Mapping Terrain Properties and Traversal Risk by Walking

Robotic exploration plays a crucial role in expanding our understanding of the surface of Earth and other planetary bodies. In situ rover missions choose their traverses conservatively due to a lack of methods to effectively assess loose regolith strength and resulting locomotion risk, limiting their ability to explore sites of interest and resulting in missed scientific opportunities. Thus, we propose using legged robotic "scouts" to measure the surface properties of areas deemed too risky for rovers to traverse without a priori assurance of success. Recent work has demonstrated the capability of a direct-drive robotic leg to accurately estimate the substrate strength and rheology using only proprioceptive measurements from the joints. Here we expand on this capability to a locomoting quadrupedal robot, to measure substrate properties from every step and enable dense measurements (<20 cm between steps) along transects covering 2D areas. By integrating this unprecedented sample density with a Gaussian process regression, the legged scout can create terrain property maps and corresponding traversal risk landscapes that can be used by other rovers for planning and decision-making, widening our scientific reach on other worlds.