A design of CMAC-based fuzzy logic controller with fast learning and accurate approximation

Abstract

This paper proposes a CMAC-based fuzzy logic controller (FLC) with a fast learning capability and an accurate approximation ability. The proposed CMAC-based FLC has the fast learning capability because it pursuits the local generalization and only a small number of activated units in the network are participated in the forward and backward computation. It also produces an accurate input-output approximation ability, because it adjusts the MFs model parameters of the input and output variables simultaneously and it considers both centers and widths of output membership functions to compute a crisp defuzzified value. Application to the truck backer-upper control problem of the proposed CMAC-based FLC is presented. Simulation results validate the fast learning and the accurate approximation of the proposed CMAC-based FLC. (C) 2002 Elsevier Science B.V. All rights reserved.