On-Line Evolutionary Optimization of Fuzzy Control System based on Decentralized Population

Abstract

This paper presents a real time evolutionary optimization method of the fuzzy control system by using the decentralized population technique. The presented method generates a new population for each rule of the fuzzy control system in a decentralized manner and updates each of them on-line during operation by using the normalized accelerated evolutionary programming technique. For each sampling rime, only the rules associated with the current states are updated, and thus, all of the fuzzy rules independently evolve to their optimal ones. As a result, the overall fuzzy control system evolves toward the suboptimal fuzzy control system on-line with significantly reduced evolution time. The developed optimization technique has: been applied to the mobile robot navigation problem to show its capabilities. Simulation and experimental results show the feasibility and the optimization capability of the method.