Wearable exoskeleton robot control using radial basis function-based fixed-time terminal sliding mode with prescribed performance

Abstract

This paper tackles the problem of robust and accurate fixed-time tracking in human–robot interaction and deals with uncertainties. This work introduces a control approach for a wearable exoskeleton designed specifically for rehabilitation tasks. The approach combines prescribed performance control-based fixed-time terminal sliding mode with a neural network. Its main objectives are to achieve trajectory tracking, reduce chattering, ensure fixed-time stability, and maintain robustness against uncertainties. The controller includes a radial basis function neural network to estimate unknown dynamics and incorporates prescribed performance criteria. This enables precise joint space trajectory tracking, even in the presence of uncertain dynamics and disturbances. The prescribed performance ensures that the trajectory tracking error evolves within prescribed limits. The combined neural network and fixed-time terminal sliding mode technique are proposed to ensure robustness and fixed-time convergence. The closed-loop stability is analyzed using the Lyapunov theory, and a new fixed-time convergence is provided. Numerical simulations demonstrate a reduction tracking compared to another advanced SMC technique, while experimental results on a 7-DoF ETS-MARSE exoskeleton show better tracking with control torques free of chattering compared with two advanced SMC techniques.