6-DOF Localization in 3D Feature Points Maps for LiDARs of Small FoV

Abstract

LOAM (lidar odometry and mapping) plays an important role in the field of mobile robot and relocalization based on prior maps is essential for LiDAR-based navigation. In this paper, we propose a system of LOAM and real-time relocalization only based on a low cost Lidar with small FoV (field of view). By taking effort on the matching method, the problem of drift at the sharp turn is settled and the robustness in map-building is improved. To accelerate the matching process in relocalization, our relocalization algorithm is based on the feature points map, not the full map. Finally we demonstrate our system to be accurate and robust by experiments in our school.