Using Potential Field Function With a Velocity Field Controller to Learn and Reproduce the Therapist's Assistance in Robot-Assisted Rehabilitation

Rehabilitative and assistive practices usually elicit intense and repetitive exercises. Thus, there has been an increasing interest in robotic systems as they are robust and cost-effective in comparison to conventional physical motor-therapy with a therapist. These robots have applications in therapeutic and in-home environments, where there is a necessity for a user-friendly procedure to program the robots for a specific task easily. Our group has suggested robot learning from demonstration (LfD) as an intuitive procedure to program robots via short-term physical interaction in rehabilitation and assistive applications. In this article, a therapist assists a patient, and cooperatively performs a task on a robotic manipulator. Then, using a nonparametric potential field function, the therapist's motion, and interaction force (assistance/resistance) is modeled time-independently via a convex optimization algorithm. Next, in the therapist's absence, the robot provides the patient with the same level of interaction force provided by the therapist along the trajectory. A velocity field controller is also designed to compensate and regulate the patient's deviation from the velocity observed in the demonstration phase. Finally, the efficacy, advantages, and stability of the proposed framework are evaluated in three different experimental scenarios involving spring arrays and an individual with cerebral palsy (CP).