In this paper, we propose the method for efficient real-time estimation of the odometry using a rotating 2D laser scanner and a scan-matching method. Our scan-matching method uses the odometry model, constituted 9-DoF parameters of positions, postures and linear velocities. And all parameters are estimated by minimizing the cost-function obtained distances between point clouds to surfels. In the experiment, we evaluated the accuracy of our method when the robot passes on a rough terrain space and a foresty work road. As result, The maximum position error is 0.19 m of 30 m distance traveled, compared with the RTK-GNSS output in a rough terrain space. And more highly accurately than a conventional method in a foresty work road.
J-STAGEがリニューアルされました! https://www.jstage.jst.go.jp/browse/-char/ja/
