Fuzzy Versus Neuro-fuzzy Identification of Active Front Steering of a Vehicle on Slippery Roads

Abstract

Due to the high rate changes in the handling of cars, the use of an auxiliary identification process to design efficient controllers is of importance. Many identification algorithms have been proposed in the literature, which generally performs well under normal situations, but does not show acceptable performance in uncertain conditions. In this article, due to the nature of the neuro-fuzzy networks in identifying and predicting uncertain conditions, an adaptive neuro-fuzzy identification algorithm is proposed to steer vehicles at the uncertain slippery condition of roads. A set of data for three well-known manoeuvres of vehicle dynamics at conventional conditions was collected to train the algorithm using adaptive neuro-fuzzy inference system of MATLAB. Using back propagation of error as the learning algorithm, the parameters of the algorithm were modified regarding uncertain conditions. Making an analogy, the performance of the proposed identification scheme was compared to the untrained fuzzy one. In regular situations, the results were almost identical, but in uncertain ones such as slippery roads, the performance of the proposed neuro-fuzzy algorithm was much better.