Learning laws for neural-network implementation of fuzzy control systems

A method of designing adaptive fuzzy control systems using structured neural networks is discussed. The basic idea is to implement a rule-based fuzzy control system with a neural network consisting of two subnetworks of pattern recognition, and fuzzy reasoning and control synthesis. The neural network is arranged such that the structure and operations of the original fuzzy control system can be fully retrieved from its network implementation. Equipped with the learning capability of neural networks, this implementation provides a mechanism to refine the existing rules and generate new rules for fuzzy control. It also opens a way for fuzzy control systems to exploit neural networks for carrying out their inherent parallel computation and suppressing the memory space required by their knowledge bases and inference programs.<
>