Fuzzy model identification by means of multiobjective genetic programming

摘要

Fuzzy model identification of high-dimension, non-linear systems is a complex, non-trivial task. Current techniques for designing Takagi-Sugeno (TS) fuzzy systems deal with the choice of input-output variables and the number and shape of membership functions whilst assuming that predefined linear structures will locally represent the system under investigation. A method for the derivation of optimal fuzzy system structures based upon non-linear local system representations is proposed. This technique also aims to reduce model complexity. In order to tackle the issue of structural optimality of the fuzzy system, a mutliobjective genetic programming approach is used in this study. Copyright ~direct 2000 IFAC