Genetic fuzzy logic controllers

Abstract

The conventional fuzzy logic controller (CFLC) is limited in application, because its logic rules and membership functions have to be preset with expert knowledge. To avoid such drawbacks, a genetic fuzzy logic controller (GFLC) is proposed by employing an iterative evolution algorithm to promote the learning performance of logic rules and the tuning effectiveness of membership functions from examples In sequence. In addition, an encoding method is developed to overcome the difficulties in dealing with numerous constraints while employing genetic algorithms in tuning membership functions. A case of GM car-following behaviors is experimented to verify the applicability and robustness of GFLC. The results demonstrate that GFLC can predict the car-following behaviors precisely. Due to the similarity between fuzzy neural networks (FNN) and GFLC, a comparison is also made and the results indicate that GFLC performs superior to FNN.