A chord-angle-based approach with expandable solution space to 1-degree-of-freedom (DOF) rehabilitation mechanism synthesis

Abstract

Rehabilitation robots have been proven to be an effective tool for patient motor recovery in clinical medicine. Recently, few degrees of freedom (DOFs), especially 1-DOF, rehabilitation robots have drawn increasing attention as the complexity and cost of the control system would be significantly reduced. In this paper, the mechanism synthesis problem of 1-DOF rehabilitation robots is studied. Traditional synthesis methods usually aim at minimizing the trajectory error to generate a mathematically optimal solution, which may not be a practically feasible solution in terms of engineering constraints. Therefore, we propose a novel mechanism synthesis approach based on chord angle descriptor (CAD) and error tolerance expansion to generate a pool of mechanism solutions from which mathematically and practically optimal solutions can be selected. CAD is utilized for its capability to represent the same-shaped trajectories of different mechanisms in a unified way, and it is robust to the noise in the rehabilitation trajectory acquired by motion capture systems. Then a library of mechanism trajectories is established with compressed representations of CAD via an auto-encoder algorithm to speed up the matching between mechanism and rehabilitation trajectory where the matching error tolerance can be adjusted according to practical rehabilitation specifications. Finally, a design example of a 1-DOF rehabilitation robot for upper-limb training is provided to demonstrate the efficacy of our novel approach.