Modeling and experiments in fuzzy control

Abstract

Fuzzy logic controllers can be programmed with minimal knowledge about the system dynamics and are capable of tuning their structure starting with simple rules. This capability was experimentally demonstrated for DC motor controlled mechanical systems using two load cases. In each case, the controller was initialized with rudimentary rules and experimentally tuned online using gradient descent techniques. Such control strategies provide a convenient framework to incorporate human experience and expert knowledge. A self-organizing tuning algorithm is proposed for fuzzy controllers to compensate for imprecision in modelling and for system nonlinearities.<>