Prescribed-Time ESO-Based Robust Repetitive Learning Control for Cable-Driven Rehabilitation Exoskeleton Robots

This article proposes a prescribed-time extended state observer (PTESO) based robust repetitive learning control (RLC) scheme for cable-driven rehabilitation exoskeleton robots (CDRER). Based on the repetitive characteristics of rehabilitation training, the uncertain dynamics in CDRER are divided into two separated parts, i.e., periodic part and nonperiodic part. The periodic part is involved in an unknown bounded desired input, and a fully saturated repetitive learning law with a continuous function is proposed, such that the estimation of the desired input is strictly confined within a predefined region and its continuity is guaranteed. Then, by constructing a bounded and continuous gain function in the ESO design, the observation error of the nonperiodic part converges into a small adjustable domain at a preset time and maintains it in the later time without switching, such that the demand of rapid disturbance rejection in the exoskeleton is satisfied. The proposed control scheme enables the CDRER to provide satisfactory steady-state tracking performance as specified by the therapist during repetitive rehabilitation tasks, and comparative experiments on a cable-driven knee exoskeleton validate its effectiveness.