標識情報およびスキャン情報を用いた未知環境におけるナビゲーション

抄録

Localization and navigation are an important factor for autonomous mobile robots. As progress in CPU and GPU performance has been made, scan matching can be the major method of localization. Before scan matching, a dense map of the activity area is necessary, and to do path planning, the coordinate value of self-position and goal is efficient. However, we don’t need such map nor coordinate value. Even in environment never been to, we can reach to the destination if there are sign symbols showing the direction of the destination in the route even if the sign does not show the precise direction. This paper proposes a method using the name and direction of the destination, scan data of the wall, and odometry calculated from encoder and IMU. This method was tested in 2 courses in a 3D simulation environment representing Shinagawa station made by Gazebo. As a result, the robot model was able to reach goal-1 without any help, and goal-2 with 0.8 times help by human in average.