Robust Mapping and Localization in Offline 3D Point Cloud Maps

Abstract

Aiming at the degradation of lidar, we propose a Robust Mapping and Localization (RMAL) method, which combines the classic Extended Kalman Filter (EKF) algorithm with the back-end pose graph optimization for 3D real-time mapping. Utilizing the complementary advantages of multiple sensors, the robustness of the mapping method is enhanced. In addition, we choose to save the feature keyframes and the corresponding optimal pose transformations as the offline map during the mapping process. Cooperating with subsequent mapping again, we can improve the positioning accuracy of the robot in the offline map. Finally, we also conduct experimental tests in different real scenarios, and the results verify the robustness and engineering practicability of the proposed method.