Study Of Fuzzy Neural Network Control Based On Reinforcement Learning And Its Application

The structure identification of Fuzzy Neural Network is studied in this paper by using reinforcement learning. According to the strength of each rule, reinforcement learning determines the selection of the rules. The simulation results show that this method has better adaptability compared with the fuzzy K-means clustering algorithm and the self-organization competition neural network. In addition, two approaches are proposed to improve the quality of the regulations: while establishing a new rule, the trial function of reinforcement learning is applied to search the rule of fuzzy control in order to improve the quality of the regulation; while deleting the rule with lowest value function, the stability of the system is enhanced by gradually decreasing the width of the membership function firstly. The simulation results certificate the validity of the above approaches.