A new approach to adaptive membership function for fuzzy inference system

Abstract

A novel adaptive neuro-fuzzy control (ANFC) system using neural network based fuzzy reasoning is proposed to make a fuzzy logic control system more adaptive and more effective. In most cases, the design of a fuzzy inference system relies on the method in which an expert or a skilled human operator works in that special domain. However, if he has no expert knowledge in any nonlinear environment, it is difficult to control in order to optimize. Thus, the proposed adaptive structure for the fuzzy reasoning system can be controlled more adaptively and more effectively in a nonlinear environment for changing input membership functions and output membership functions. ANFC can be adapted to a proper membership function for nonlinear plants, based on a minimum number of rules and an initial approximate membership function. A rotary inverted pendulum system is simulated to demonstrate the efficiency of the proposed ANFC.