The study proposes a LiDAR odometry method that leverages environmental landmarks to enhance pose estimation precision. It involves two sequential pose estimation stages: horizontal and vertical. The algorithm evaluates the similarity between landmarks considering their dimension and shape. Tested with the KITTI dataset and data from an unmanned ground vehicle, the results show that this method outperforms existing LiDAR odometry solutions in terms of positioning accuracy. The proposed approach not only improves localization precision but also constructs a detailed global map made of the extracted landmarks.
Publication date: 29 Dec 2023
Project Page: http://ieeexplore.ieee.org
Paper: https://arxiv.org/pdf/2312.16787
