Article Figures & Data
Figures
- FIGURE 1
Graphical overview of the lidar odometry performed via scan matching After each new point cloud is received from the sensor, the scan-matching algorithm computes the change in vehicle pose
, as well as the associated state-error covariance matrix P. The dead reckoning solution accumulates these data. - FIGURE 2
Voxel point distributions featuring (A) sensor noise, (B) surface roughness, and (C) extended surfaces
Representative point-cloud standard deviations in the wall-normal and tangent directions (σn and σt, respectively) are listed for each case.
- FIGURE 3
Illustration of a wall (seen from above) passing through two rounded voxels Lidar measurements of the surface are shown as dots. Shaded ellipsoids encode voxel statistics, with centers indicating the sample mean and principal axes sized to twice the standard deviation.
- FIGURE 4
Sigma-point refinement for identifying extended surfaces within a single voxel A covariance ellipse is shown in relation to a wireframe voxel. The sigma points along the ellipsoid’s three principal axes (labeled I, II, and III) are identified by small dots. Only axis III is eliminated by our sigma-point exclusion text, as both of its sigma points lie outside the voxel boundaries.
- FIGURE 5
Spherical voxels with adaptive radial binning Point distributions within each cell are shown as ellipsoids.
- ALGORITHM 1
ICET Scan Registration
- FIGURE 6
Synced forward-facing images from LeddarTech PixSet (Déziel et al. (2021)), drive 20200721_144638_part36_1956_2229 through Old Montreal, Quebec
- FIGURE 7
ICET demonstration for real-world PixSet data: (left) ICET odometry trajectory, shown in white, with an HD map generated by using the odometry to stitch scans together; (right) vehicle trajectory projected on the north–east plane, as measured by three methods: GPS ground truth, ICET odometry, and NDT odometry
- FIGURE 8
ICET performance for simulated CODD data: (left) ICET odometry trajectory, shown in white, with an HD map generated by using the odometry to stitch scans together; (right) vehicle trajectory projected on the north–east plane, including the known truth from the simulation (dashed line) as well as ICET and NDT odometry measurements (solid lines)
- FIGURE 9
NDT- and ICET-generated estimates for forward translation and rotation about the vertical axis, compared with ground truth values
Results were generated with the CODD simulation, and hence, the ground truth and error values are known exactly. Shaded regions show the 2σ error bound for each frame predicted by each algorithm.
- FIGURE 10
Corner-case simulations including scans for (left) a T-intersection, (middle) a tunnel, and (right) a large open field
- FIGURE 11
Sensitivity study: Performance as a function of voxel resolution
- FIGURE 12
Three examples of constructing the projection matrix L
Tables
x (cm) y (cm) z (cm) ϕ (deg) θ (deg) ψ (deg) NDT: RMSE 0.0128 0.101 0.0158 0.00131 0.00218 0.00104 NDT: Predicted 0.0122 0.0522 0.0147 0.00127 0.00199 0.00102 ICET: RMSE 0.0122 0.0709 0.0159 0.00134 0.00213 0.00103 ICET: Predicted 0.0124 0.0736 0.0152 0.00129 0.00215 0.00105 x (cm) y (cm) z (cm) ϕ (deg) θ (deg) ψ (deg) NDT: RMSE 0.0182 10.493* 0.0113 0.00147 0.00148 0.000987 NDT: Predicted 0.0164 0.0655 0.0106 0.00140 0.00142 0.000881 ICET: RMSE 0.0171 Marked DNU 0.0110 0.00144 0.00151 0.000920 ICET: Predicted 0.0170 Marked DNU 0.0107 0.00143 0.00153 0.000901 ↵* Output error unbounded for this solution component
x (cm) y (cm) z (cm) ϕ (deg) θ (deg) ψ (deg) NDT: RMSE 10.447* 10.424* 0.0116 0.00127 0.00128 1.73* NDT: Predicted 0.181 0.181 0.0115 0.00127 0.00128 0.0121 ICET: RMSE Marked DNU Marked DNU 0.0117 0.00128 0.00127 Marked DNU ICET: Predicted Marked DNU Marked DNU 0.0115 0.00128 0.00128 Marked DNU ↵* Output error unbounded for this solution component
Jump to section
- Article
- Abstract
- 1 INTRODUCTION
- 2 DEFINING THE PROBLEM
- 3 IMPLEMENTATION
- 4 VERIFICATION TESTING: REAL LIDAR DATA SET
- 5 VERIFICATION TESTING: HIGH-FIDELITY SIMULATED DATA SET
- 6 VERIFICATION TESTING: ABSTRACTED GEOMETRIES
- 7 CONCLUSION
- HOW TO CITE THIS ARTICLE
- CONFLICT OF INTEREST
- ACKNOWLEDGMENTS
- A APPENDIX: PARAMETRIC STUDY OF SPHERICAL VOXEL RESOLUTION
- B APPENDIX: CONSTRUCTING THE VOXEL PROJECTION MATRIX
- Footnotes
- REFERENCES
- Figures & Data
- Supplemental
- References
- Info & Metrics
Related Articles
Cited By...
- No citing articles found.