Informed Autonomous Exploration of Subterranean Environments

Autonomous exploration is highly challenging in subterranean applications due to the constraints imposed by the nature of the environments (e.g., dead-end branches, unstructured regions, narrow passages and bifurcations). Robots need to constantly balance their exploration objectives with measures to ensure safety. We present an informed exploration approach to address these challenges, which exploits a reachability graph to represent the environment's structure and adaptive navigation to find collision-free motions. Our system makes the inspection task tractable and maximizes the information acquired about the environment while preserving safety. We evaluate our navigation and exploration techniques against several challenging cave scenarios reconstructed using real data. Our experimental results demonstrate that our method enables the robot to make informed decisions and perform exploration more efficiently than existing techniques in confined spaces.