Robot Localization Method for 3D Point Cloud Maps
SONG Yonglei
JIN Jianchen
Abstract：In order to address the problem of large time cost of robot position tracking in 3D point cloud maps,this paper pro-poses a new point cloud descriptor extraction method and a new Monte Carlo localization method for 3D point cloud maps by fusing particle filtering.The method first proposes a new SCS descriptor designed according to the structure of the LiDAR beam and the sim-ilarity measure between the corresponding descriptors,and then calculates the displacement of the particles by the robot wheel ta-chometer and updates the particle weights by using the descriptors of the robot location as the observation data and the descriptors of the particle location in the 3D point cloud map as the similarity measure.The pose of the robot in the 3D point cloud map is estimat-ed.The average distance error is only 0.637 6 m and the average heading error is 3.738 5°in the dataset of campus environment.Af-ter deployment on the actual vehicle,it is found that the localization effect is good and has practical application value.
Keywords：3D point cloud maprobot localizationpoint cloud descriptorpose estimation
Publication Date：2025-09-20
Online Publishing Date：2025-12-23(First online date of this platform, not the publication date of the document)
Pages：7( 2477-2483 )