Optimizing non-assisted body part movements for robot-assisted therapy

Learning to control muscles in movement is essential in rehabilitation. The use of biofeedback and robotics to induce muscle activity has been investigated in recent years. The human musculoskeletal system has complex inter-limb interactions, which have been simplified in previous studies by immobilizing non-assisted body parts. This study proposes a framework to induce the desired muscle activity pattern and provide visual feedback to the user by optimizing the reference trajectory of the non-assisted body part in robotic rehabilitation. In the proposed framework, an individual model learning and trajectory optimization method was utilized to consider the constraints of the rehabilitation time slot. Its performance was verified through experiments on 12 healthy subjects. The results show improved effectiveness and feasibility, achieved by reducing the discrepancies between targeted and induced muscle activations, compared to the baseline, which did not optimize non-assisted body part movements. • Aimed to induce target muscle activation patterns in robot-assisted therapy. • Proposed a novel framework to optimize non-assisted body part movements • Developed user-specific model learning and posture trajectory optimization. • Verified the tracking accuracy of muscle activity patterns on 12 subjects. • Demonstrated a improvement in tracking performance compared to the baseline.