NASA has selected Charles River Analytics and the Worcester Polytechnic Institute (WPI) to advance autonomous swarming efforts for robots exploring the Moon.
Under the Autonomous Swarming for Teams of Exploration Robots (ASTER) project, a team of Charles River scientists are developing artificial intelligence (AI), aiming to enable lunar robots to work together.
Developed as part of the $750,000 agreement with NASA, the AI technology design is inspired by a psychological principle called “theory of mind,” which describes the ability to recognize differences between self and others.
Charles River assert that the ability for robots to work together provides NASA with new ways of studying the lunar surface. With an individual robot, exploration is limited to one point at a single time and is constrained by the health of the robot and its sensors. With multiple robots, each robot can spread out to read signals at a distance, using tools such as ground-penetrating radar, or they can explore and collect samples across a broad area.
Robots update their knowledge of the lunar surface based on what they each perceive and what they hear from other robots. This means that each robot may not have a complete picture at all times, but their shared knowledge allows them to estimate updates that are useful for allocating tasks until they can acquire new, actual information.
Heterogeneous robots introduce the idea that each robot may have a unique set of skills. For example, a fast robot with a sensor can find likely mineral deposits over a large area, and slow robots that are built for digging and collecting samples can follow up at specific locations.
ASTER extends the notion of heterogeneity to include the health and status of robots, including factors such as a robot’s battery charge or degradation of its wheels or sensors. The algorithms that the team is developing consider all of these factors when deciding which robot is best suited to perform tasks that are required to support the mission.
Charles River state that autonomous coordination is especially critical when the robots are far from human command and control centers, and communications systems are limited or delayed by long distances. This challenge will grow as NASA continues to explore the Moon, Mars, and beyond.
Previously, Charles River researchers completed a feasibility study and developed algorithms for autonomous task allocation as robots’ health and status change during their mission.
Now, in Phase II, the researchers are enhancing their work to explore challenging issues with task allocation so they can perfect these algorithms and test them on real-world robots. The work builds on the Collaborative Autonomy Framework that Charles River Analytics has developed to compare and evaluate new ideas in multirobot task allocation and coordination.
Charles River partnered with Dr. Carlo Pinciroli of WPI’s Novel Engineering for Swarm Technologies (NEST) Lab for this effort.
Dr. Spencer Lynn, Senior Scientist at Charles River Analytics and Principal Investigator on the ASTER effort, commented; “Robots can determine what to do next based on their beliefs about what the rest of the swarm is doing.”
David Koelle, Director of Engineering and Swarm Cognition Lead of ASTER, added; “Collaborative autonomy in space opens the door to so many opportunities for exploring our solar system. With multiple small robots, we can send specialized robots that are equipped for different environments, and we can replace or add to robots that are already on the planets or moons.”
