Subsurface Prospecting by Planetary Drones

Recurring slope linae (RSL), such as those in Newton Crater on Mars, methane plumes in hazardous Martian terrain, and water ice discovered during the LCROSS experiment in the Moon?s permanently shadowed Cabeus Crater drive the need for a new generation of robotic explorers that access, probe, extract, and return resources from extreme terrains. These robots must possess sufficient system-level autonomy to operate without human guidance due to latency constraints over vast distances, and must also have perceptual capabilities to analyze sensor measurements and the belief state to make decisions about where to explore and whether a target is worth sampling. This enhanced exploration capability takes advantage of perceptual models that can encode the probability of the existence of a resource given material properties estimated from current and prior sensor measurements. The proposed program innovates novel perceptual models and exploration algorithms that maximize the likelihood of detecting resources if they are present and enables robots to make decisions about where to loiter in order to sample terrain for a particular resource. Beyond topical research, the program will ruggedize Phase 1 software to operate in the presence of sensor and state uncertainty, integrate the capabilities on physical robots, and demonstrate results in relevant, subterranean field test. Besides RSL and craters, the research enables exploration and access of cryovolcanoes, steep and deep gullies, and canyons. Terrestrial applications include the detection of radiation in contaminated facilities or explosive gases and flammable dust in mines, surveying urban canyons, and exploring bunkers and caves.