Neural‐Network‐Based Adaptive Sliding Mode Control for Robotic Prescribed‐Instant Trajectory Tracking

Abstract

This article develops an innovative prescribed‐instant sliding mode control strategy, specifically designed to ensure robust trajectory tracking performance for robotic manipulators under model uncertainties and actuator saturation limitations. The proposed control architecture guarantees precise trajectory tracking while enforcing prescribed‐instant convergence to an arbitrary time‐varying reference profile at a user‐specified instant. Neural networks are utilized to approximate the unknown parameters of the overall system model, eliminating the requirement for precise dynamic model parameters in the controller design. The tracking errors are then leveraged to construct an auxiliary system, which is used to implement a new prescribed‐instant sliding mode controller. The proposed control scheme augments robustness and accelerates convergence, while mitigating the impact of actuator saturation in practical implementations. Furthermore, it provides the flexibility to prescribe the convergence time a priori, thereby ensuring stringent trajectory‐tracking accuracy. Its prescribed‐instant convergence is theoretically guaranteed via Lyapunov analysis, and the practical effectiveness of the method is substantiated through experimental studies.