Fuzzy-Neural Network Controllers using the Mean-Value-based Functional Reasoning

Abstract

Most of conventional fuzzy controllers use the input-data-based functional reasoning and the simplified reasoning, so that they can not rationally design any parameter in the conclusion of the fuzzy reasoning by some ways. Therefore, it is not necessarily effective for shortening the computational time and reducing the number of learning parameters to realize such fuzzy reasonings as a fuzzy-neural network.In this paper, a fuzzy-neural network controller is proposed by applying a mean-value-based functional reasoning, in which the mean values are used in constructing a conclusion function when a Gaussian function is applied for the antecedent membership function. Since the initial values of any parameter except for the mean-values in the conclusion can be rationally designed as the parameters, which are independent of the control rules and represent a stable switching plane (or line) in the VSS control, the learning time and the number of learning parameters in the conclusion are shown to be reduced drastically, compared with those due to the input-data-based founctional reasoning and the simplified reasoning, in which the initial parameters in the conclusion depend on the control rules. The effectiveness of the proposed method is illustrated by computer simulations for the tracking control problem of a mobile robot with two independent drive wheels.