Chapter Two discusses about other existing swarming optimization algorithms in detail. It mainly focuses on the FOA and LFOA which are the related topic of interest.

The logic of these algorithms is explained in detail, and the disadvantages and advantages are critically discussed.

Chapter Three explains in details about the features of IFOA. It covers the limitations that can be observed in LFOA and discusses appropriate solutions in solving them. The programming code of IFOA is written in MatlabTM file without using any Matlab Toolbox. All of the display of iterations are manually designed and coded.

Chapter Four analyses the performance of IFOA by benchmarking it to other well- known algorithms. Mostly, it focuses on the improvement made by comparing the proposed algorithm to FOA and LFOA. The benchmarking is done in MatlabTM software which is widely used in engineering application.

Chapter Five contains conclusion and recommendation. Summary of the dissertation and possible improvements are discussed.

(22)

8

CHAPTER TWO LITERATURE REVIEW

2.1 INTRODUCTION

This chapter explains about swarm intelligence which includes swarm optimization and swarm robotics. Fruit fly optimization algorithm (FOA) and Lévy-flight fruit fly optimization algorithm (LFOA) are further revealed in this chapter as well. Their abilities and limitations are analysed and discussed.

2.2 SWARM BEHAVIOR IN BIOLOGY

Before understanding artificial swarm intelligence, it is necessary to understand how this principle came into existence. Swarm intelligence in engineering imitate the natural behaviours that can be observed in nature as quoted:

“…swarm intelligence is biology. For millions of years many biological systems have solved complex problems by sharing information with group members. By carefully studying the underlying individual behaviours and combining behavioural observations with mathematical or simulation modelling we are now able to understand the underlying mechanisms of collective behaviour in biological systems. We use examples from the insect world to illustrate how patterns are formed, how collective decisions are made and how groups comprised of large numbers of insects are able to move as one”

(Beekman et al., 2008).

In swarm behaviour, the agents can communicate to and help one another to complete specific tasks that are assigned by the users which are impossible or too

(23)

9

difficult to complete individually (Blum, and Li, 2008). The system is inspired by collective behaviour of social insects or animal societies, such as ants, termites, bees, wasps, flocks of birds, or schools of fish. However, even though some systems may look similar to one another, it cannot be assumed that the system will behave similarly. The real biological inspiration does not come from some basic similarities of the systems; instead sophisticated and exquisite differences in different systems are the characteristics that needed to be observed. Nevertheless, many biological swarms have two characteristics that they share in common which are crucial for their survival ability. These characteristics need to be studied in detail in order to develop a functional artificial swarm system; these characteristics are: decentralized decision making and coordinated movement (Beekman et al., 2008).

2.2.1 Decentralised Decision Making in Insects

Insects, such as bees, warps, ants, and termites, do live together within a small nest.

The requirement for information sharing and transfer among the group members are essential for them to survive. As such, it is impossible for it to live alone in the community. Their actions are tuned and harmonized which allow them to achieve adaptive behaviour that allows all of them to work together. This is of utmost important for the insect colonies as they need to make several collective and unified decisions to survive. For instance, they must make important decision on where to forage, where to relocate to, when to reproduce, and how to split the tasks among the members. Obviously, these decisions and actions are influenced mainly by local information obtained from interactions with individuals as well as the surrounding environment, which were found in the study of Bonabeau et al. (1997) and Camazine et al. (2001) (as cited in Beekman et al., 2008). This mean that everyone would obtain

(24)

10

different local information depending on its situation, but this would contribute to the global decision and shape the decision of the whole colonies. In this context, it can be said that the decision making the insect societies is decentralized.

2.2.2 Coordinated Movement in Animals

In many animal species that live in groups, individuals travel over long distance in groups for seasonal migration or searching for food sources. Some species even hunt for prey or escape from predators by coordinating their movements with one another.

These movements are not controlled by a single leader within each group, rather, they are the movements that organized naturally through the local interactions between individuals (Şahin et al., 2008). Self-organized movement help the swarms avoid dangers that can harm them, and it can also help them in the foraging process as well.

These movements can be observed in both insects and animals, such as in colony of bee, flock of bird, school of fish, or swarm of bat.

2.3 TYPE OF SWARM INTELLIGENCE

Swarm intelligence (SI) is part of an artificial intelligence (AI) discipline that involves either multiple independent robots or artificial agents. The properties of SI mimic those of natural behaviours found in swarm of living organisms. SI can be categorized into two types depending on its objectives; the first type is swarm optimization which is mainly used in data-mining or solving mathematical problems, while the second type is swarm robotic which involve physical independent robots that coordinate with one another to complete determined tasks. These two categories are further explained in more detail below.

Kulliyyah o

Rujukan

DOKUMEN BERKAITAN

The vehicle steering control system consists of Fuzzy Logic Control (FLC) and the Proportional, Integral and Derivative (PID) control are built in cascade, in which FLC is used

In this paper, Modified Particle Swarm Optimization (MPSO) technique is demonstrated for the reduction of real power losses in the radial network using NR.. The presented technique

A direct control based maximum power point tracking method for photovoltaic system under partial shading conditions using particle swarm optimization algorithm..

In fact, all steps of genetic algorithm (GA), particle swarm optimization (PSO) and imperialistic competitive algorithm (ICA) will be well mentioned in details and

This paper present a novel optimization method, Time Varying Acceleration – Rank Evolutionary Particle Swarm Optimization (TVA- REPSO) in solving optimum generator sizing for

This paper presents a study on optimal location and sizing of multiple FACTS devices based on Particle Swarm Optimization (PSO) for minimization of transmission

Thus, the purpose of this work is to adapt the SMC technique for the control of ASS, where particle swarm optimization (PSO) algorithm is utilized to design the sliding surface

In this paper, a novel scheme based on particle swarm optimization (PSO) algorithm is proposed to improve the variable speed control of IFOC in TIM.. The PSO