Local Geometric Information Representation and Uncertainty Analysis in LiDAR SLAM

Kai Huang; Junqiao Zhao; Tiantian Feng

doi:10.3788/CJL241023

Chinese Journal of Lasers, Volume. 52, Issue 6, 0600003(2025)

Local Geometric Information Representation and Uncertainty Analysis in LiDAR SLAM

Kai Huang¹, Junqiao Zhao^2、*, and Tiantian Feng¹

Author Affiliations

¹College of Surveying and Geo-Informatics, Tongji University, Shanghai 200092, China

²School of Computer Science and Technology, Tongji University, Shanghai 200092, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(16)

Fig. 1. Illustration of the LOAM framework

Download full size

Fig. 2. Illustration of the LIO framework

Download full size

Fig. 3. Recent studies based on FAST-LIO2

Download full size

Fig. 4. Characterization of local geometric information of single-frame point clouds in some mainstream LiDAR SLAM systems

Download full size

Fig. 5. Characterization of local geometric information of map in some mainstream LiDAR SLAM systems

Download full size

Fig. 6. Structure of the map data in some mainstream LiDAR SLAM systems

Download full size

Fig. 7. Illustration of the point uncertainty model

Download full size

Fig. 8. Uncertainty models of point cloud in some mainstream LiDAR SLAM systems

Download full size

Fig. 9. Illustration of single-frame point cloud on NTU VIRAL sequence spms_03. The points are colored according to their uncertainty

Download full size

Fig. 10. Mapping comparison between LOG-LIO2 and FAST-LIO2 based on localization results in the M2DGR sequence street_07 at the same point cloud resolution

Download full size

Table 1. Comparison of different LiDAR SLAM algorithms for experiments

View table

Table 1. Comparison of different LiDAR SLAM algorithms for experiments

Name of the algorithm	Scan points representation	Map		Uncertainty
Name of the algorithm	Scan points representation	Structure	Representation	Factor	Propagation
FAST-LIO2^［55］	Coordinate	ikd-tree	Single point	Isotropic noise
Faster-LIO^{［57，80］}	Coordinate	iVOX	Single point	Isotropic noise
Point-LIO^［81］	Coordinate	iVOX	Single point	Isotropic noise
LOG-LIO^［59］	Coordinate， normal	ikd-tree	Points distribution	Isotropic noise
PV-LIO	Coordinate	Adaptive voxel	Points distribution	Range， bearing	BALM^［75］
LOG-LIO2^［82］	Coordinate， normal	Adaptive voxel	Points distribution， normal	Range， bearing	Incremental
LOG-LIO2^［82］	Coordinate， normal	Adaptive voxel	Points distribution， normal	Incident angle， roughness	Incremental

Table 2. Details of the M2DGR, NTU VIRAL, and Newer College datasets

View table

Table 2. Details of the M2DGR, NTU VIRAL, and Newer College datasets

Dataset	Mobile platform	LiDAR		f of IMU /Hz	Ground truth	Environments	Duration /s
Dataset	Mobile platform	Type	f /Hz	f of IMU /Hz	Ground truth	Environments	Min	Max
M2DGR^［116］	Wheeled robot	Velodyne-32	10	150	RTK， 3D laser tracker	Indoor， outdoor	127	1227
NTU VIRAL^［122］	UAV	Ouster-16	10	385	3D laser tracker		181	583
Newer College^［124］	Handheld	Ouster-64	10	100	6-DoF ICP		120	2180

Table 3. Experimental results on NTU VIRAL dataset: RMSE at specified map resolutions for each algorithm
View table
Table 3. Experimental results on NTU VIRAL dataset: RMSE at specified map resolutions for each algorithm
Algorithm Specified map resolution RMSE /m
LOG-LIO2 Adaptive 0.039
PV-LIO Adaptive 0.044
LOG-LIO 0.5 m 0.045
Point-LIO 0.5 m 0.041
Faster-LIO 0.5 m 0.084
FAST-LIO2 0.75 m 0.047
0.50 m 0.039
0.25 m 0.045

Table 4. Experimental results on M2DGR and Newer College datasets: RMSE at specified map resolutions for each algorithm

View table

Table 4. Experimental results on M2DGR and Newer College datasets: RMSE at specified map resolutions for each algorithm

Algorithm	Specified map resolution	RMSE /m
Algorithm	Specified map resolution	M2DGR	Newer College
LOG-LIO2	Adaptive	0.626	0.219
PV-LIO	Adaptive	0.747	0.242
LOG-LIO	0.4 m	0.682	0.275
Point-LIO	0.4 m	0.727	0.266
Faster-LIO	0.4 m	1.115	0.237
FAST-LIO2	0.6 m	0.943	0.254
	0.4 m	1.028	0.254
	0.2 m	1.110	0.265

Table 5. Experimental results on NTU VIRAL dataset: average running time at specified map resolutions for each algorithm
View table
Table 5. Experimental results on NTU VIRAL dataset: average running time at specified map resolutions for each algorithm
Algorithm Specified map resolution Running time /ms
LOG-LIO2 Adaptive 26.165
PV-LIO Adaptive 28.536
LOG-LIO 0.5 m 28.428
Point-LIO 0.5 m 11.367
Faster-LIO 0.5 m 6.655
FAST-LIO2 0.75 m 9.017
0.50 m 12.036
0.25 m 18.036

Table 6. Experimental results on M2DGR and Newer College datasets: RMSE at specified map resolutions for each algorithm

View table

Table 6. Experimental results on M2DGR and Newer College datasets: RMSE at specified map resolutions for each algorithm

Dataset	Specified map resolution	Average running time /ms
Dataset	Specified map resolution	M2DGR	Newer College
LOG-LIO2	Adaptive	45.376	57.820
PV-LIO	Adaptive	54.610	62.965
LOG-LIO	0.4 m	58.319	46.932
Point-LIO	0.4 m	25.514	25.488
Faster-LIO	0.4 m	12.894	14.315
FAST-LIO2	0.6 m	19.350	16.748
	0.4 m	29.038	23.482
	0.2 m	46.210	45.677

Tools

Get Citation

Copy Citation Text

Kai Huang, Junqiao Zhao, Tiantian Feng. Local Geometric Information Representation and Uncertainty Analysis in LiDAR SLAM[J]. Chinese Journal of Lasers, 2025, 52(6): 0600003

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: reviews

Received: Jul. 2, 2024

Accepted: Nov. 18, 2024

Published Online: Mar. 18, 2025

The Author Email: Junqiao Zhao (zhaojunqiao@tongji.edu.cn)

DOI:10.3788/CJL241023

CSTR:32183.14.CJL241023

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Comparison of different LiDAR SLAM algorithms for experiments

Table 1. Comparison of different LiDAR SLAM algorithms for experiments

Table 2. Details of the M2DGR, NTU VIRAL, and Newer College datasets

Table 2. Details of the M2DGR, NTU VIRAL, and Newer College datasets

Table 3. Experimental results on NTU VIRAL dataset: RMSE at specified map resolutions for each algorithm

Table 3. Experimental results on NTU VIRAL dataset: RMSE at specified map resolutions for each algorithm

Table 4. Experimental results on M2DGR and Newer College datasets: RMSE at specified map resolutions for each algorithm

Table 4. Experimental results on M2DGR and Newer College datasets: RMSE at specified map resolutions for each algorithm

Table 5. Experimental results on NTU VIRAL dataset: average running time at specified map resolutions for each algorithm

Table 5. Experimental results on NTU VIRAL dataset: average running time at specified map resolutions for each algorithm

Table 6. Experimental results on M2DGR and Newer College datasets: RMSE at specified map resolutions for each algorithm

Table 6. Experimental results on M2DGR and Newer College datasets: RMSE at specified map resolutions for each algorithm