Experimental Data-Driven Angular Velocity Control for a Soft Pneumatic Wrist Exoskeleton for Motion Assistance in Rehabilitation

Soft robotic devices have proven to be highly effective in supporting stroke patients by offering adaptive, gentle, and precise assistance for rehabilitation and mobility improvement. Their inherent flexibility and compliance make them ideal for promoting recovery in individuals with neurological impairments. This paper presents a methodology for utilizing experimental data to develop a control model for a pneumatic soft robotic wrist exoskeleton. The exoskeleton is equipped with capacitive-type flexible sensors to measure angular displacement and employs a proportional-integral (PI) controller to regulate wrist flexion and extension movements. To establish reliable control parameters, a linear dynamic state-space model for the wrist exoskeleton is derived from experimental data. The system identification method enabled the development of state-space models that closely matched the experimental system with minimal error. A simulation model of angular velocity control is created in MATLAB/Simulink, incorporating the estimated state-space model and PI controller. This simulation is used to fine-tune the PI parameters, which are then programmed into a microcontroller to control the soft robotic wrist exoskeleton. Experimental results demonstrated angular velocity tracking with average errors of 0.18-0.26 deg/sec across different configurations and velocity commands, validating the effectiveness of the system identification approach and simulation-based tuning in reducing manual tuning efforts while achieving precise control of the required angular velocities.