Adaptive Dynamic Programming-Based Control of an Ankle Joint Prosthesis

The potential of an adaptive dynamic programming (ADP)-based control strategy for learning the human gait dynamics in real-time and generating control torque for a prosthetic ankle joint is investigated in this paper. This is motivated by the desire for control strategies which can adapt in real-time to any gait variations in a noisy environment while optimizing some gait related performance indices. The overall amputated leg-prosthetic foot system is represented by a link-segment model with the kinematic patterns for the model are derived from human gait data. Then a learning-based control strategy including an ADP-based controller and augmented learning rules is implemented to generate torque which drives the prosthetic ankle joint along the designed kinematic patterns. Numerical results show that with the proposed learning rules, the ADP-based controller is able to maintain stable gait with robust tracking and reduced performance indices in spite of measurement/actuator noises and variations in walking speed. Promising results achieved in this paper serve as the starting point for the development of intelligent ankle prostheses, which is a challenge due to the lack of adequate mathematical models, the variations in the gait in response to the walking terrain, sensor noises and actuator noises, and unknown intent of users.