Detecting Grasping Sites in a Martian Lava Tube: Multi-Stage Perception Trade Study for ReachBot

This paper presents a trade study analysis to design and evaluate the perception system architecture for ReachBot. ReachBot is a novel robotic concept that uses grippers at the end of deployable booms for navigation of rough terrain such as caves and lava tubes. Previous studies on ReachBot have discussed the overall robot design, number of deployable booms, and gripper mechanism design; however, analysis of the perception and sensing system remains underdeveloped. Because ReachBot can extend and interact with terrain over long distances on the order of several meters, a robust perception and sensing strategy is crucial to identify grasping locations and enable fully autonomous operation. This trade study focuses on developing the perception trade space and realizing such perception capabilities for a physical prototype. This work includes analysis of: (1) multiple-range sensing strategies for ReachBot, (2) sensor technologies for subsurface climbing robotics, (3) criteria for sensor evaluation, (4) positions and modalities of sensors on ReachBot, and (5) map representations of grasping locations. From our analysis, we identify the overall perception strategy and sensor modality configuration for a fully-instrumented case study mission to a martian lava tube: LiDAR and radar in the body, with a 3D camera at the distal end of each boom. We also identify sensors conducive to benchtop testing and further prototype hardware development.