Byrne ( 50222465 ) GA - Optimisation of a Fuzzy Logic Controller

Declaration I hereby declare that the work contained within this report, except where otherwise stated, is entirely my own and has not been previously submitted in whole, or in part, to any other educational institution. Acknowledgements I would like to express my sincere gratitude to the following; University) for her advice, guidance and support throughout the entire duration of this project. Tallaght) for his help in relation to some of the mathematical aspects of Genetic Algorithms and decision theory. insights into fuzzy controller design and the loan of numerous texts! for his advice on conventional control theory. Abstract The Knowledge Base of a Fuzzy Logic Controller (FLC) encapsulates expert knowledge and consists of the Data Base (membership functions) and Rule-Base of the controller. Optimization of both of these Knowledge Base components is critical to the performance of the controller and has traditionally been achieved through a process of trial and error. Such an approach is convenient for FLCs having low numbers of input variables (e.g. 2-3), however for greater numbers of inputs, more formal methods of Knowledge Base optimization are required. Genetic Algorithms (GAs) provide such a method. They are stochastic, but directed, numerical search methods which use operators consistent with evolutionary theory and find application in many disciplines, particularly function optimization problems. Although not learning algorithms in the strictest sense, GAs can be applied to learning tasks. In this study, the tripartite, Simple Genetic Algorithm (SGA) was evaluated as a feasible solution to the offline optimization/tuning of a simulated Fuzzy Logic Controller (FLC). The efficacy of this approach will be tested by comparison of the GA-FLC's performance in controlling a position-control system, to that of a conventional controller, as well as a heuristically-tuned, FLC.