Evolutionary Fuzzy Control Using Rule-based Multi-Objective Genetic Algorithms

This paper addresses data-driven multi-objective fuzzy controller (FC) design problems using rule-based multi-objective genetic algorithms (GAs). The objectives considered in the design of FCs include minimization of the number of fuzzy rules and control accuracy between controlled plant outputs and reference outputs. In the evolutionary FC design process, data are assumed to be online generated without off-line collection in advance. To optimize the number of rules, a rule-coded solution representation and a rule-based mutation operation are introduced into a typical multi-objective non-dominated sorting GA (NSGA II). Simulation results on nonlinear plant control problems verify effectiveness the proposed multi-objective FC design approach.