Robotic Knee Tracking Control to Mimic the Intact Human Knee Profile Based on Actor-Critic Reinforcement Learning

We address a state-of-the-art reinforcement learning (RL) control approach to automatically configure robotic prosthesis impedance parameters enable end-to-end, continuous locomotion intended for transfemoral amputee subjects. Specifically, our actor-critic based RL provides tracking of knee mimic the intact profile. This is significant advance from previous automatic tuning which have centered on regulation with designer prescribed profile as target. In addition presenting complete algorithm direct heuristic dynamic programming (dHDP), we provide an analytical framework controller constrained inputs. show that proposed possesses several important properties, such weight convergence networks, Bellman (sub)optimality cost-to-go value function and input, practical stability human-robot system under input constraint. further systematic simulation using realistic simulator, OpenSim, emulate how dHDP enables level ground walking, walking different terrains at paces. These results not only theoretically suitable, but also practically useful.