Summary Multi-Agent systems typically utilise simple, predictable agents. The usage of such agents in large systems allows for complexity to be achieved through the interaction of these agents. It is feasible, however, to utilise intelligent agents in smaller systems, allowing for more agent complexity and hence a higher degree of realism in the multi-agent model. By utilising the TD( ) Algorithm to train feedforward neural networks, intelligent agents were successfully trained within the reinforcement learning paradigm. A methodology for stabilising this typically unstable neural network training was found through first looking at the relatively simple problem of Tic-Tac-Toe. Once a stable training methodology was arrived at, the more complex task of tackling a multi-player, multi-stage card-game was tackled. The results illustrated that a variety of scenarios can be realistically investigated through the multi-agent model, allowing for solving of situations and better understanding of the game itself. Yet more startling, owing to the agent’s design, the agents learned on their own to bluff, giving much greater insight into the nature of bluffing in such games that lend themselves to the act.

This thesis concerns the design, construction, control, and testing of a novel self-contained soft robotic vehicle; the JenniFish is a free-swimming jellyfish-like soft robot that could be adapted for a variety of uses, including: low frequency, low power sensing applications; swarm robotics; a STEM classroom learning resource; etc. The final vehicle design contains eight PneuNet-type actuators radially situated around a 3D printed electronics canister. These propel the vehicle when inflated with water from its surroundings by impeller pumps; since the actuators are connected in two neighboring groups of four, the JenniFish has bi-directional movement capabilities. Imbedded resistive flex sensors provide actuator position to the vehicle’s PD controller. Other onboard sensors include an IMU and an external temperature sensor. Quantitative constrained load cell tests, both in-line and bending, as well as qualitative free-swimming video tests were conducted to find baseline vehicle performance capabilities. Collected metrics compare well with existing robotic jellyfish. / Includes bibliography. / Thesis (M.S.)--Florida Atlantic University, 2016. / FAU Electronic Theses and Dissertations Collection

No description available.

HVAC (Heating Ventilating and Air-Conditioning) system is multivariate, nonlinear, and shares time-varying characteristics. It poses challenges for both system modeling and performance optimization. Traditional modeling approaches based on mathematical equations limit the nature of the optimization models and solution approaches. Computational intelligence is an emerging area of study which provides powerful tools for modeling and analyzing complex systems. Computational intelligence is concerned with discovery of structures in data and recognition of patterns. It encompasses techniques such as neural networks, fuzzy logic, and so on. These techniques derive rules, patterns, and develop complex mappings from the data. The recent advances in information technology have enabled collection of large volumes of data. Computational intelligence embraces biology-inspired paradigms like evolutionary computation and particle swarm intelligence in solving complex optimization problems. Successful applications of computational intelligence have been found in business, marketing, medical and manufacturing domains. The focus of this thesis is to apply computational intelligence approach in modeling and optimization of HVAC systems. In this research, four HVAC sub-systems are investigated: the AHU (Air Handling Unit), VAV (Variable Air Volume), ventilation system, and thermal zone. Various computational intelligence approaches are used to identify parameters or problem solving. Energy savings are accomplished by minimizing the cooling output, reheating output or fan running time as well as on-line monitoring. One contribution of the research reported in the thesis is the use of computational intelligence algorithms to establish nonlinear mappings among different parameters. Another major contribution is in using heuristics algorithms to solve multi-objective optimization problems.

Recent years have seen increasingly growing interest in energy conservation. Industrial systems involving large energy consumption are receiving intensive attentions from both academia and industry on optimizing control strategies for potential energy savings. This thesis investigates energy efficiency of two industrial systems, the heating, ventilating and air-conditioning (HVAC) system, and the wastewater pumping system. Both systems are known as dynamic, nonlinear, and multivariate, which are of great challenge for system modeling and performance optimization. Traditional approaches, usually relying on physical equations and mathematical programming, show limited abilities in dealing with complex system modeling and optimization. As an emerging science with an abundance of successful applications in industrial, business, medical areas, data mining has proven its powerful capabilities in nonlinear system modeling and complex pattern recognition. Successful and effective applications of data mining algorithms, such as multilayer perceptron neural network, support vector machine, and boosting tree have been reported in literature and expanded to complex system modeling. Computational intelligence has been an emerging and promising area over these years for its capability of solving difficult optimization problems, for instance, mixed integer nonlinear programing problems. Computational intelligence has been tremendously applied in providing optimal or near-optimal solutions within limited computation time in different kinds of optimization problems. This thesis mainly focuses on employing computational intelligence to generate optimal control strategies in the stated industrial systems. The main contribution of this research lies in utilizing computational intelligence to solve the mixed integer nonlinear programming optimization models built by data mining algorithms. Another strength of this thesis is establishing the unified framework of applying data mining and computational intelligence to real-world system control and optimization.

Symmetry has been a very useful tool to researchers in various scientific fields. At its most basic, symmetry refers to the invariance of an object to some transformation, or set of transformations. Usually one searches for, and uses information concerning an existing symmetry within given data, structure or concept to somehow improve algorithm performance or compress the search space. This thesis examines the effects of imposing or inducing symmetry on a search space. That is, the question being asked is whether only existing symmetries can be useful, or whether changing reference to an intuition-based definition of symmetry over the evaluation function can also be of use. Within the context of optimization, symmetry induction as defined in this thesis will have the effect of equating the evaluation of a set of given objects. Group theory is employed to explore possible symmetrical structures inherent in a search space. Additionally, conditions when the search space can have a symmetry induced on it are examined. The idea of a neighborhood structure then leads to the idea of opposition-based computing which aims to induce a symmetry of the evaluation function. In this context, the search space can be seen as having a symmetry imposed on it. To be useful, it is shown that an opposite map must be defined such that it equates elements of the search space which have a relatively large difference in their respective evaluations. Using this idea a general framework for employing opposition-based ideas is proposed. To show the efficacy of these ideas, the framework is applied to popular computational intelligence algorithms within the areas of Monte Carlo optimization, estimation of distribution and neural network learning. The first example application focuses on simulated annealing, a popular Monte Carlo optimization algorithm. At a given iteration, symmetry is induced on the system by considering opposite neighbors. Using this technique, a temporary symmetry over the neighborhood region is induced. This simple algorithm is benchmarked using common real optimization problems and compared against traditional simulated annealing as well as a randomized version. The results highlight improvements in accuracy, reliability and convergence rate. An application to image thresholding further confirms the results. Another example application, population-based incremental learning, is rooted in estimation of distribution algorithms. A major problem with these techniques is a rapid loss of diversity within the samples after a relatively low number of iterations. The opposite sample is introduced as a remedy to this problem. After proving an increased diversity, a new probability update procedure is designed. This opposition-based version of the algorithm is benchmarked using common binary optimization problems which have characteristics of deceptivity and attractive basins characteristic of difficult real world problems. Experiments reveal improvements in diversity, accuracy, reliability and convergence rate over the traditional approach. Ten instances of the traveling salesman problem and six image thresholding problems are used to further highlight the improvements. Finally, gradient-based learning for feedforward neural networks is improved using opposition-based ideas. The opposite transfer function is presented as a simple adaptive neuron which easily allows for efficiently jumping in weight space. It is shown that each possible opposite network represents a unique input-output mapping, each having an associated effect on the numerical conditioning of the network. Experiments confirm the potential of opposite networks during pre- and early training stages. A heuristic for efficiently selecting one opposite network per epoch is presented. Benchmarking focuses on common classification problems and reveals improvements in accuracy, reliability, convergence rate and generalization ability over common backpropagation variants. To further show the potential, the heuristic is applied to resilient propagation where similar improvements are also found.

Symmetry has been a very useful tool to researchers in various scientific fields. At its most basic, symmetry refers to the invariance of an object to some transformation, or set of transformations. Usually one searches for, and uses information concerning an existing symmetry within given data, structure or concept to somehow improve algorithm performance or compress the search space. This thesis examines the effects of imposing or inducing symmetry on a search space. That is, the question being asked is whether only existing symmetries can be useful, or whether changing reference to an intuition-based definition of symmetry over the evaluation function can also be of use. Within the context of optimization, symmetry induction as defined in this thesis will have the effect of equating the evaluation of a set of given objects. Group theory is employed to explore possible symmetrical structures inherent in a search space. Additionally, conditions when the search space can have a symmetry induced on it are examined. The idea of a neighborhood structure then leads to the idea of opposition-based computing which aims to induce a symmetry of the evaluation function. In this context, the search space can be seen as having a symmetry imposed on it. To be useful, it is shown that an opposite map must be defined such that it equates elements of the search space which have a relatively large difference in their respective evaluations. Using this idea a general framework for employing opposition-based ideas is proposed. To show the efficacy of these ideas, the framework is applied to popular computational intelligence algorithms within the areas of Monte Carlo optimization, estimation of distribution and neural network learning. The first example application focuses on simulated annealing, a popular Monte Carlo optimization algorithm. At a given iteration, symmetry is induced on the system by considering opposite neighbors. Using this technique, a temporary symmetry over the neighborhood region is induced. This simple algorithm is benchmarked using common real optimization problems and compared against traditional simulated annealing as well as a randomized version. The results highlight improvements in accuracy, reliability and convergence rate. An application to image thresholding further confirms the results. Another example application, population-based incremental learning, is rooted in estimation of distribution algorithms. A major problem with these techniques is a rapid loss of diversity within the samples after a relatively low number of iterations. The opposite sample is introduced as a remedy to this problem. After proving an increased diversity, a new probability update procedure is designed. This opposition-based version of the algorithm is benchmarked using common binary optimization problems which have characteristics of deceptivity and attractive basins characteristic of difficult real world problems. Experiments reveal improvements in diversity, accuracy, reliability and convergence rate over the traditional approach. Ten instances of the traveling salesman problem and six image thresholding problems are used to further highlight the improvements. Finally, gradient-based learning for feedforward neural networks is improved using opposition-based ideas. The opposite transfer function is presented as a simple adaptive neuron which easily allows for efficiently jumping in weight space. It is shown that each possible opposite network represents a unique input-output mapping, each having an associated effect on the numerical conditioning of the network. Experiments confirm the potential of opposite networks during pre- and early training stages. A heuristic for efficiently selecting one opposite network per epoch is presented. Benchmarking focuses on common classification problems and reveals improvements in accuracy, reliability, convergence rate and generalization ability over common backpropagation variants. To further show the potential, the heuristic is applied to resilient propagation where similar improvements are also found.

In this thesis a Simulator of the Immune System (IS) is developed. The implemented models of the IS refines and extends models of existing Artificial IS. / In der Arbeit wird der Prototyp eines Immunsystem-Simulators erstellt, der die vorhandenen Modelle Künstlicher Immunsystem verfeinert und die Untersuchung der Einflüsse verschiedener Parameter erlaubt.

Στην διατριβή αυτή εξετάζεται η αποδοτικότητα αλγορίθμων υπολογιστικής νοημοσύνης σε προβλήματα αριθμητικής βελτιστοποίησης, αναπτύσσονται τροποποιήσεις και βελτιώσεις των μεθόδων και εισάγεται ένα νέο σχήμα της μεθόδου Βελτιστοποίησης με Σμήνος Σωματιδίων, το οποίο ενοποιεί διαφορετικές εκδόσεις της συνδυάζοντας τα χαρακτηριστικά τους, μαζί με την θεωρητική ανάλυσή του. / The main goal of this thesis was the investigation of the performance of computational intelligence algorithms on numerical optimization problems, the development of modifications and improvements of the algorithms, as well as the development of a new scheme of the Particle Swarm Optimization algorithm that harnesses its main variants, along with its theoretical analysis.

The ill-effects of stress due to fatigue significantly impact the welfare of individuals and consequently impact overall corporate productivity. This study introduces a simplified model of stress in the workplace using agent-based simulation. This study represents a novel contribution to the field of evolutionary computation. Agents are encoded initially using a String Representation and later expanded to multi-state Binary Decision Automata to choose between work on a base task, special project or rest. Training occurs by agents inaccurately mimicking behaviour of highly productive mentors. Stress is accumulated through working long hours thereby decreasing productivity performance of an agent. Lowest productivity agents are fired or retrained. The String representation for agents demonstrated near average performance attributed to the normally distributed tasks assigned to the string. The BDA representation was found to be highly adaptive, responding robustly to parameter changes. By reducing the number of simplifications for the model, a more accurate representation of the real world can be achieved.