Upper Extremity Control Using Reinforcement Learning

Case Western Reserve University, Cleveland OH

Investigators

Linked publications & trials

Abstract

DESCRIPTION (provided by applicant): Functional Electrical Stimulation (FES) is the electrical activation of peripheral nerves and muscles to restore motor function lost due to injury or disease (e.g., spinal cord injury, stroke, cerebral palsy). Optimal implementation of FES requires muscle activation patterns such that resulting movements functionally duplicate those of healthy individuals. Current FES technology utilizes fixed stimulation patterns, and movements do not yet have the smoothness, flexibility, and efficiency of normal human movement. A more desirable approach would be to develop intelligent control systems that can self-optimize by learning from experience. Reinforcement learning is a technique in which learning rules and reinforcement of successful parameters are used to achieve improvements in control over time. To date, reinforcement learning has not been used to refine control of upper extremity FES systems. For my project, I will initially apply reinforcement learning methods to FES control using a computer model of a FES-activated arm. In the subsequent stage of this project, I will apply reinforcement learning to the control of FES systems implanted in paralyzed individuals and quantify the improvements in arm movement quality.