Vision Based Power Assist Control for Functional Behavior of Robotic Wheelchair

This paper presents a vision based novel approach of power assist motion control for robotic wheelchair. In the proposed motion controller, the human force and wheelchair velocity are estimated only from visual information. This approach using vision and force based assist control makes it possible to facilitate sophisticated interactions between human force and dynamic environments such as human following assist, obstacle avoidance and so on. For velocity estimation, three dimensional information is obtained by stereo images, and the optical flow vectors are also used for self-velocity estimation at real-time, and its velocity is used as feedback signal of velocity controller of wheelchair. Furthermore the force based assist function is integrated into vision based velocity controller. Then human force is also estimated by vision based reaction force observer. The validities of the power assist control are verified by several experimental results.