Adptive Heading Control of Underactuated Unmanned Surface Vehicle Based on Improved Backpropagation Neural Network

Abstract Aiming at the challenges to the accurate and stable heading control of underactuated unmanned surface vehicles arising from the nonlinear interference caused by the overlay and the interaction of multi interference, and also the uncertainties of model parameters, a heading control algorithm for an underactuated unmanned surface vehicle based on an improved backpropagation neural network is proposed. Based on applying optimization theory to realize that the underactuated unmanned surface vehicle tracks the desired yaw angle and maintains it, the improved momentum of weight is combined with an improved tracking differentiator to improve the robustness of the system and the dynamic property of the control. A hyperbolic tangent function is used to establish the nonlinear mappings an approximate method is adopted to summarize the general mathematical expressions, and the gradient descent method is applied to ensure the convergence. The simulation results show that the proposed algorithm has the advantages of strong robustness, strong anti-interference and high control accuracy. Compared with two commonly used heading control algorithms, the accuracy of the heading control in the complex environment of the proposed algorithm is improved by more than 50%.