Autonomous Cave Exploration using Aerial Robots

Abstract

In this paper we present the complete system design for an aerial robot capable of autonomous exploration inside natural cave environments. Cave networks involve diverse and complicated topologies, complex geometries and degraded conditions rendering the process of robotic mapping a particularly daunting adventure. In response to these challenges, we outline the core algorithmic modules relating to localization and mapping, exploration path planning and control, alongside the developed perception and computing solutions onboard an aerial robot tailored to undertake such complex tasks given no prior information for the cave environments in which it is deployed. A set of extensive results is presented including both simulation studies in multi-branching and maze-like cave environments, as well as field experiments inside the Moaning Caverns natural cave environment in California, US.