Search

Automatic generation of art has aided artists in creating novel 2-dimensional artforms in a field known as evolutionary art. The development of technology on 3D printing has attracted people of different fields to designing 3D shapes for art display, an architecture design, a prototype or even an aiding tool. However, current automatic generation algorithms have focused primarily on 2D art only and have not successfully crossed the realm into 3D. This paper introduces an automatic method for generation of 3D artforms using electroencephalogram (EEG). Interactive evolutionary computing (IEC) is the typical method used to evolve art, however, IEC is often reported to cause significant fatigue among its users due to its repetitive and time-consuming characteristics. With the combination of brain-computer interface (BCI) and IEC, the hope is to speed up the interactive process, hence reducing user fatigue and at the same time be able to produce promising solutions. The proposed approach focuses on understanding human emotions in order to identify user preferences of the 3D shapes automatically and further evolve the selected artform. The use of a natureinspired shape generation formula, called the Gielis supershape, is used as the encoding for generation of 3D art and is shown to be able to bring forth large varieties of shapes during the automatic evolutionary optimization process.

Emotions play an important role in our daily life. AntonioDamasiofamously stated: “We are not thinking machines that feel; rather we are feeling machines that think”. Human emotions can be recognized through facial expression, speech and gesture. The use of electroencephalograms (EEGs) to understand and recognize human emotion has been widely studied, where those recognition techniques greatly benefit in human-computer interaction (HCI). In this investigation, we study the use of EEGs to recognize emotions. Pleasant and unpleasant images are used as stimuli to elicit human pleasant and unpleasant emotions.The brainwaves are recorded using a 9-electrode medical grade wireless EEG headset from Advance Brain Monitoring (ABM), the B-alert X10. The features comprising alpha, beta, gamma, theta, and delta bands are then extracted from the recorded brainwaves using time-frequency analysis. Different channels and rhythms are used in support vector machine (SVM) and K-nearest neighbors (KNN) classifiers to train and classify pleasant vs. unpleasant mind states. The best accuracy obtained was 70.43% using SVM with alpha and beta rhythmsfromchannel F3 and Fz, and also with gamma rhythms from channel POz.

Over the last decade, various mobile robots have been developed and widely used in myriad sectors. However, the vast majority of mobile robots are manually designed where the designers must have the preliminary knowledge of the interaction between the robots with the environment. Additionally, the high complexity involved in the design of the kinematics and controllers of a mobile robot has always been the biggest challenge for the researchers and practitioners alike. Thus, the task of designing a robot can be considered very demanding and extremely challenging. In this research, an artificial evolution approach utilizing Single-Objective Evolutionary Algorithm (SOEA) and Multi-Objective Evolutionary Algorithm (MOEA) respectively are investigated in the automatic design and optimization of the morphology of a Six Articulated-Wheeled Robot (SAWR) with climbing ability. Results show that SOEA is able to produce optimized SAWR with climbing ability while MOEA is able to produce a set of Pareto optimal solutions which provide users with a choice of solutions for trade-off between the objectives of morphology size and climbing performance. The Pareto optimal set of solutions are the smallest SAWR with least climbing ability to biggest SAWR with the best climbing motion. The research continues by transferring the evolved solutions from simulation to the real world using 3D printing. The body, legs and wheels of the evolved robots are printed by a 3D printer and assembled with sensors, servos and motors for real world testing. Results show that the fabricated real world SAWRs are able to perform the climbing motion with the average accuracy of 85.8% in comparing to the performance in simulation.

Emotion is an important part of human and it plays important role in human communication. Nowadays, as the use of machine getting more common, the human computer interaction (HCI) has become important. The understanding of user could bring across a better aiding machine. The exploration of using EEG in understanding human is widely studied for benefit in several fields such as neuromarketing and HCI. In this study, we compare the use of 2 different stimuli (3D shapes with motion vs. 2D emotional images that are static) in attempting to classify positive versus negative feelings. A medical-grade 9-electrode Advance Brain Monitoring (ABM) B-alert X10 is used as the brain-computer interface (BCI) acquisition device to obtain the EEG signals. 4 subjects are involved in recording brain signals during viewing 2 types of stimuli. Feature extraction is then applied to the acquired EEG signals to obtain the alpha, beta, gamma, theta and delta rhythms as features using time frequency analysis. Support vector machine (SVM) and K-nearest neighbors (KNN) classifiers are used to train and classify positive and negative feelings for both stimuli using different channels and rhythms. The average accuracy of 3D motion shapes are better than the average accuracy of the 2D static emotional images for both SVM and KNN with 69.88% and 56.35% using SVM for 3D motion shapes and emotional images respectively, and also 65.31% and 55.45% using KNN for 3D motion shapes and emotional images respectively. This study shows that the parietal lobe are more informative in the classification of 3D motion shapes while the Fz channel of the frontal lobe is more informative in classification of 2D static emotional images.

By hybridizing Evolutionary Programming (EP) with Interactive Evolutionary Algorithm (IEA), game rules and its playing environment will be automatically generated for an arcade-type game that can be played on the Android mobile platform. In this study, mutation rates of 0.7 and 0.9 are used to generate both the game rules and the game environment for the mobile game. Players are used as the evaluator instead of the conventional mathematical fitness functions and hence the motivation for using high mutation rate is that they are able to generate higher levels of diversity during the optimization runs. This interactive mode of game-playing cum evaluation will enable the creation of games that can fit the user’s preferences as well as styles of game-playing. Experiments show a very positive result where very good evaluation scores were obtained from the users. This shows that with a high mutation rate, the hybridized EP with IEA approach can generate rules and environments that are well-accepted and liked by human players.

The objective of this study is to focus on the automatic generation of game artificial intelligence (AI) controllers for Ms. Pac-Man agent by using artificial neural network (ANN) and multiobjective artificial evolution. The Pareto Archived Evolution Strategy (PAES) is used to generate a Pareto optimal set of ANNs that optimize the conflicting objectives of maximizing Ms. Pac-Man scores (screen-capture mode) and minimizing neural network complexity. This proposed algorithm is called Pareto Archived Evolution Strategy Neural Network or PAESNet. Three different architectures of PAESNet were investigated, namely, PAESNet with fixed number of hidden neurons (PAESNet_F), PAESNet with varied number of hidden neurons (PAESNet_V), and the PAESNet with multiobjective techniques (PAESNet_M). A comparison between the single- versus multiobjective optimization is conducted in both training and testing processes. In general, therefore, it seems that PAESNet_F yielded better results in training phase. But the PAESNet_M successfully reduces the runtime operation and complexity of ANN by minimizing the number of hidden neurons needed in hidden layer and also it provides better generalization capability for controlling the game agent in a nondeterministic and dynamic environment.

This paper presents an initial approach for creating autonomous controllers for the car racing game using a hybrid technique. The Differential Evolution (DE) algorithm is combined with Feed-forward Artificial Neural Networks (FFANNs) to generate the required intelligent controllers in a well-known car racing game, namely The Open Racing Car Simulator (TORCS). TORCS is used as a platform in most of the IEEE conference competitions. The main objective of this research is to test the feasibility of the DE implementation in TORCS platform. The literature showed that the application of DE in Real Time Strategy game returned promising results in evolving the required strategy gaming controllers. Interestingly, there is still no study thus far that has been conducted in applying DE into TORCS game platform. This research result shows that DE performed well in TORCS even though a very simple fitness function was used. This indicates that DE has well-tuned the neural network weights to generate optimal and sub-optimal controllers in TORCS.

Recently, the growth of Artificial Intelligence (AI) has provided a set of effective techniques for designing computer-based controllers to perform various tasks autonomously in game area, specifically to produce intelligent optimal game controllers for playing video and computer games. This paper explores the use of the competitive fitness strategy: K Random Opponents (KRO) in a multiobjective approach for evolving Artificial Neural Networks (ANNs) that act as controllers for the Ms. Pac-man agent. The Pareto Archived Evolution Strategy (PAES) algorithm is used to generate a Pareto optimal set of ANNs that optimize the conflicting objectives of maximizing game scores and minimizing neural network complexity. Furthermore, an improved version, namely PAESNet_KRO, is proposed, which incorporates in contrast to its predecessor KRO strategy. The results are compared with PAESNet. From the discussions, it is found that PAESNet_KRO provides better solutions than

This paper presents the design and evaluation of a full AI controller for Real-Time Strategy (RTS) games using techniques from Evolutionary Computing (EC). The design is novel in its use of a modified Pareto Differential Evolution (PDE) algorithm for bi-objective optimization of the weights of an Artificial Neural Network (ANN) controller when only single-objective optimization has so far been studied. The two main aims of this research are to: (1) develop controllers capable of defeating opponents of varying difficulty levels, which may assist in commercial RTS AI development, and (2) minimize the number of neurons used in the ANN architecture, an issue primarily of efficiency. Experimental results using the popular Warcraft III platform demonstrate success with both aims: the optimized controller was able to win any battle using only a minimal number of hidden neurons, but sub-optimal controllers were able to provide opponents of any intermediate difficulty.

The main objective of this paper is to investigate online evolution of military unit combination strategies for winning an offensive rush in a real-time strategy (RTS) game. A modified version of Evolutionary Programming (EP) is used as the evolutionary optimizer while WARGUS is used as the RTS gaming environment. Evolution of the military unit combinations is conducted online, which means that optimization is taking place while a particular round of the RTS game is still in progress. Empirical tests show that the online evolution of military unit combination strategies is possible using EP and was able to mount successful offensive campaigns on a reliable basis against three respective built-in, human-crafted AI strategies provided in WARGUS.

Adaptive modulation techniques in wireless communications are paramount technologies designed to thrive in fading and noisy environments. Methods as such require an accurate estimate of the channel condition at the receivers' end in order for the transmitter to adapt appropriately. With that in mind, a communication link which changes the degree of modulation scheme according to the estimated signal-to-noise ratio (SNR) values is proposed. Hence, the objective of the system is to achieve acceptable quality in a changing communication link. In this paper, spectral efficiency and average bit error rate of the overall system is measured. Monte Carlo simulations of different signals and channel conditions corroborate our analysis and discussion.

Evolutionary multi-objective optimization (EMO) has gained popularity and it has been successfully applied in several research areas. Based on the literature review conducted, EMO approach has not been applied in any Go game application. In this study, artificial neural networks (ANNs) are evolved with an EMO algorithm, Pareto Archived Evolution Strategies (PAES) for computer player to learn and play the 7x7 board Go game against GNU Go. In this study, two conflicting objectives are investigated: first, maximize the ability of neural player to play the Go game and second, minimize the complexity of the ANN by reducing the hidden units. Several comparative empirical experiments were conducted that showed EMO which optimize two distinct and conflicting objectives outperformed the single-objective (SO) optimization which only optimized the first objective with no pressure selection on the second objective.

Over the last decade, various mobile robots have been developed and widely used in myriad sectors. However, the vast majority of mobile robots are manually designed where the designers must have the preliminary knowledge of the interaction between the robots and the environment. Additionally, the high complexity involved in the design of the kinematics and controllers of a mobile robot has always been the biggest challenge for researchers and practitioners alike. Thus, the task of designing a robot can be considered very demanding and extremely challenging. In this research, an artificial evolution approach utilizing Single-Objective Evolutionary Algorithm (SOEA) and Multi-Objective Evolutionary Algorithm (MOEA) respectively are investigated in the automatic design and optimization of the morphology of a Six Articulated-Wheeled Robot (SAWR) with climbing ability. Results show that the SOEA is able to produce optimized SAWR with climbing ability while the MOEA is able to produce a set of Pareto optimal solutions which provide users with a choice of solutions for trade-off between the objectives of morphology size and climbing performance. The Pareto optimal set of solutions are the smallest SAWR with the least climbing ability to the biggest SAWR with the best climbing ability. The research continues by transferring the evolved solutions from simulation to the real world using 30 printing. The body, legs and wheels of the evolved robots are printed by a 30 printer and assembled with sensors, servos and motors for real world testing. Results show that the fabricated real world SAWRs were able to perform the climbing motion with an average accuracy of 80.9% compared to the performance in simulation.

Numerous traditional board games such as Backgammon, Chess, Tic-Tac-Toc, Othello, Checkers, and Go have been used as research test-beds for assessing the performance of myriad computational intelligence systems including evolutionary algorithms (EAs) and artificial neural networks (ANNs). Approaches included building intelligent search algorithms to find the required solutions in such board games by searching through the solutions space stochastically. Recently, one particular type of search algorithm has been receiving a lot of interest in solving such kinds of game problems, which is the multi-objective evolutionary algorithms (MOEAs). Unlike single-objective optimization based search algorithms, MOEAs are able to find a set of non-dominated solutions which trades-off among all the conflicting objectives. In this study, the utilization of a multi-objective approach in evolving ANNs for Go game is investigated. A simple three layered feed-forward ANN is used and evolved with Pareto Archived Evolution Strategies (PAES) for computer players to learn and play the small board Go games.

This paper presents an empirical comparison of two sigmoidal-type activation functions in evolutionary artificial neural network models. They are the log-sigmoid and hyperbolic tangent sigmoid activation functions which were investigated in order for evolving neural network controllers to play a classic video game. A Hill-Climbing Neural Network (HillClimbNet) was developed using the hill-climbing method together with a feedforward neural network to automatically create an intelligent controller that can play the screen-capture of Ms. Pac-man arcade game. The experimental results showed that that the HillClimbNet with log-sigmoid outperforms the HillClimbNet with hyperbolic tangent sigmoid when used in the hidden and output layers of the network when the agent plays the game.

This paper presents the research results found for the utilization of a Genetic Algorithm (GA) technique in evolving a set of Artificial Neural Networks (ANNs) weights which functions as controller in deciding what type of unit that should be spawned for winning against the opponent in a RTS game called Warcraft 3 (custom map). The elitism concept is applied during the optimization processes in order to avoid losing good solutions. The experimentation results show clearly a group of mixed randomized opponent can be defeated by the generated AI army. Hence, it is proof that GA is capable to act as a tuning technique in generating the required controllers in RTS game. Furthermore, the neural controllers generated are able to decide the best group of army used in defeating the opponent.

In this paper we present the possibilities of implementing evolution algorithm on a mobile platform. To inspect such possibilities we have created a Pac-man like mobile game that is implemented together with evolutionary programming. Here the interest does not only lie on performing the evolution algorithm on the mobile platform but also to examine the possibilities in implementing interactive evolution and create outputs that are based on the user preferences via reactive feedback, thereby creating a game design that is based upon the users' preferences. By explicitly examining the interactive evolution algorithm on a new testing platform, we have successfully created game rules that are based on the users' playing preferences.

Differential Evolution is known for its simplicity and effectiveness as an evolutionary optimizer. In recent years, many researchers have focused on the exploration of Differential Evolution (DE). The objective of this paper is to show the evolutionary and population dynamics for the empirical testing on 3-Parents Differential Evolution (3PDE) for unconstrained function optimization (Teng et al. 2007). In this paper, 50 repeated evolutionary runs for each of 20 well-known benchmarks were carried out to test the proposed algorithms against the original 4-parents DE algorithm. As a result of the observed evolutionary dynamics, 3PDE-SAF performed the best among the preliminary proposed algorithms that included 3PDE-SACr and 3PDE-SACrF. Subsequently, 3PDE-SAF is chosen for the self-adaptive population size for testing dynamic population sizing methods using the absolute (3PDE-SAF-Abs) and relative (3PDE-SAF-Rel) population size encodings. The final result shows that 3PDE-SAF-Rel produced a better performance and convergence overall compared to all the other proposed algorithms, including the original DE. In terms of population dynamics, the population size in 3PDE-SAF-Abs exhibited disadvantageously high dynamics that caused less efficient results. On the other hand, the population size in 3PDE-SAF-Rel was observed to be approximately constant at ten times the number of variables being optimized, hence giving a better and more stable performance.

Problem statement: The retail sales of computer and video games have grown enormously during the last few years, not just in United States (US), but also all over the world. This is the reason a lot of game developers and academic researchers have focused on game related technologies, such as graphics, audio, physics and Artificial Intelligence (AI) with the goal of creating newer and more fun games. In recent years, there has been an increasing interest in game AI for producing intelligent game objects and characters that can carry out their tasks autonomously. Approach: The aim of this study is an attempt to create an autonomous intelligent controller to play the game with no human intervention. Our approach is to use a simple but powerful evolutionary algorithm called Evolution Strategies (ES) to evolve the connection weights and biases of feed-forward Artificial Neural Networks (ANN) and to examine its learning ability through computational experiments in a non-deterministic and dynamic environment, which is the well-known arcade game, called Ms. Pac-man. The resulting algorithm is referred to as an Evolution Strategies Neural Network or ESNet. Results: The comparison of ESNet with two random systems, Random Direction (RandDir) and Random Neural Network (RandNet) yields promising results. The contribution of this work also focused on the comparison between the ESNet with different mutation probabilities. The results show that ESNet with a high probability with high mean scores recorded compared to the mean scores of RandDir, RandNet and ESNet with a low probability. Conclusion: Overall, the proposed algorithm has a very good performance with a high probability of automatically generating successful game AI controllers for the video game.

A number of studies have demonstrated the capability of ANNs for the required robot behaviors by using an evolutionary optimization technique in generating more complex robot controllers. Interestingly however, there is still a serious lack of research in exploring the application of Evolutionary Multi-objective Optimization (EMO) algorithm in evolutionary robotics. In this paper, we investigate the utilization of a multi-objective approach in evolving artificial neural networks for a simulated autonomous Khepera robot. The generated neural network acts as a controller for radio frequency localization behavior of a Khepera robot. There are two conflicting objectives to be optimized during the evolution: (1) maximize the Khepera robot's behavior for homing towards a RF signal source and (2) minimize the number of hidden neurons used in the robot. The testing results showed the robots were capable to achieve the objective with very few hidden neurons used. Furthermore, the genetic structures of the generated controllers have been further analyzed using the Hinton analysis and the results obtained are presented next.

This paper explores the idea of combining the hillclimbing concept into feed-forward artificial neural networks (ANN) to develop intelligent controllers to play the Ms. Pacman game. The resulting algorithm is referred to as the HillClimbingNet. A comparison with a random system, called RandNet is conducted on the same problem. We also present a survey of the effects of two most popular probability density functions, uniform and Gaussian distributions/mutators on the introduced algorithm. The results clearly indicate the strong potential of the hill-climbing strategy as a direct search method in tandem with a Gaussian-based mutator to optimize the ANN for playing Ms. Pac-Man. ©2010 IEEE.

In this paper, we investigate the utilization of a multi-objective approach for evolving artificial neural networks (ANNs) that act as controllers for a radio frequency (RF) based collective box-pushing task of a group of virtual E-puck robots simulated in a 3D, physics-based environment. The modified Pareto-frontier Differential Evolution (PDE) algorithm is used to generate the Pareto optimal sets of ANN that optimize the conflicting objectives of maximizing the virtual E-puck robots' behaviors for pushing a box towards a wall based on RF-localization as well as minimizing the number of hidden neurons used in its feed-forward ANN controller. A new fitness function used during the collective robotics' optimization process is proposed. The experimentation results showed the virtual E-puck robots were capable of moving towards to the RF signal area. Thereafter, the robots were capable in self-assembling in the signal source area as well as completing the box-pushing towards the target wall with very small neural network architectures. Then, the genetic structures of the generated controllers have been further analyzed using the Hinton analysis. Hence, the results demonstrated that the utilization of the PDE approach in evolutionary robotics can be practically used to generate neural-based controllers for collective robotics' behaviors.

Solving multi-objective optimization problems (MOPs) using evolutionary algorithms (EAs) has been gaining a lot of interest recently. Go is a hard and complex board game. Using EAs, a computer may learn to play the game of Go by playing the games repeatedly and gaining the experience from these repeated plays. In this project, artificial neural networks (ANNs) are evolved with the Pareto Archived Evolution Strategies (PAES) for the computer player to automatically learn and optimally play the small board Go game. ANNs will be automatically evolved with the least amount of complexity (number of hidden units) to optimally play he Go game. The complexity of ANN is of particular importance since it will influence the generalization capability of the evolved network. Hence, there are two conflicting objectives in this study; first is maximizing the Go game fitness score and the second is reducing the complexity in the ANN. Several comparative empirical experiments were conducted that showed that the multi-objective optimization with two distinct and conflicting fitness functions outperformed the single-objective optimization which only optimized the first objective with no selection pressure selection on the second objective. © 2010 IEEE.

Fouling has been a problem since the invention of heat exchanger. The presence of such deposits on heater surface can reduce the efficiency of the heater, which leads to plant shut down and subsequently loss of production. Fouling in the dairy industry is the most severe as daily cleaning of heat exchangers is often required, compared to other industries which only need yearly cleaning. Dairy fouling is caused by the thermal denaturation of betalactoglobulin at temperatures above 70oe. Therefore, this work was carried out to help the understanding of how transport phenomena in a heater affects the protein denaturation process. Modelling of the concentration of the native protein, denatured protein and aggregated protein during heat treatment of dairy product was successfully performed using finite element computational method. Based on the simulated concentration prOfiles, it is clear that native protein is not the contributor to fouling. It was also found that aggregated protein solely cannot be the main foulant. However, there is also possibility where depoSition is caused by both denature and aggregated proteins, but it is more complicated to predict the deposition pattern unless the ratio of the two depositing proteins involved is known through experiment. Experiments were also carried out to validate the simulation work, however, due to the raw materials used, results were found to be inconsistence.

Smartphones have recently become very popular among mobile users. With more powerful smartphones being released by manufacturers around the world, it has encouraged more users to own this gadget. Hence it has opened up a whole new market with its mass among users and attracted a large number of smartphone application developers. It has also attracted the attention of academic researchers to start investigating in this new field. Evolutionary computing has been around for the past few decades and this is a good chance for researchers to try and use evolutionary computing techniques on another platform. Testing on different platforms would allow us to view the capabilities of evolutionary computing to perform under limited computational resources such as that of smartphones. The aim of this research is to develop an application that contains evolutionary computing techniques that will be used by the user as an entertainment tool. Evolutionary Programming is used as the main evolutionary technique in this study. An arcade-type game has been created that serves as a test-bed. This game contains no rules so that it can allow Evolutionary Programming to generate the rules to be played by the mobile users. Since it involves human users, adding Interactive Evolutionary Algorithm together with Evolutionary Programming has allowed further rules to be generated that are based on the user's preferences. The preliminary testing has proven that Evolutionary Programming and lEA can be run on the mobile platform without any major difficulties. Through the testing conducted, it has been observed that finding a suitable mutation rate and population size is very vital. This is due to the probability of the generated rules having the ability to be propelled out of a local optimum with more variety and flexibility in generating different offspring that might contain better results. The results from the experiments has shown that mutation rates of 0.7 and 0.9 generates higher and better

The aim of this paper is to use a simple but powerful evolutionary algorithm called Evolution Strategies (ES) to evolve the connection weights and biases of feed-forward artificial neural networks (ANN) and to examine its learning ability through computational experiments in a non-deterministic and dynamic environment, which is the well-known arcade game called Ms. Pac-man.The resulting algorithm is referred to as an Evolution Strategies Neural Network or ESNet.This study is an attempt to create an autonomous intelligent controller to play the game.The comparison of ESNet with two random systems, Random Direction (RandDir) and Random Neural Network (RandNet) yields promising results.

This paper investigates the application of selfadaptation genetic algorithm on a flexible and configurable heuristic decision making framework that can tackle the problem of bidding across multiple auctions that apply different protocols (English, Vickrey and Dutch) by using an autonomous agent to search for the most effective strategies (offline). Our study shows that self-adaptation genetic algorithm performance is much better than conventional genetic algorithm. An empirical evaluation on the effectiveness of genetic algorithm and self-adaptation genetic algorithm for searching the most effective strategies in the heuristic decision making framework are discussed in this paper. © 2009 IEEE.

This paper discussed the utilization of a multiobjective approach for evolving artificial neural networks (ANNs) that act as a controller for radio frequency (RF)-localization behavior of a virtual Khepera robot simulated in a 3D, physics-based environment. The Pareto-frontier Differential Evolution (PDE) algorithm is used to generate the Pareto optimal sets of ANNs that optimize the conflicting objectives of maximizing the virtual Khepera robot's behavior for homing towards a RF signal source and minimizing the number of hidden neurons used in its feed-forward ANNs controller. A fitness function used for mobile robot RFlocalization behavior is proposed. The experimentation results showed the virtual Khepera robot was able to navigate towards signal source with using only a small number of hidden neurons. Furthermore, the Pareto-frontier solutions have been utilized for robustness testing purposes in the environment differs as that used during evolution. The results showed the PDE-EMO algorithm can be practically used in generating the required robot controllers for RF-localization behavior. © 2009 IEEE.

This chapter introduces two algorithms for multiobjective optimization. These algorithms are based on a state-of-the-art Multiobjective Evolutionary Algorithm (MOEA) called Strength Pareto Evolutionary Algorithm 2 (SPEA2). The first proposed algorithm implements a competitive coevolution technique within SPEA2. In contrast, the second algorithm introduces a cooperative coevolution technique to SPEA2. Both novel coevolutionary approaches are then compared to the original SPEA2 in seven scalable DTLZ test problems with 3 to 5 objectives. Overall, the optimization results show that the two proposed approaches are superior to the original SPEA2 with regard to the average distance of the nondominated solutions to the true Pareto front, the diversity of the obtained solutions and also the coverage level. In addition, t-tests have been conducted to validate the significance of the improvements obtained by the augmented algorithms over the original SPEA2. Finally, cooperative coevolution is found to be better than competitive coevolution in terms of enhancing the performance of the original SPEA2. © 2009 Springer-Verlag Berlin Heidelberg.

The proliferation of online auctions has caused the increasing need to monitor and track multiple bids in multiple auctions. As a solution to the problem, an autonomous agent was developed to work in a flexible and configurable heuristic decision making framework that can tackle the problem of bidding across multiple auctions that apply different protocols (English, Vickrey and Dutch). Due to the dynamic and unpredictable nature of online auctions, the agent utilizes genetic algorithm to search for effective solution. Instead of using the conventional genetic algorithm, this paper investigates the application of deterministic dynamic adaptation genetic algorithm and self adaptive genetic algorithm to search for the most effective strategies (offline). An empirical evaluation on the comparison between the effectiveness of self-adaptive genetic algorithm and deterministic dynamic adaptation genetic algorithm for searching the most effective strategies in the online auction environment are discussed in this paper. ©2009 IEEE.

This article describes a simulation model in which a multi-objective approach is utilized for evolving an artificial neural networks (ANNs) controller for an autonomous mobile robot. A mobile robot is simulated in a 3D, physics-based environment for the RF-localization behavior. The elitist Pareto-frontier Differential Evolution (PDE) algorithm is used to generate the Pareto optimal set of ANNs that could optimize two objectives in a single run; (1) maximize the mobile robot homing behavior whilst (2) minimize the hidden neurons involved in the feed-forward ANN. The generated controllers are evaluated on its performances based on Pareto analysis. Furthermore, the generated controllers are tested with four different environments particularly for robustness assessment. The testing environments are different from the environment in which evolution was conducted. Interestingly however, the testing results showed some of the mobile robots are still robust to the testing environments. The controllers allowed the robots to home in towards the signal source with different movements’ behaviors. This study has thus revealed that the PDE-EMO algorithm can be practically used to automatically generate robust controllers for RFlocalization behavior in autonomous mobile robots.

The utilization of a multi-objective approach for evolving artificial neural networks that act as the controllers for phototaxis and radio frequency (RF) localization behaviors of a virtual Khepera robot simulated in a 3D, physics-based environment is discussed in this chapter. It explains the comparison performances among the elitism without archive and elitism with archive used in the evolutionary multi-objective optimization (EMO) algorithm in an evolutionary robotics study. Furthermore, the controllers’ moving performances, tracking ability and robustness also have been demonstrated and tested with four different levels of environments. The experimentation results showed the controllers allowed the robots to navigate successfully, hence demonstrating the EMO algorithm can be practically used to automatically generate controllers for phototaxis and RF-localization behaviors, respectively. Understanding the underlying assumptions and theoretical constructs through the utilization of EMO will allow the robotics researchers to better design autonomous robot controllers that require minimal levels of human-designed elements.