NASA recently conducted field tests of artificial intelligence that could make future missions to explore the Solar System more efficient.
In July 2026, at the Virginia Tech Transportation Institute test site in Blacksburg, Virginia, the ASTRA (Adaptive Sensing Technology for Responsive Autonomy) team, together with partners from Noblis, tested a fleet of three robots.
The drone selected a zone of interest based on initial targets set by humans. The system then assessed risks and scientific value, choosing which robot was best suited to the task.
The fleet continuously surveyed the environment, adapted its plan to new information and reported back to the humans. It did not abandon the original task, but could bring in an additional robot to study an unexpected find.
The goal of the project is to create a reliable extension of the human science team where direct control is impossible: on distant missions or in hazardous regions of the Moon and Mars.
The system takes into account not only objective data — distance, speed, robot availability — but also subjective priorities: how important new information is compared with the original plan and whether the risk is acceptable.
"We want missions to be able to do what they couldn't before, because there was no one on board who simultaneously weighs all these criteria," says Bethany Theiling, ASTRA project lead at Goddard Space Flight Center.
The tests showed that the fleet is capable of making decisions on its own, drawing on its own capabilities and the team's priorities.
Aurora Engineering and the University of Tulsa also took part in the tests. The project aims to have robots return more scientific data and accelerate discoveries without constant human intervention.
The term "AI swarm" (or "swarm technology", "swarm intelligence") is professional jargon, a calque of the English AI swarm or swarm robotics.
AI — artificial intelligence.
Swarm — comes from the word "swarm" (as in bees, ants, locusts or a flock of birds).
