Maintaining subject engagement during robotic rehabilitation with a minimal assist-as-needed (mAAN) controller

One challenge of robotic rehabilitation interventions is devising ways to encourage and maintain high levels of subject involvement over long duration therapy sessions. Assist-as-needed controllers have been proposed which modulate robot intervention in movements based on measurements of subject involvement. This paper presents a minimal assist-as-needed controller, which modulates allowable error bounds and robot intervention based on sensorless force measurement accomplished through a nonlinear disturbance observer. While similar algorithms have been validated using healthy subjects, this paper presents a validation of the proposed mAAN control algorithm's ability to encourage user involvement with an impaired individual. User involvement is inferred from muscle activation, measured via surface electromyography (EMG). Experimental validation shows increased EMG muscle activation when using the proposed mAAN algorithm compared to non-adaptive algorithms.