Fuzzy controllers for semi-active suspension system generated through genetic algorithms

This paper presents two new methods to generate fuzzy controllers through genetic algorithms (GAs). The main difference between them lies in the initial state of the GAs. This means whether or not to use experts' knowledge for the fuzzy controllers. The method without the knowledge focuses on finding appropriate fuzzy control rules. The other focuses on selecting adequate combinations of input variables, which are usually determined a priori by the designer of the controller in the conventional methods. To set it performance index is the only procedure for the designer of the controllers. In both methods, parameters for the membership functions are tuned through the genetic operations. These parameters are encoded into the chromosomes. A new approach called GA with a local improvement mechanism is applied to the authors' methods. The comparisons with other conventional GAs are also discussed. The authors' approach shows superiority when particular parts of the chromosome have special meanings. To verify the effectiveness of the authors' approach, generation of fuzzy controllers for a semi-active suspension system is simulated.