Insights into Modeling Functional Trajectories for Robot-Mediated Daily Living Exercise Environments

Abstract

Robotic therapy systems for stroke rehabilitation typically facilitate the practice of point-to-point reaching tasks utilizing straight trajectories that have zero velocity and acceleration at the beginning and end. In order to understand how to support the performance of real-life activities within a robot-mediated therapy environment, we studied trajectories created by the non-dominant wrist of able-bodied persons during more functional tasks such as drinking. In this paper, we report on how these more functional movements differ from key assumptions used to define algorithms for trajectory planning in robot therapy environments. We note a significant difference in curvature in both the XY and XZ planes of reaches towards real or imagined objects compared to the minimum jerk trajectories predicted for point-to-point reaches. We discuss implications of these findings and how they may influence the modeling of more natural trajectories within a more functional, task-oriented, robotic training system such as the activities of daily living exercise robot (ABLER) therapy environment