Bilateral constrained control for prosthesis walking on stochastically uneven terrain

Abstract

The stochastic characteristic of uneven terrain is the leading cause of stochastic ground reaction forces, hindering gait coordination between the lower-limb prosthesis and healthy limb. Therefore, dedicated modeling and control methods for prosthesis walking on uneven terrain should be urgently provided. Noting that the stochastic feature originates from uneven ground contact, this paper first applies data-driven modeling to accurately represent the stochastic ground reaction forces. Then, an extended stochastic dynamic model for the amputee walking on uneven terrain is established based on this representation. Accordingly, a stochastic-gait-coordination (SGC) oriented control architecture is proposed and realized via the bilateral constrained adaptive neural controller (BC-ANC), where BC is designed to ensure disturbance boundness, and ANC is employed to handle system uncertainty. Rigorous stochastic stability analysis of the proposed architecture, i.e., BC-ANC, is also provided by deriving an asymmetric barrier Lyapunov function based on the backstepping technique. The study reveals that the controlled states in the stochastic dynamic system promise semi-globally uniformly ultimately bounded (SGUUB) in probability. Furthermore, prediction studies are deployed, showing that BC-ANC is more efficient than existing controllers.