Research Analyzer
← Back ICRA 2026

Robust Localization in Large-Scale Symmetric Environments through Dynamic Topological Mapping

Rafael Flor Rodríguez-Rabadán, Sergio Lafuente-Arroyo, Saturnino Maldonado-BascÃ3n, Carlos Gutiérrez Álvarez, Roberto J. LÃ3pez-Sastre

PDF

AI summary

Key figure (auto-extracted from paper)
A two-layer topological framework combining LiDAR structure and sequence-based visual matching enables robust, error-free indoor localization and automatic lifelong map updates in dynamic environments.
Topological localization Visual place recognition Dynamic environments Lifelong mapping Robot autonomy Scene graph

Problem

Visual place recognition degrades in large-scale indoor settings due to perceptual aliasing from structural symmetry, ambiguous open areas, and dynamic environmental changes that break static maps.

Approach

The system builds a two-layer scene graph that merges LiDAR-derived structural layouts with RGB sequence descriptors, using edge-weight attenuation to detect topological changes and trigger autonomous map updates.

Key results

  • Zero localization error across diverse real-world conditions
  • Automatic detection and mapping of dynamic topological changes
  • Scalable performance in large, symmetric environments
  • SuperGlue maximizes accuracy while SIFT offers optimal cost-accuracy trade-off

Why it matters

Enables reliable, lifelong robot autonomy in complex public spaces without manual map maintenance.

Abstract

Visual place recognition in large-scale, indoor en- vironments often suffers from perceptual aliasing due to struc- tural symmetries and dynamic changes. This work presents a robust hierarchical topological mapping framework designed for long-term robot autonomy. Our system integrates multi- modal data (including 2D LiDAR, odometry, and RGB imagery) into a two-layer architecture. First, a Layout Layer is designed to capture the geometric structure of the environment. Then, a Visual Layer is used to encode image sequences. A key con- tribution is the dynamic map maintenance mechanism, which monitors the attenuation of edge weights to detect environmen- tal transitions, such as the opening or closing of doors. This allows for seamless lifelong updates without human intervention in large-scale environments. We evaluate our approach using various visual descriptors (e.g. SuperGlue, Patch-NetVLAD, and SeqVLAD) within a sequence-based matching pipeline. Experimental results in a 750 m2 real-world facility demonstrate that the proposed method achieves high discrimination and scalability, even in challenging open areas and symmetric corridors. This framework provides a reliable solution for assistive robotics navigating complex, evolving public spaces.

Index terms

Localization Vision-Based Navigation Mapping

Related papers