Research Analyzer
← Back ICRA 2026

Breaking the Static Assumption: A Dynamic-Aware LIO Framework Via Spatio-Temporal Normal Analysis

Zhiqiang Chen, Cedric Le Gentil, Fuling Lin, Minghao Lu, Qiyuan Qiao, Bowen Xu, Yuhua Qi, Peng Lu

PDF

AI summary

Key figure (auto-extracted from paper)
Integrating spatio-temporal normal analysis directly into the registration loop breaks the pose-detection circular dependency, enabling robust localization in highly dynamic and geometrically sparse environments.
Dynamic LIO Spatio-temporal analysis Point cloud registration SLAM Dynamic object detection Autonomous navigation

Problem

Conventional Lidar-Inertial Odometry assumes static scenes and fails when moving objects dominate, creating a circular dependency where accurate pose estimation requires reliable static features, yet detecting dynamic objects requires precise pose information. This leads to severe localization drift and mapping errors in real-world dynamic settings.

Approach

The method embeds spatio-temporal normal analysis directly into the iterative closest point algorithm to simultaneously estimate pose and classify dynamic points without separate detection modules. It supplements this with a dual-map architecture and a spatial consistency check to filter false positives and construct a reliable static map.

Key results

  • Novel dynamic-aware ICP algorithm that jointly solves pose estimation and dynamic point classification
  • Dual-map architecture balancing temporal density for normal calculation with spatial richness for localization
  • Spatial consistency verification method that effectively filters false positives in static map construction
  • Open-source code and a new benchmark dataset for challenging dynamic environments

Why it matters

Enables reliable robot navigation and mapping in real-world scenarios with heavy traffic or moving objects, advancing the practical deployment of autonomous systems.

Abstract

This paper addresses the challenge of Lidar-Inertial Odometry (LIO) in dynamic environments, where conventional methods often fail due to their static-world assumptions. Tra- ditional LIO algorithms perform poorly when dynamic objects dominate the scenes, particularly in geometrically sparse environ- ments. Current approaches to dynamic LIO face a fundamental challenge: accurate localization requires a reliable identification of static features, yet distinguishing dynamic objects necessitates precise pose estimation. Our solution breaks this circular depen- dency by integrating dynamic awareness directly into the point cloud registration process. We introduce a novel dynamic-aware iterative closest point algorithm that leverages spatio-temporal normal analysis, complemented by an efficient spatial consis- tency verification method to enhance static map construction. Experimental evaluations demonstrate significant performance improvements over state-of-the-art LIO systems in challenging dynamic environments with limited geometric structure. The code and dataset are available at https://github.com/thisparticle/btsa.

Index terms

Localization SLAM Mapping

Related papers