Performance-based assistance control for robot-mediated upper-limbs rehabilitation

Clinical research results demonstrated that robot-mediated training for the recovery of subjects after the stroke had the benefit of being able to carry out high-intensity repetitive and task-oriented tasks, which could effec-tively facilitate the rehabilitation effect. However, these subjects may have different training requirements as the disability levels vary from subject to subject at different stages of rehabilitation. Therefore, to meet the reha-bilitation training needs of subjects in different rehabilitation stages, a performance-based assistance control strategy for safe robot-mediated therapy is proposed in this paper. The strategy contains three training modes that are switched by the trajectory tracking error, the zero interaction force mode (ZIF), assist-as-needed (AAN) mode, and restriction interaction region (RIR) mode. In ZIF mode, the subject can freely refer to their motion trajectory, and the robot will only offer fixed assistance only when the subject is slacking. In AAN mode, the robot would offer a variable level of assistance force by adjusting the stiffness coefficient to assist subjects in accordance with their motor abilities. To assure safety, a sizable assisting force is applied in RIR mode. To validate and evaluate the effectiveness and performance of the proposed assistance control strategy, a set of task -oriented continuous movement experiments based on an end-effector robot-mediated training system were performed by five able-bodied subjects. The preliminary experiment results demonstrated that the proposed control strategy works properly and could provide corresponding assistance and adjust the working modes ac-cording to the subject's motor performance. It was found that the robot provided more assistance as the increasing trajectory tracking error led to a decrease in motor performance. Additionally, adaptive adjustment of the stiffness coefficient allowed the proposed assistance control strategy to induce more active effort. It's worth noting that the parameters of the proposed strategy may be personalized and adjusted for different subjects to satisfy various requirements.

Automated summary

Clinical research has shown that robot-mediated training is beneficial for stroke recovery by allowing high-intensity repetitive and task-oriented tasks. However, as the disability levels and rehabilitation stages vary among subjects, different training requirements need to be met. This paper proposes a performance-based assistance control strategy that includes three training modes switched by trajectory tracking error, namely, the zero interaction force mode, assist-as-needed mode, and restriction interaction region mode.