Modified Learning of T-S Fuzzy Neural Network Control for Autonomous Underwater Vehicles

Abstract

In this paper, an improved Takagi-Sugeno (T-S) Fuzzy Neural Network (FNN) based on modified learning is proposed for the motion control of Autonomous Underwater Vehicles (AUV). Aiming to improve the control precision and adaptability of T-S fuzzy model, a fuzzy objective is used to update the fuzzy rules and the proportion factor on-line. A modified learning of network is developed by back-propagating the error between the actual response and the desired output of the vehicle, which allows us to train the network exactly on the operational range of the plant, and consequently effectively compensates the slow convergence of BP algorithm. Finally, simulations on the ldquoMini-AUVrdquo show that the control scheme can greatly speed up the response of the vehicle with pretty stability, which makes it possible to implement the real-time control for AUV with FNN.