Neural Network-Based Fixed-Time Sliding Mode Control for Robotic Systems With Disturbances
Abstract
This article addresses the trajectory tracking control problem of robotic systems under parameter uncertainties and external disturbances. First, the fixed-time stability of a second-order system is established, and based on this, an integral sliding variable is introduced to ensure fixed-time stability while avoiding singularity issue. To enhance tracking performance, an adaptive neural network is designed to estimate the uncertainties and external disturbances. Consequently, a neural network-based nonsingular fixed-time sliding mode control (FTSMC) is proposed. The fixed-time stability of the system under the proposed control scheme is proved based on homogeneity and Lyapunov stability theory. Finally, the effectiveness and superiority of the proposed method are verified by numerical simulation and comparative experiments on Franka Emika Panda robotic platform. Note to Practitioners—This work is motivated by the high-precision and high-response requirements for robot trajectory tracking control in engineering practice. We propose a novel NN-based FTSMC. To avoid the singularity, the traditional FTSMC can only achieve that the state converges to the neighborhood of the origin. Different from this, the proposed scheme not only guarantees the fixed-time stability by designing an integral sliding mode, but also guarantees that the states converge to the origin, which means it has higher accuracy. Moreover, leveraging the universal approximation property of neural networks, this method designs an adaptive neural network observer that—unlike most conventional observers—imposes no constraints on either the disturbances magnitude or its derivative bounds. Consequently, the proposed approach exhibits enhanced robustness against more generalized disturbances, which means it can be applied to more general work scenarios.