A LiDAR Localization Method for Mobile Robots Based on Fast Search and Outlier Mitigation

LiDAR localization gains more popularity by providing state information for robots in GPS-denied environments. Improving both the accuracy and computational efficiency is an ongoing effort for all LiDAR localization algorithms. In this paper, we propose a LiDAR localization method composed of a fast direct registration method and a sliding-window-based IMU propagator in a tightly coupled way. Also the factor graph optimization based on outlier mitigation is served as a back-end for robust state estimation. Our approach is extensively evaluated on multiple public datasets and also analyzed in the aspect of running time. Experimental results show that our system achieves high accuracy in various datasets while maintaining a real-time performance.