Optimal fuzzy control: Design of fuzzy controllers through knowledge discovery and optimization

Mark William Palmer, Marquette University

Abstract

This dissertation introduces a new method to the study of optimal fuzzy control. This new fuzzy controller optimization method combines several optimization methods with concepts from optimal control and applies them to the problem of minimizing a cost function describing the operation of a fuzzy control system. In addition to improving fuzzy controller performance, the method has the benefit of adding to and improving upon the controller's rule-based knowledge. The new method further improves upon previous optimal fuzzy control methods by allowing for a greater breadth of fuzzy controller designs and by yielding designs that are effective across a range of initial system conditions. Brief overviews of the related fields of fuzzy control, optimal control, optimization, and the previous work in optimal fuzzy control are included. The fuzzy controller optimization method is then presented in detail, followed by demonstrations of the application of the method to control problems. Evaluations of the method for control effectiveness and computational complexity are also included.

This paper has been withdrawn.