Research Analyzer
← Back ICRA 2026

Onion-LO: Why Does LiDAR Odometry Fail across Different LiDAR Types and Scenarios?

Xiaolong Cheng, Keke Geng, Zhichao Liu, Tianxiao Ma, Ye Sun

PDF

AI summary

Key figure (auto-extracted from paper)
Onion-LO delivers robust, parameter-free LiDAR odometry across diverse sensors and environments using an adaptive spatial partitioning structure.
LiDAR odometry SLAM adaptive point cloud processing generalizable navigation real-time localization Onion Ball

Problem

Existing LiDAR odometry systems fail or require heavy manual tuning when deployed across different LiDAR types and dynamic scenarios due to rigid parameter settings and environment-dependent feature extraction.

Approach

The framework uses an 'Onion Ball' structure to dynamically analyze point cloud distribution, enabling adaptive downsampling, segmentation, and real-time optimization parameter tuning without manual configuration.

Key results

  • Outperforms state-of-the-art methods in accuracy and robustness across five datasets
  • Validated across 11 LiDAR sensors and 8 diverse scenarios
  • Enables real-time operation on onboard processors without manual tuning
  • Provides plug-and-play compatibility with existing odometry frameworks

Why it matters

Enables reliable, out-of-the-box navigation for autonomous systems using heterogeneous LiDAR sensors in dynamic real-world environments.

Abstract

LiDAR odometry is a fundamental technology for autonomous navigation. However, existing LiDAR-based odometry methods typically demand extensive manual parameter tuning and remain prone to instability when deployed across varying LiDAR types and environments. This letter focuses on the essence of point clouds and introduces a fast, highly adaptable, and robust LiDAR odometry framework named Onion-LO. Onion-LO demonstrates strong compatibility with various LiDAR types and reliable operation across diverse scenarios. This is facilitated by an onion-like point cloud processing structure termed Onion Ball. The Onion Ball supports multi-threaded implementation, efficiently executing point cloud distribution analysis, segmentation, and downsampling. In addition, we design an adaptive optimization strategy for local map management and iterative optimization, which effectively enhances the system's robustness and accuracy. Extensive experiments on five datasets demonstrate that Onion-LO outperforms existing state-of-the-art methods regarding localization accuracy and robustness. Additional evaluations across 11 LiDAR sensors and 8 diverse scenarios further confirm its strong generalization capability. Our method is designed for practical deployment and supports real-time operation on onboard processors. We open-source the code on https://github.com/huashu996/Onion-LO.

Index terms

SLAM Localization Field Robots

Related papers