Answers are grounded in this paper and cite their sources.
Simulation study demonstrates modular robotic exploration and artifact localization across subterranean environments, indicating high urban accuracy but degraded coverage in steep, cluttered terrain.
Subterranean environments such as caves, tunnels, and underground urban structures present significant challenges for search and rescue operations due to hazards like unstable terrain, limited visibility, and the absence of GPS signals. To address these challenges, autonomous mobile robots have emerged as a viable solution. This paper presents a modular autonomy stack for autonomous underground exploration, object detection and localization in simulated hazardous subterranean environments. The system integrates three core modules: LIO-SAM for lidar-IMU odometry, GBPlanner2 as the exploration planner with a pure-pursuit tracker for path following, and YOLOv3 for real-time artifact detection with depth-based localization. We evaluate the integrated stack in multiple virtual circuits (cave, tunnel, and urban) and report circuit-wise performance, including odometry drift, exploration coverage, and detection errors. The system performs best in the urban circuit, where it localizes multiple artifacts with low horizontal drift. Performance degrades in steep or cluttered terrain, where limited observability and collision-checking errors reduce coverage and increase failure rates.
No takes yet. Share an insight, caveat, or question.
Bagot et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: