An adaptive hybrid fuzzy-neural controller

This paper presents investigations toward adaptive fuzzy control using neural networks for the control of nonlinear and time variable physical processes. By using a hybrid structure, the fuzzy logic controller can be composed of its traditional components and two neural networks. The first neural network works as an adaptive rule estimator for online acquisition and modification of the fuzzy rules as it was presented in a previous work. Here a second neural network is proposed as a further development of the hybrid structure which implements the defuzzification function and works as an adaptive neural defuzzifier. The standard one step gradient descent learning procedure is implemented and the output from a conventional feedback controller is used for tuning the parameters of the neural defuzzifier. Comparison with the traditional fuzzy technique is discussed. The results obtained by simulation show the better performance of the proposed hybrid control structure.