Hierarchical Optimal Switching Tracking Control for Rehabilitative Walker Considering Human-Robot Interaction Environments

This study investigates a hierarchical optimal switching tracking control method intended to facilitate cyclic switching training in a rehabilitative walker. Reinforcement learning (RL) is utilized to optimize the performance of the human-robot system during the switching process, and it is designed on the basis of the switched stochastic configuration networks (SSCN) approximations under switchable observer-actor-evaluator (SOAE) online learning framework. The switching tracking control is designed hierarchically through the optimal backstepping technique to compensate for the influence of dynamically switching motion environments on tracking performance, thereby achieving simultaneous trajectory and velocity tracking. A redundant input switching model, accounting for human-robot interaction environments, is established by decomposing the coefficient matrices that affect the patient's position and posture. In addition, a time-based cyclic switching mode is developed, in which different training tasks are alternated at time intervals prescribed by the physiatrist. Finally, simulation analysis and experimental results demonstrate that the proposed method can flexibly and effectively support rehabilitation training.