Vision-based human-following control using optical flow field for power assisted wheelchair

This paper describes a power assisting control with human following function for robotic wheelchair. In our past research, visual feedback control of the wheelchair has been proposed for power assisting according to environmental changes, and it is useful for power assisted wheelchair to integrate several driving assistance, such as obstacle avoidance, following motion only from unified motion controller. In this paper, we propose human following function on the vision-based power assist controller. The proposed method employs optical flow vectors in the field of camera view, and it can be realized only from human leg view. In case that the camera system is mounted at lower position, then the camera image only includes human lower limb. The proposed method allows for stably tracking the walking human even in such a case. The validity of our approach is evaluated by several experimental results.