Optimization of Fuzzy Logic Controller for Trajectory Tracking Using Genetic Algorithm

Most Fuzzy Logic Controllers (FLCs) to date are working based on expert knowledge derived from heuristic knowledge of experienced operators. Conventional fuzzy logic controllers have poor adaptability due to invariable Membership Function (MF) parameters and fixed rule set. Conventional manual coded FLCs use only expert knowledge bases and do poorly with complex problems, especially with large numbers of input variables. We have developed FLCs using a Genetic Algorithm (GA) to automatically acquire knowledge that we call a genetic-fuzzy in which the GA is used to adaptively generate fuzzy rules and simultaneously selecting an appropriate MF shape. We also evaluate different membership functions in the fuzzy logic control. FLC sensibility is analysed and compared for different membership functions. We compare our proposed genetic-fuzzy approach to such existing methods, including as a manually coded conventional method, conventional method with complementary membership function, and a neuro-fuzzy method on a widely used test bed; backing up a truck reversal problem. Simulation results have shown our proposal to be superior to existing widely used methods.