Abstract This paper introduces a novel stiffness-based Assist-as-Needed (AAN) control strategy aimed at optimizing the balance between robotic assistance and patient engagement in wrist rehabilitation. Central to the proposed approach is a joint stiffness estimation technique that enables precise, direction-specific assessment of the patient's motor condition. The derived stiffness profile is utilized to adaptively modulate key AAN control parameters, specifically the radius and velocity of the error tolerance circle, thereby individualizing assistance according to the patient's functional capacity. Additionally, the framework facilitates sensorless implementation of AAN control by employing a series elastic actuator, which permits accurate torque control and interaction force estimation using only encoder feedback. This design choice minimizes hardware complexity and cost, while maintaining desirable system characteristics such as backdrivability and compliance. Experimental results validate the efficacy of the proposed method, demonstrating its superiority over conventional AAN approaches in minimizing unnecessary robotic assistance and enhancing voluntary motor involvement. The presented strategy offers a scalable, quantitatively informed control paradigm suitable for precision rehabilitation across diverse impairment levels.
Lin et al. (Thu,) studied this question.