A method for design of a hybrid neuro-fuzzy control system based on behaviour classification

This paper presents a new method for design of a neuro-fuzzy control system based on behaviour classification. Its main idea is to describe characters of the system by defining behaviours, such as “oscillation”, “overdamping”, “underdamping” and so forth, because control signals from a fuzzy logic controller are determined by responses of the system rather than its analytical models and the fuzzy controller could yield same control results for a set of systems with similar response behaviours. For these response behaviours, therefore, we find a set of equivalent simplified systems for optimization of a neuro-fuzzy control system. The advantage of this method is that we do not need to care about convergence problem during adaptation process when this method is used to design a neuro-fuzzy controller Finally, we use the proposed method to design neuro-fuzzy controllers for a two-link manipulator