Reinforcement Learning Control of Functional Electrical Stimulation of the upper limb: a feasibility study.

摘要

Controllers for Functional Electrical Stimulation are still not able to produce natural movements of the paretic arm. In this work, Reinforcement Learning was used to design a non-linear controller for a hybrid upper limb robotic system thought for stroke rehabilitation. The performance of the controller was tested on one healthy subject during elbow extensions in the horizontal plane. Experimental results showed an absolute position error <0.7° for a maximum range of motion of 40° and stability against perturbation induced by simulated muscle spasms. Promising results must be confirmed on a broader population.