Design of a proportional plus derivative neuro fuzzy controller

The transformation of an expert's knowledge to control rules in a fuzzy logic controller has not been formalized and arbitrary choices concerning, for example, the shape of membership functions have to be made. The quality of a fuzzy controller can be drastically affected by the choice of membership functions. Thus, methods for tuning fuzzy logic controllers are needed. Neural networks and fuzzy logic are combined to solve the problem of tuning fuzzy logic controllers. The neuro fuzzy controller uses the neural network learning techniques to tune the membership functions while keeping the semantics of the fuzzy logic controller intact. Both the architecture and the learning algorithm are presented for a general neuro fuzzy controller. From this, procedures to design proportional and proportional plus derivative neuro fuzzy controllers are obtained. A step by step algorithm for offline training is presented.