The US Air Force Research Laboratory (AFRL) is turning to artificial intelligence to help keep satellites healthy in orbit, exploring a system designed to spot early warning signs of failure and explain exactly how it reached its conclusions.
Under a Cooperative Research and Development Agreement, AFRL will adapt PiLogic’s “Exact AI” to analyze spacecraft electrical subsystems, helping engineers better understand system behavior and improve satellite reliability.
Unlike conventional “black box” AI systems, Exact AI uses probabilistic inference to deliver transparent, mathematically grounded reasoning with measurable confidence levels.
The company said this approach produces precise results while requiring significantly less computing power than traditional large language models.
For its part, AFRL will provide access to its satellite testing platform, allowing PiLogic to validate and refine the AI system under operationally relevant conditions.
“This collaboration accelerates our ability to make satellite systems smarter, safer, and more predictable,” said Johannes Waldstein, chief executive at PiLogic.
“Space missions demand both reliability and transparency. Our technology gives engineers a clear, explainable understanding of system behavior, essential not just for defense, but for the rapidly expanding commercial satellite economy.”
Autonomy in Focus
AFRL sees the initiative as a step toward higher levels of autonomy in future satellite systems.
Joseph Melville, Satellite Autonomy Lead at AFRL, said the goal is to build “first-of-their-kind, highly autonomous satellite systems” that are transparent at their core and can be trusted by human operators.
“We at the Air Force Research Laboratory, Small Satellite Portfolio, are excited to evaluate the next generation of autonomy for satellite health monitoring with true causal understanding,” he added.
