Development of Adaptive Impedance Control Strategy for an Upper-Limb Cable-Driven Rehabilitation Robot

Abstract

Cable-driven rehabilitation robots (CDRR) offer low inertia and a large workspace, making them well-suited for upper-limb therapy in bedridden patients. Assist-as-Needed (AAN) control promotes neuroplasticity by providing minimal support only when needed, but may not sufficiently challenge users with position-dependent weakness, such as frozen shoulder. To address this, we propose an Assist–Resist Adaptive Control (ARAC) strategy that switches between assistive stiffness and resistive damping, tuning impedance via radial basis functions updated by an adaptive law with forgetting and bias terms. ARAC was implemented on a custom 3-DOF cable-driven planar CDRR with inline load cells and quadratic tension allocation. Eight participants used an elastic fixture to simulate angle-dependent shoulder weakness, which limited their unaided range of motion. Compared to AAN, ARAC significantly increased interaction forces in resistive phases and overall EMG activity, without compromising the full range of motion. These results demonstrate that adaptive control in CDRRs can enable personalized rehabilitation by combining assistance and resistance.