An Extended Evolutionary Approach Toward Optimal Design of Fuzzy Systems

摘要

This paper proposes an extended evolutionary approach in which genetic algorithms and hierarchical modules are used to automatically design fuzzy systems. The basic idea is to define fuzzy system parameters and operators, organized in hierarchical modules which interact in a evolutionary process to produce, at the final generation, a fuzzy system behavior close to the optimal desired one. In this work, besides the automatic definition of some important parameters, like number of fuzzy rules and shape of membership functions, the inference mechanism may also be obtained through evolution. Some intrinsic characteristics of fuzzy systems concerning function approximation are analyzed and simulation examples are carried out to attest the efficiency of the automatic design process.