Research on loop detection algorithm for voxelized generalized iterative closest point
Synchronous localization and mapping (SLAM ) is one of the key technologies for large-scale localization and mapping.Aiming at the large cumulative error problem in outdoor large scene environment mapping,which leads to low positioning precision and map ghosting and drift,a loop detection algorithm based on voxelized generalized iterative closest point(VGICP)optimization is proposed.This method extends the generalized iterative closest point(GICP)algorithm to calculate multiple local points in voxels,which ensures accuracy and avoids costly nearest neighbor search.The proposed method is added to the complete SC-LeGO-LiDAR odometry and mapping in real-time(LOAM)framework and tested using the KITTI dataset 05 sequence.The experimental results show that the trajectory estimated by the optimization algorithm has a high coincidence with the real trajectory,and the maximum absolute pose error(APE)and relative pose error(RPE)have decreased by 46.4％ and 18.8％ respectively;The mean square error decreases by 17.7％ and 19.9％.The optimization algorithm can further improve the mapping precision and reduce the attitude drift error.
NETL
NSTL
万方数据
