IPNL Intelligent Positioning & Navigation Lab Search EN 简 繁

GLIO: Tightly-coupled gnss/lidar/imu integration for continuous and drift-free state estimation of intelligent vehicles in urban areas

Liu, X., Wen, W., Hsu, L. T.

IEEE Transactions on Intelligent Vehicles (2023)

journal Q1 Featured page
DOI
3D trajectory comparison for GLIO and other GNSS, LiDAR, and IMU integration methods.
Figure 4 visual detail: 3D trajectory comparison across GNSS, LiDAR, and IMU integration methods.

Summary

GLIO tightly couples GNSS, LiDAR, and IMU in a single factor graph for continuous drift-free state estimation in urban areas.

Figures

After the system pipeline, this figure explains the LiDAR constraint design that supports smoother and more globally consistent optimization.

Scan-to-map and scan-to-multiscan LiDAR association schemes.
Figure 2: LiDAR factor association through scan-to-map and scan-to-multiscan schemes.

The factor graph shows how GNSS, Doppler, IMU, and LiDAR constraints are organized across the first and second optimization stages.

Two-stage factor graph structure with GNSS, INS, Doppler, scan-to-map, and scan-to-multiscan factors.
Figure 3: factor graph structure for first-stage fusion and second-stage optimization.

The results begin with the Tsim Sha Tsui sequence, comparing trajectory consistency between GLIO variants and baseline methods.

3D trajectory comparison in the Tsim Sha Tsui sequence.
Figure 4: Tsim Sha Tsui trajectory comparison across RTKLIB, LIO, LIO-GNSS, GLIO-SS, and GLIO-DS.

These error plots quantify the same sequence and show where loosely coupled systems degrade while tighter fusion improves.

Tsim Sha Tsui positioning error plots comparing GLIO variants and baselines.
Figures 5 and 6: Tsim Sha Tsui positioning errors and scan-to-multiscan frame comparison.

The Whampoa case extends the evaluation to longer dense-urban driving with tunnels, traffic, and severe GNSS/LiDAR challenges.

Whampoa table, trajectory, and positioning error plots for GLIO and baseline methods.
Figures 7 and 8: Whampoa trajectory and error results under dense urban driving.

Key idea. GLIO tightly couples GNSS, LiDAR, and IMU in a single factor graph, so absolute GNSS positioning and drift-free LiDAR-inertial odometry constrain each other — global accuracy without the long-run drift of LiDAR-only systems.

Impact. Delivers continuous, drift-free state estimation for intelligent vehicles across urban areas, and is a flagship example of the lab’s multi-sensor fusion stack for real platforms.