APPROXIMATION-BASED NEURAL NETWORK AND FUZZY LOGIC CONTROL

摘要

The universal approximation properties of neural networks and fuzzy logic networks are exploited in designing nonlinear adaptive controllers for systems with unknown dynamics. These controllers are model-free in the sense that no detailed knowledge of the underlying dynamics is needed. The controller parameters are tuned on-line and guarantee overall closed-loop performance of the system. The proposed method circumvents the standard assumptions like linearity-in-the-pararneters that limit the utility of conventional adaptive control techniques. In this framework, the equivalence of neural networks and fuzzy logic networks from a closed-loop control perspective is established.