Autonomous navigation of mobile robot based on teaching and playback using trinocular vision

The purpose of the study is to realize autonomous navigation of a mobile robot in an indoor environment by teaching and playback method, without making any environmental map in advance. The strategy of navigation in the study is that the robot runs based on its odometry information and corrects its position using landmark information obtained from a stereo vision sensor. We use a trinocular vision system to solve easily the stereo matching problem and choose vertical lines in the environment as landmarks because of their easiness to detect and sufficient existence in the environment. In the paper, the method of teaching and playback is described and some experimental results are also shown.