LEADER: Learning Reliable Local-to-Global Correspondences for LiDAR Relocalization

TL;DR

LEADER is a robust LiDAR relocalization framework that uses a geometric encoder and a reliability loss to improve pose estimation accuracy in complex environments.

Contribution

It introduces a geometric encoder and a reliability loss to enhance the robustness and accuracy of learning-based LiDAR relocalization methods.

Findings

LEADER outperforms state-of-the-art methods on Oxford RobotCar and NCLT datasets.

Achieves 24.1% and 73.9% relative reductions in position error.

Demonstrates robustness in challenging scenes with noise and outliers.

Abstract

LiDAR relocalization has attracted increasing attention as it can deliver accurate 6-DoF pose estimation in complex 3D environments. Recent learning-based regression methods offer efficient solutions by directly predicting global poses without the need for explicit map storage. However, these methods often struggle in challenging scenes due to their equal treatment of all predicted points, which is vulnerable to noise and outliers. In this paper, we propose LEADER, a robust LiDAR-based relocalization framework enhanced by a simple, yet effective geometric encoder. Specifically, a Robust Projection-based Geometric Encoder architecture which captures multi-scale geometric features is first presented to enhance descriptiveness in geometric representation. A Truncated Relative Reliability loss is then formulated to model point-wise ambiguity and mitigate the influence of unreliable predictions. Extensive experiments on the Oxford RobotCar and NCLT datasets demonstrate that LEADER outperforms state-of-the-art methods, achieving 24.1% and 73.9% relative reductions in position error over existing techniques, respectively. The source code is released on https://github.com/JiansW/LEADER.