Infrastructure Robotics: a Technology Enabler for Lunar In-situ Resource Utilization, Habitat Construction and Maintenance

NASA’s planned permanent return to the Moon by the year 2018 will demand advances in many technologies. Just as those pioneers who built a homestead in North America from abroad, it will be necessary to use the resources and materials available on the Moon, commonly referred to as in-situ resource utilization. This benefit would come in a number of ways. Required payloads would be smaller as supplies would be available at the mission site. This would allow for much less expensive missions as the launch fuel-to-payload weight ratio is greater than 9:1. In addition, this would free up valuable payload space for other instruments and tools, allowing more effective and higher-return missions to be undertaken. Preparation would nonetheless be required well in advance of manned missions. In this concept study we put forward a role for robotic precursor missions that would prepare for the arrival of astronauts, serving to establish methods of collecting oxygen, water and various other critical resources. It will also be of principal importance to perform site preparation to set up a power generation center necessitating excavation of trenches, foundations, radiation shielding, landing and launch sites. We explore the potential role for autonomous, multirobot excavation solutions for these infrastructure development tasks. As part of this analysis, we compare various excavation platforms using an integrated real-time 3D dynamics simulator and autonomous control techniques for lunar surface interactions. Traditional human-designed controllers lack the ability to adapt in-situ (without human intervention), particularly when faced with environmental uncertainties and changing mission priorities which were unaccounted for during design. By contrast, use of Artificial Neural Tissues (a machine learning approach) to produce autonomous controllers requires much less human supervision. These controllers only require a single global fitness function (akin to a system goal) and can perform autonomous task decomposition through a Darwinian selection process. This novel quantitative approach combining real-time 3D simulation with machine learning provides an alternative to the often disputed and unreliable qualitative predictions of terrestrial excavation solutions applied to the lunar surface. Besides an autonomous infrastructure robotics concept, we also consider traditional approaches including teleoperated single and multirobotic systems. Some of the advantages of the autonomous multirobot approach to excavation over the traditional ones are analyzed in terms of launch mass, power, efficiency, reliability, verifiability and overall mission cost.