Odometer-Agnostic Drift Correction Using OpenStreetMap Lane Geometry

Joaquin Caballero, Emilio Garcia-Fidalgo, Alberto Ortiz, Jarno Ralli · N/A · 2026

Summary

Despite significant progress in odometry estimation, long-term drift remains a fundamental limitation of incremental pose integration, especially in large-scale or loop-free environments.

Abstract Summary

Despite significant progress in odometry estimation, long-term drift remains a fundamental limitation of incremental pose integration, especially in large-scale or loop-free environments. Existing map-assisted methods can reduce drift, but often depend on dense maps, sensor-specific processing, or complex matching pipelines. We propose a lightweight open-source, odometry-agnostic correction method that aligns short trajectory segments to OpenStreetMap (OSM) lane centerlines. By formulating drift correction as a direct alignment between recent odometry and sparse lane geometry, the method enables efficient online operation without dense priors or expensive preprocessing. Experiments with LiDAR and visual odometry backends demonstrate consistent improvements, with particularly strong gains under severe drift.

Key Points

Despite significant progress in odometry estimation, long-term drift remains a fundamental limitation of incremental pose integration, especially in large-scale or loop-free environments.
Existing map-assisted methods can reduce drift, but often depend on dense maps, sensor-specific processing, or complex matching pipelines.
We propose a lightweight open-source, odometry-agnostic correction method that aligns short trajectory segments to OpenStreetMap (OSM) lane centerlines.
By formulating drift correction as a direct alignment between recent odometry and sparse lane geometry, the method enables efficient online operation without dense priors or expensive preprocessing.
Experiments with LiDAR and visual odometry backends demonstrate consistent improvements, with particularly strong gains under severe drift.

Method

Our approach can be understood as a continuous 2D alignment procedure that corrects odometry drift by registering variable-length segments of the estimated trajectory to lane-level map points extracted from OSM. Specifically, we use lane centerlines as lightweight geometric landmarks and solve for an incremental rigid transform in $\text{SE}(2)$ that refines the odometry output over time.

Experimental Setup

This section describes the experimental protocol used to evaluate the proposed OSM lane-based odometry correction method. We evaluate the method in two complementary ways. First, we apply it to multiple baseline odometry backends to demonstrate that the correction is independent of the underlying odometry source.

Second, we compare against recent OSM-assisted drift-correction methods on KITTI Odometry, where published results are available for common baselines.

Experimental Setup — Odometer-Agnostic Drift Correction Using OpenStreetMap Lane Geometry — Fig. 2: Illustration of lane–pose matching. The red curve represents the trajectory, the yellow point is the pose under evaluation, the blue segment represents the normal projection direction, and light-blue points correspond to valid candidate correspondences obtained from the intersection between the projection and neighboring lane segments. Finally, point C is selected as the correct matching point.

Results

In this section, we evaluate the proposed correction approach in three stages. First, we compare against prior OSM-assisted odometry-correction methods on the KITTI sequences commonly reported in the literature. Second, we evaluate the same correction framework across multiple odometry backends and datasets to show that the method is independent of the underlying odometry source.

Finally, we report the computational cost of the correction module. Table I compares our correction with the LiDAR–OSM drift-correction methods of Li et al.

Results — Odometer-Agnostic Drift Correction Using OpenStreetMap Lane Geometry — Fig. 3: Representative KITTI-360 corrections using OSM lane centerlines. Left: LiODOM seq. 07. Right: Basalt seq. 02.

Sources and Demonstrations

Figures are reproduced from Caballero et al. See the full paper for experimental details and the project page, when the paper links one, for demonstrations and videos.