A Learning Fuzzy Controller Using a Neural Network

Abstract

A fuzzy control has a distinguished feature that the control is capable of incorporating expert's control rules using their linguistic descriptions. Several researches have been done on the learning capability of neural networks for their applications of fuzzy control. But, there has been no controller which can automatically identify the rules in the consequences and tune the membership functions in the premises.
This paper presents a fuzzy controller using a neural network. The new controller can automatically identify the control rules and tune the membership functions utilizing expert's control data. Identification capability of the new fuzzy controller is examined using simple numerical data. The results show that the network can identify nonlinear systems to the same extent of precision as conventional fuzzy modeling methods do. And for demonstrating the capability of the new fuzzy controller, we simulate step responses of a control system which incorporates the fuzzy controller and a controlled object of a first-order system with a dead time.