MLP neural network-based recursive sliding mode dynamic surface control for trajectory tracking of fully actuated surface vessel subject to unknown dynamics and input saturation

• Published in: Neurocomputing (Amsterdam) Vol. 377; pp. 103 - 112
• Main Authors: Shen, Zhipeng, Bi, Yannan, Wang, Yu, Guo, Chen
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 15.02.2020
• Subjects: Input saturation; Recursive sliding mode control; The minimal learning parameter; The Nussbaum function; Trajectory tracking; The Nussbaum function; The minimal learning parameter; Recursive sliding mode control; Input saturation; Trajectory tracking
• ISSN: 0925-2312; 1872-8286
• DOI: 10.1016/j.neucom.2019.08.090

In this work, we present a MLP neural network-based recursive sliding mode dynamic surface control scheme for a fully actuated surface vessel with uncertain dynamics and external disturbances, where the control input are required to be constrained. First of all, the minimum learning parameter (MLP) neural networks (NN)-based are designed to enhance the robustness against model uncertainties. Subsequently, an adaptive law is employed to compensate neural networks approximation errors and disturbances. The recursive sliding mode control method combined with dynamic surface control (DSC) is designed to eliminate repeated derivative of virtual control laws and enhance systems robustness. A smooth hyperbolic tangent function is incorporated with the control scheme to reduce the risk of actuator saturation. At the same time, the Nussbaum function is used to compensate for the saturation function and ensure the stability of system. We show that under the proposed control method, despite the presence of system uncertainties and disturbances, the tracking errors can converge into arbitrarily small neighborhoods around zero, while the constraint requirements on the control force and torque will not be violated. By using the Lyapunov function, it is proven that the proposed control method can guarantee the uniform boundedness of all the closed loop signals. Finally, simulation results further demonstrate the effectiveness of the proposed method.