移動ロボットのためのLiDARデータの遮蔽にロバストな自己位置推定法

抄録

For autonomous navigation of mobile robots, localization of the position and orientation is a fundamental issue. For this issue, we assume a mobile robot equipped with a 3D LiDAR sensor. Scan matching with the 3D LiDAR enables the robot to achieve localization. However, the scan data is sometimes occluded by obstacles. In this case, since landmarks in the map are not observed by the robot, scan matching causes a localization error. For this localization challenge, we first use octrees to distinguish the occluded scan data. For scan matching, these scan data are then removed. For the reduced scan data, we further introduce a local map based on online SLAM. Through the experiments, we show that the robot is able to robustly localize even in a dense environment with obstacles.