Realization of stair ascent and motion transitions on prostheses utilizing optimization-based control and intent recognition

This paper presents a systematic methodology for achieving stable locomotion behaviors on transfemoral prostheses, together with a framework for transitioning between these behaviors-both of which are realized experimentally on the self-contained custom-built prosthesis AMPRO. Extending previous results for translating robotic walking to prosthesis, the first main contribution of this paper is the gait generation and control development for realizing dynamic stair climbing. This framework leads to the second main contribution of the paper: a methodology for motion intent recognition, allowing for natural and smooth transitions between different motion primitives, e.g., standing, level walking, and stair climbing. The contributions presented in this paper, including stair ascent and transitioning between motion primitives, are verified in simulation and realized experimentally on AMPRO. Improved tracking and energy efficiency is seen when the online optimization based controller is utilized for stair climbing and the motion intent recognition algorithm successfully transitions between motion primitives with a success rate of over 98%.