Advances in fuzzy adaptive control

Abstract

Fuzzy control has proven effective for complex, nonlinear, imprecisely-defined processes for which standard models and controls are impractical or cannot be derived. Deriving fuzzy control is, however, often difficult and time-consuming. Furthermore, severe problems of high-dimensionality are usually encountered in the implementation of control for systems with multiple inputs and outputs. More efficient and systematic methods for knowledge acquisition and fuzzy controller synthesis are needed. Effective, intelligent automated systems, such as adaptive fuzzy controllers, capable of learning from process data as well as incorporating linguistic data, possess significant advantages that make them attractive candidates for the much needed technology. An adaptive fuzzy controller could automatically generate a set of fuzzy control rules and improve on them as the control process evolves. In this paper, we survey major results on the development of intelligent, adaptive fuzzy control. Particular focus is given to methods which combine the learning capabilities of neural networks with fuzzy logic control, as these appear to be most promising.