Autonomous navigation of a mobile robot based on environmental information observed during a past human-guided movement

It is necessary for a robot to know environmental information in order to move autonomously. In our method, we first guide a mobile robot to a destination by remote control. During this movement, the robot collects environmental information and makes a map along the path. Then the robot quickly reaches the destination autonomously using the map. We use stereo vision to obtain range data for detecting objects in the surrounding environment. Since the computational cost of finding correspondence is high, we use an image processor for real time processing. The map is used to find the region where the robot can move safely. We therefore record on the map only the nearest object boundary point. After human-guided movement, we compute the shortest path to reach the destination because the human-guided path is not always the best. Subsequently, the robot moves autonomously along the computed path. The robot observes the surrounding environment to avoid obstacles and to estimate its position in the map.