Multimodal Perception-Based Adaptive Admittance Control for Cable-Driven Lower Limb Rehabilitation Robot
To address the coupling between ambiguous sEMG-based intention recognition and mechanical safety constraints in cable-driven lower-limb rehabilitation, this study proposes a multimodal perception-based adaptive admittance control framework. Mechanical, interaction, and physiological indicators, including stiffness, end-effector position, coupling force, velocity, joint motion, and sEMG, were fused to estimate active participation, passive/flaccid tracking, and spasticity risk. The identified state was then used to adjust admittance parameters online, while an RBFNN-SMC inner loop improved trajectory tracking. Experiments showed high-damping, low-stiffness protection during spasticity bursts and milder responses during passive conditions. Compared with PID, RBFNN-SMC reduced tracking errors, supporting safer adaptive rehabilitation control.