Active Loop Closure for OSM-guided Robotic Mapping in Large-Scale Urban Environments

TL;DR

This work tackles the challenge of large-scale urban mapping by integrating OpenStreetMap (OSM) priors with a real-time active SLAM framework. It introduces an active loop closure mechanism that re-plans GPS trajectories on the OSM graph to trigger loop closures and back-end optimization, thereby reducing pose uncertainty and improving map quality. The methodology combines global path planning via CPP and Hierholzer’s algorithm with uncertainty-aware replanning rooted in a pose-graph Laplacian and the $D\text{-}opt$ criterion, alongside terrain-aware local planning and traversability analysis. Real-world experiments across diverse urban scenarios demonstrate faster loop closures, lower online pose uncertainty, and significant translation error reductions, confirming the practical value of OSM-guided active SLAM for robust large-scale mapping. The approach offers a scalable pathway for accurate autonomous mapping in complex environments with global guidance and real-time adaptability.

Abstract

The autonomous mapping of large-scale urban scenes presents significant challenges for autonomous robots. To mitigate the challenges, global planning, such as utilizing prior GPS trajectories from OpenStreetMap (OSM), is often used to guide the autonomous navigation of robots for mapping. However, due to factors like complex terrain, unexpected body movement, and sensor noise, the uncertainty of the robot's pose estimates inevitably increases over time, ultimately leading to the failure of robotic mapping. To address this issue, we propose a novel active loop closure procedure, enabling the robot to actively re-plan the previously planned GPS trajectory. The method can guide the robot to re-visit the previous places where the loop-closure detection can be performed to trigger the back-end optimization, effectively reducing errors and uncertainties in pose estimation. The proposed active loop closure mechanism is implemented and embedded into a real-time OSM-guided robot mapping framework. Empirical results on several large-scale outdoor scenarios demonstrate its effectiveness and promising performance.

Figures (5)

• Figure 1: System overview of OSM-guided robot mapping framework.
• Figure 2: Illustration of two different graph structures constructed by two different global routes. The left graph corresponds to the OSM data. (a) and (b) are two routes obtained by the CPP solver as an example. The blue solid lines represent the edges and the blue dashed lines represent the nodes where the robot has traversed duplicates, indicating potential loop closures at these locations.
• Figure 3: The process of waypoint generation. First, the semantic local sub-map is transformed into a 2D grid map after updating the grid state. Subsequently, the road area is extracted from the free-state regions of the grid map. Morphological operations are then applied to refine the road area. Finally, the skeleton is extracted and the convex hull vertices are calculated to generate waypoints within the current perception range.
• Figure 4: Experiment comparison results for Env.5. (a), (b): w/o ALC the map is distorted in the wall area. (c) The exploration trajectory of w/ALC, where the color of the trajectory indicates the order of exploration, and the white lines represent the prior OSM graph. (d) real world satellite images. (e), (f): w/ALC explored this area with routes: 1$\rightarrow$2$\rightarrow$3$\rightarrow$5$\rightarrow$4$\rightarrow$1$\rightarrow$2$\rightarrow$3$\rightarrow$7$\rightarrow$6$\rightarrow$5$\rightarrow$3$\rightarrow$start. w/o ALC explored this area with routes: 1$\rightarrow$2$\rightarrow$1$\rightarrow$4$\rightarrow$5$\rightarrow$6$\rightarrow$7$\rightarrow$3$\rightarrow$5$\rightarrow$3$\rightarrow$start. The yellow straight line in the yellow dashed box represents the occurrence of loop closure detection.
• Figure 5: Instant uncertainty with travel distance