Research on RBF Neural Network for Discrete Fuzzy Control

Abstract

On the basis of studying the working principle of discrete fuzzy control, according to the structural characteristic and the ability of RBF neural network approaching to any unknown function, the clustering and nonlinear mapping function of RBF neural network are used to implement discrete fuzzy inference and control. If only a RBF neural network is suitably designed and well trained by a discrete numerical set of the corresponding relation between inputs and outputs induced by fuzzy set theory, it can entirely act as a fuzzy controller. The function of RBF neural network and fuzzy control theory are organically integrated in the paper. Research results show that RBF neural network provides a simple and effective realizing method for fuzzy control, and using fuzzy theory greatly reduces the number of training samples of RBF neural network, the integration of RBF neural network and fuzzy theory is both mutually beneficial and very perfect.