Assist-as-needed control of a soft rehabilitation robot for the finger using an interaction torque observer

This paper introduces an Assist-as-Needed (AAN) control strategy for a soft finger rehabilitation robot using a fiber-reinforced bending (FRB) actuator. Most conventional rehabilitation controllers focus solely on position or force control and fail to incorporate the patient’s active participation during the rehabilitation process. This lack of engagement limits the effectiveness of patient-specific rehabilitation, highlighting the need for interaction-based strategies. To address this gap, we propose a novel AAN approach that combines a nonlinear disturbance observer (NDO) and a sliding mode controller (SMC), designed using the dynamic model of the FRB actuator. The NDO estimates the patient–robot interaction torque in real-time without using external sensors or modeling the patient’s finger, while the SMC ensures robust and accurate position tracking. This interaction-based assistance enables the system to dynamically adjust the level of assistance based on the patient’s voluntary efforts, while the controller simultaneously works to guide the finger movement toward the rehabilitation trajectory, promoting a more active and patient-specific rehabilitation process. Additionally, the stability of the closed-loop system is analytically proven using Lyapunov theory. The proposed method is validated both on a finger-like mechanism simulating various patient scenarios and on a wearable rehabilitation glove used by a participant. Experimental results confirm that the AAN controller achieves superior trajectory tracking and provides adaptive assistance aligned with the patient’s capability, outperforming conventional model-free controllers.