Development of an Assist-As-Need Controller for an Upper-Limb Exoskeleton With Voluntary Torque Estimate

This study considers the development of an assist-as-need torque controller for an exoskeleton for stroke rehabilitation. Studies have shown that active patient participation improves the patient’s recovery from stroke. Assist-as-need control, providing the patient with the assistance they need to complete a task, is desirable, as the assistance can be varied to maximise patient participation. However, research is limited, and current methods cannot guarantee optimal assistance as non-zero assistive forces are still provided to subjects that are capable of completing the task unassisted. This study proposes a control system to vary and optimise the assistance for a subject based on their capabilities. A particle filter developed from previous research is used to estimate the subject’s voluntary effort. The assistive torque is determined from a target torque and the voluntary effort. The controller is shown to be effective, as zero assistance is provided to a subject capable of completing the task unassisted. Additionally, the assistance will increase if the subject fatigues. Using the estimate of the subject’s strength, the assistive torque can be accurately set to maximise a patient’s participation, and therefore, the assist-as-need controller can lead to improved therapeutic results.