Interactive control against obstacle for power assisted wheelchair by vision-based reaction force observer

This paper describes a novel approach to interactive power assisting control cooperating with visual information for robotic wheelchair. In the proposed approach, both encoder-based and vision-based reaction force observer are designed for obtaining the virtual force due to the environmental change in the field of view. The vision-based observer using the optical flow field of visual scene can estimate the external force including the inertial force due to the moving obstacle. The virtual force is defined as the difference between both reaction force observer outputs, and it is useful for assisting the avoiding operation against obstacle while power assisting control. The validity of our approach is evaluated by several experimental results.