Obstacle detection for a bipedal walking robot by a fisheye stereo

Having a sense of its environment is essential for a bipedal robot to walk around autonomously in unknown and uneven indoor environments. In a previous study, we measured the floor height necessary for stable walk by processing images acquired by a fisheye stereo mounted on the chest of a bipedal robot. In order to avoid collisions between the bipedal robot and environmental obstacles, very wide field of sensing area is necessary because the bipedal robot can move sideways. In this study, we developed a method for checking the existence of an object on the barrier, which is the front half surface of a cylinder surrounding the body of the robot, by using the fisheye stereo images. By combining this with the occupancy grid method, we can detect an object that is likely to collide with the walking robot body even while walking sideways. The results of a real experiment conducted using the bipedal robot validate the proposed method.