A Practical Prescribed-Time Control Strategy for Knee Rehabilitation Exoskeletons With a Series Elastic Actuator

Abstract In recent years, advances in sports medicine have significantly improved rehabilitation strategies for exercise-induced injuries. Among them, robot-assisted rehabilitation systems have emerged as an effective approach for knee joint recovery due to their precise and controllable training capabilities. Knee exoskeletons equipped with series elastic actuators (SEAs) improve the safety of human-robot interaction and reduce the risk of joint injury by using compliant elements to absorb unexpected external impacts. However, the integration of SEAs introduces several control challenges, including modeling uncertainties, friction, and external disturbances, which degrade model accuracy and control performance. To cope with these unknown nonlinearities, this paper employs a radial basis function neural network for real-time approximation. In addition, a prescribed-time Lyapunov-based stability criterion is incorporated to guarantee system convergence within a prescribed time. To reduce redundant data transmission and communication burden caused by frequent control updates, a dynamic event-triggered mechanism (DETM) is developed, significantly lowering the control update frequency. Rigorous Lyapunov-based analysis confirms that all signals in the closed-loop system remain bounded and achieve uniform convergence within the prescribed time. Simulation results further demonstrate the effectiveness of the proposed control scheme.