Performance-Based Iterative Learning Control for Task-Oriented Rehabilitation: A Pilot Study in Robot-Assisted Bilateral Training

Active participation from human subjects can enhance the effectiveness of robot-assisted rehabilitation. Developing interactive control strategies for customized assistance is therefore essential for encouraging human-robot engagement. However, existing human-robot interactive control strategies lack precise evaluation indicators with effective convergence method to steadily and rapidly customize appropriate assistance during task-oriented training. This study proposes a performance-based iterative learning control algorithm for robot-assisted training, which aims at providing subject-specific robotic assistance to encourage active participation. Three performance indicators based on a Fugl-Meyer assessment (FMA) regression model are introduced to associate clinical scales with robot-based measures, and a fuzzy logic is employed for comprehensive performance evaluation. To increase efficient training time, a piecewise learning rate-based iterative law is applied to quickly converge to a subject-specific control parameter session by session. The proposed strategy is preliminarily estimated for a case of bilateral upper limb training with an end-effector-based robotic system. The experimental results with human subjects indicate that the proposed strategy can obtain appropriate parameters after only several iterations and adapt to random perturbations (like muscle fatigue).