Study On Soft Computing Based Intelligent Controller

Study On Soft Computing Based Intelligent ControllerAbstractIn this thesis, Soft Computing including fuzzy control, neural network (NN), genetic algorithms (GA) and their combination are discussed systematically. Cerebella model articulation controller (CMAC) among NN is mainly studied here. Firstly, the basic principle, characteristics and their shortcomings of fuzzy control, CMAC and genetic algorithm are illuminated. Secondly, according to their invariable characteristics, several combinations of above-mentioned algorithms are proposed, i. e. Fuzzy-GA control algorithm, Fuzzy-CMAC control algorithm, GA-FCMAC control algorithm. Human Simulating Intelligent Control (HSIC ) are introduced as comparison with above approaches. Typical nonlinear system: simple inverted pendulum and double inverted pendulum are simulated. The characteristics and applying range of each algorithm are discussed and compared.The simulation results illustrate that the combination of two or more soft computing based intelligent control algorithms can preserve their merits and counteract their defects. Due to Fuzzy-CMAC has strong learning ability, it is adapted to time-variable plant. As a random searching algorithm, genetic algorithms based controller is adapted to the plant which is ill-defined and very difficult to be controlled, though the learning time is long. The simulation results demonstrate the effectiveness and feasibility of each integrated intelligent control algorithm presented in the thesis.