WASHINGTON — Slingshot Aerospace developed an artificial intelligence system that monitors the behavior of satellites in orbit and identifies anomalous spacecraft within large constellations.
Slingshot, a space-tracking and data analytics firm, said the AI system, called Agatha, was developed under a Defense Advanced Research Projects Agency (DARPA) contract.
“Agatha represents a major advance in using AI for space domain awareness,” said the company’s vice president of strategy and policy Audrey Schaffer.
Last year Slingshot Aerospace revealed how it used AI to track unusual maneuvers by Russia’s Luch Olymp-K-2 inspector satellite.
Agatha is a different use of AI, Schaffer told SpaceNews. It’s optimized to find anomalous behavior amid large satellite constellations and verify that the satellites are operating as intended.
Dylan Kesler, director of data science and AI at Slingshot, said this tool can find “needles in the haystack” and perform tasks that are near impossible for human analysts.
He explained Agatha was trained on over 60 years of simulated constellation data and successfully transitioned from simulated data to real-world applications, identifying non-threatening anomalous spacecraft within operational commercial satellite networks. Agatha’s findings were validated by the respective satellite operators, said Kesler.
The AI detects anomalies such as a malfunctioning spacecraft or an intentionally nefarious ‘wolf in sheep’s clothing’ spacecraft, he said. As more companies and national governments continue to deploy large constellations, “it’s important to be able to verify that satellites are operating within the constellations’ stated purposes.”
Slingshot won a DARPA contract of undisclosed value after submitting a white paper in response to a Broad Agency Announcement. The company said it also contributed resources to the project.
Schaffer said the Agatha AI is in operation in Slingshot’s platform used by customers, and the company is in discussions with a number of U.S. government and allied organizations, and commercial satellite operators about “how we can offer Agatha as part of our advanced space situational awareness services.”
“This could be quite a powerful tool for U.S. Space Command to monitor everything happening in space,” said Schaffer.
The system would flag a potential threat, for example, if it detects that a satellite has a different mission, or a larger mass than the rest of the satellites in the constellation, which would require it to perform additional maneuvers in order to stay in the same orbit. Outliers could also be identified by their communications patterns, Schaffer explained. If a rogue satellite is eavesdropping, it could be flagged because it’s sending data to ground stations in countries that have a cooperative relationship with the owner operator.
Kesler noted that the AI does not need to be cued on what to look for. “It can detect things that we haven’t necessarily defined yet. We don’t tell it what to look for, and it figures it out.”
