The Air Force Research Laboratory is betting $750,000 on a new system that Albuquerque-based data analytics firm RS21 built to prolong satellite operations through remote monitoring and intervention.
The system – dubbed the Space Prognostic AI Custodian Ecosystem, or SPAICE – applies artificial intelligence to detect problems before they arise, allowing ground operators to intervene in advance of failures to extend satellite functionality, said RS21 Chief Technology Officer Kameron Baumgardner.
SPAICE has already shown enough promise for AFRL to entirely skip its normal “Phase I” proof-of-concept seed funding and proceed to a “Direct to Phase II” small business innovation research grant. That allows RS21 to immediately build and install the new system for real-world testing on a satellite that was launched from the International Space Station earlier this year, Baumgardner said.
Once fully installed, SPAICE will provide a steady stream of data from the satellite and the space environment where it’s operating, including real-time information on everything from how components are functioning, heat levels inside the craft, and the satellite’s positioning to current weather in space, fuel-tank levels, and how charged the vehicle batteries are.
All that data will automatically combine with detailed information about the spacecraft itself, such as the manufacturing process used to build it, creating a “neural network” for instantaneous monitoring and assessment.
“SPAICE is based on deep-learning AI that mimics how the brain is structured,” Baumgardner told the Journal. “It looks at a set of inputs to understand how they correlate to identify patterns and predict things. In this case, we’re looking at the system health of the satellite.”
That can enable satellite operators to detect anomalies and take action to correct them before something goes wrong, Baumgardner said.
SPAICE provides automated, real-time monitoring, data-feed and analysis, but it allows operators to make final decisions on needed action to help them build trust in the information received.
“With AI, folks who have done everything themselves for years often have a hard time trusting algorithms to do things for them,” Baumgardner said. “SPAICE presents data in a way that adds context for decision making, but it doesn’t actually make decisions for operators.”
RS21 won first place last December in the annual “Hyperspace Challenge” managed by AFRL and the ABQid business accelerator, run CNM Ingenuity, which oversees all of Central New Mexico Community College’s commercial endeavors. That generated broad interest in the new AI system from government agencies and commercial satellite manufacturers and operators, said RS21 President and CEO Charles Rath.
“We believe our AI-powered platform will be transformational for industry to improve situational awareness of satellite operations and save money,” Rath told the Journal. “Our phone is ringing off the hook from government entities and satellite producers. … We believe the SPAICE system could become the standard for AI and machine learning for preventive maintenance across the industry.”