Autonomous Cave Surveying With an Aerial Robot

This article presents a method for cave surveying in total darkness using an autonomous aerial vehicle equipped with depth camera mapping, downward-facing state estimation, and forward downward lights. Traditional methods of are labor-intensive dangerous due to the risk injury when operating darkness, potential structural instability subterranean environment. Although these dangers can be mitigated by deploying robots map passages voids, real-time feedback is often needed operate safely efficiently. Few state-of-the-art, high-resolution perceptual modeling techniques attempt reduce their high bandwidth requirements work well low communication channels. To bridge this gap state-of- the-art, compactly represents sensor observations as Gaussian mixture models maintains local occupancy grid motion planner that greedily maximizes information-theoretic objective function. The approach accommodates both limited field view (FoV) cameras larger FoV LiDAR sensors extensively evaluated long duration simulations on embedded PC. An system leveraged demonstrate repeatability flight arena effects dropouts. Finally, deployed Laurel Caverns, commercially owned operated southwestern Pennsylvania, USA, wild West Virginia, USA. Videos simulation hardware results available at https://youtu.be/iwi3p7IENjE xmlns:xlink="http://www.w3.org/1999/xlink">https://youtu.be/H8MdtJ5VhyU .