Adaptive fuzzy neural network-based smooth-switching gain dynamic surface control for constrained manipulator with error-based nonlinear disturbance observer

This paper studies an adaptive fuzzy neural network-based smooth-switching gain dynamic surface control with an error-based nonlinear disturbance observer for the manipulator subject to the full-state constraints and input saturation. First, the smooth-switching gain strategy is proposed to replace the conventional static control gain. This novel strategy can dynamically update the control gains according to the tracking error and its derivative, thereby optimizing tracking performance across different response phases. Second, considering that the conventional nonlinear disturbance observer generates harmful observation peaks when there is a large tracking error, an error-based nonlinear disturbance observer is designed. This observer adaptively adjust the observer gain based on the tracking error, effectively mitigating harmful observation peaks. Furthermore, the adaptive fuzzy neural network strategy is introduced to approximate the modeling uncertainties. Finally, an auxiliary system and an asymmetric time-varying barrier Lyapunov function are established to handle the input saturation and the asymmetric time-varying full-state constraints, respectively. The comparative experiment of a two-degree-of-freedom manipulator validate that the proposed strategy can effectively constrain the state and input of the system within predefined limits. The experimental results further validate that the proposed strategy can effectively optimise the tracking performance and increase the disturbance rejection performance of the system.