Robust Lifelong Indoor LiDAR Localization using the Area Graph

TL;DR

This work tackles lifelong indoor LiDAR localization by leveraging the Area Graph, a compact topometric semantic map that encodes walls and doors. The method first global-localizes against the Area Graph using a clutter-free 2D LiDAR projection and WiFi/barometer priors, then tracks pose with a weighted point-to-line ICP that ignores clutter and emphasizes architectural features. Key contributions include clutter removal subsampling, a tailored weight function for ICP, and corridorness-based downsampling to improve corridor localization, validated in cluttered rooms and long corridors with competitive performance relative to 2D AMCL and 3D LeGO-LOAM baselines. The approach enables robust, drift-free localization in lifelong deployments without relying on dense clutter maps or IMU, offering practical benefits for campus-scale autonomous navigation and navigation stack integration.

Abstract

Lifelong indoor localization in a given map is the basis for navigation of autonomous mobile robots. In this letter, we address the problem of robust localization in cluttered indoor environments like office spaces and corridors using 3D LiDAR point clouds in a given Area Graph, which is a hierarchical, topometric semantic map representation that uses polygons to demark areas such as rooms, corridors or buildings. This representation is very compact, can represent different floors of buildings through its hierarchy and provides semantic information that helps with localization, like poses of doors and glass. In contrast to this, commonly used map representations, such as occupancy grid maps or point clouds, lack these features and require frequent updates in response to environmental changes (e.g. moved furniture), unlike our approach, which matches against lifelong architectural features such as walls and doors. For that we apply filtering to remove clutter from the 3D input point cloud and then employ further scoring and weight functions for localization. Given a broad initial guess from WiFi localization, our experiments show that our global localization and the weighted point to line ICP pose tracking perform very well, even when compared to localization and SLAM algorithms that use the current, feature-rich cluttered map for localization.

Figures (9)

• Figure 1: Open Street Map Area Graph maps for the two experimental datasets Seq01 (top) and Seq02. We can see areas demarked by pink polygons and passages (doors) as red lines, that are connections between areas and semantic information in terms of room names and passages. Yellow dots are polygon coordinates which form areas and passages.
• Figure 2: The overall framework of AGLoc.
• Figure 3: Illustration of Area Graph polygons and signed distance $sd_j$ intersection calculation.
• Figure 4: Weight function $\mathcal{W}(sd_j)$, computing a weight for the weighted point to line ICP based on the minimum signed distance (positive values behind the wall/ outside) between a point and the line its beam is intersecting with. Mostly ignoring clutter inside rooms (early cut-off of negative $sd_j$) and also ignoring long error beams, e.g. due to reflection.
• Figure 5: Illustration of problematic ICP performance in corridor-like environments. Only few (2) points match against the vertical line, but many on horizontal lines, fixing the translation in the longitudinal direction in place, preventing the correct solution. Our corridorness downsampling reduces the number of points on lines in the dominant orientation, mitigating this problem.