CompSLAM: Complementary Hierarchical Multi-Modal Localization and Mapping for Robot Autonomy in Underground Environments

TL;DR

CompSLAM addresses robust localization and mapping in GPS-denied underground environments by fusing visual, thermal, depth, inertial, and kinodynamic data in a hierarchical coarse-to-fine framework. The method comprises two main modules: VTIO for visual/thermal/inertial fusion and LIO for LiDAR-inertial odometry, both complemented by degeneracy-aware checks, factor-graph smoothing, and map management, plus inter-robot sharing and learning-based enhancements. The paper provides a comprehensive dataset from the SubT finals and releases code, enabling multi-robot mapping and long-term deployment in challenging settings. The results demonstrate resilience to sensor degradation and self-similar geometries, with practical impact for underground autonomy and collaborative robotics.

Abstract

Robot autonomy in unknown, GPS-denied, and complex underground environments requires real-time, robust, and accurate onboard pose estimation and mapping for reliable operations. This becomes particularly challenging in perception-degraded subterranean conditions under harsh environmental factors, including darkness, dust, and geometrically self-similar structures. This paper details CompSLAM, a highly resilient and hierarchical multi-modal localization and mapping framework designed to address these challenges. Its flexible architecture achieves resilience through redundancy by leveraging the complementary nature of pose estimates derived from diverse sensor modalities. Developed during the DARPA Subterranean Challenge, CompSLAM was successfully deployed on all aerial, legged, and wheeled robots of Team Cerberus during their competition-winning final run. Furthermore, it has proven to be a reliable odometry and mapping solution in various subsequent projects, with extensions enabling multi-robot map sharing for marsupial robotic deployments and collaborative mapping. This paper also introduces a comprehensive dataset acquired by a manually teleoperated quadrupedal robot, covering a significant portion of the DARPA Subterranean Challenge finals course. This dataset evaluates CompSLAM's robustness to sensor degradations as the robot traverses 740 meters in an environment characterized by highly variable geometries and demanding lighting conditions. The CompSLAM code and the DARPA SubT Finals dataset are made publicly available for the benefit of the robotics community

Figures (9)

• Figure 1: Narrow passage, traversed by the legged robot (location 4 in Figure \ref{['fig:dataset']}). Left: Registered LiDAR pointcloud (white) and CompSLAM map (color), right: Corresponding images captured by the robot's onboard RGB camera.
• Figure 2: Overview figure of the proposed complementary multi-modal SLAM system showcasing different sensor inputs and the hierarchal refinement of robot odometry for reliable pose and map estimation.the different modules and passing of the estimated odometry as the initial guess of the next module.
• Figure 3: Tracked CompSLAM-VTIO image features and overlay of LiDAR points projected into the image frame. Features with purple-colored bounding boxes have been initialized using LiDAR depth.
• Figure 4: Highlighting different ranges of map used for scan-to-submap registration as compared to map update utilizing full range of the pointcloud.
• Figure 5: Sequence captured in the DARPA SubT final circuit environment, comprising tunnel, urban, and cave domains and featuring various geometries, such as self-similar tunnels, narrow passages, and large natural caves. The sample views are taken from the robot's onboard RGB stream. Corresponding locations and the approximate robot path are indicated on the ground truth point cloud map (GT: https://github.com/subtchallenge/systems_finals_ground_truth).