Synthesis of fuzzy controllers from linguistic specifications using an evolution strategy

Abstract

The lack of a general methodology for tuning the parameters of a fuzzy controller is a major drawback to a more widespread usage of this control technique. In this paper the authors propose an approach based on an evolution strategy. The major advantage of the approach is twofold. Firstly, the possibility to specify the desired controller performance in linguistic terms. This entirely agrees with the general idea of fuzzy control being a framework to model human control knowledge. Secondly, since we start off from a given non-linear control schema, only three parameters have to be learned.<>