UAVs Beneath the Surface: Cooperative Autonomy for Subterranean Search and Rescue in DARPA SubT

Abstract

This paper presents a novel approach for autonomous cooperating UAVs in search and rescue operations in subterranean domains with complex topology. The proposed system was ranked second in the Virtual Track of the DARPA SubT Finals as part of the team CTU-CRAS-NORLAB. In contrast to the winning solution that was developed specifically for the Virtual Track, the proposed solution also proved to be a robust system for deployment onboard physical UAVs flying in the extremely harsh and confined environment of the real-world competition. The proposed approach enables fully autonomous and decentralized deployment of a UAV team with seamless simulation-toworld transfer, and proves its advantage over less mobile UGV teams in the flyable space of diverse environments. The main contributions of the paper are present in the mapping and navigation pipelines. The mapping approach employs novel map representations—SphereMap for efficient risk-aware long-distance planning, FacetMap for surface coverage, and the compressed topologicalvolumetric LTVMap for allowing multirobot cooperation under low-bandwidth communication. These representations are used in navigation together with novel methods for visibility-constrained informed search in a general 3D environment with no assumptions about the environment structure, while balancing deep exploration with sensor-coverage exploitation. The proposed solution also includes a visual-perception pipeline for on-board detection and localization of objects of interest in four RGB stream at 5 Hz each without a dedicated GPU. Apart from participation in the DARPA SubT, the performance of the UAV system is supported by extensive experimental verification in diverse environments with both qualitative and quantitative evaluation.

SphereMap: Dynamic Multi-Layer Graph Structure for Rapid Safety-Aware UAV Planning

“…The amount of misclassified space could be reduced by choosing different bounding shapes, such as polygons or 6DOF boxes, which would better approximate the segments' shapes at the expense of higher computational complexity and higher communication bandwidth. Nevertheless, for the purposes of decentralized cooperative planning, we found the 4DOF boxes to be sufficient in the DARPA SubT Challenge [22].…”

“…• a novel dynamically built graph structure (SphereMap) with clearance information that allows for rapid safetyaware path planning between any two points in known space and can handle arbitrarily structured and changing environments, • a method for quickly generating a lightweight volumetric graph representation (LTVMap) of the environment, that can be shared with other robots using low-bandwidth communication and carry mission-critical information, • experimental evaluation showing that planning on the SphereMap can be orders of magnitude faster than stateof-the-art approaches for long-distance planning, • validation of the proposed methods in real world scenarios, most importantly in the DARPA SubT competition, where robots were deployed into both real and simulated previously unvisited environments [22].…”

“…For SAR missions, additional information (such as the percentage of surfaces covered by the UAV's cameras or information about frontiers in a given region) can easily be appended to each region. This information can then be shared with other robots during the mission for cooperation purposes, and also with human operators to whom this map representation can provide a good overview of the environment and the mission progress [22].…”

“…The control and state estimation part of the deployed system is described further in [4]. The SphereMap in a more crude form (without pruning of nodes) was used also in the virtual circuit of the DARPA SubT Challenge Finals, where our team CTU-CRAS-NORLAB won the 2 nd place [22]. Lastly, the SphereMap was also tested on a large outdoor environment where a UAV autonomously explored the space around a house surrounded by heavy shrubbery, crossed a bridge into a vast open field, and safely returned home.…”

MRS Drone: A Modular Platform for Real-World Deployment of Aerial Multi-Robot Systems

“…We finished second in the virtual track, and we achieved sixth place in the competition with physical robots, where three UAVs together with six UGVs found seven artifacts in a complex confined environment. The developed approach is extensively covered by [29,68]. The platform also proved its reliability in intensive tests in intricate environments, such as vast natural caves, ruins of industrial buildings, underground fortresses, mining tunnels, and large-scale outdoor environments with dense vegetation.…”