This week, Planetary Systems AI demonstrated powerful new AI tools for identifying adversary spacecraft, analyzing their behavior, and helping U.S. Space Force (USSF) Guardians decide how to respond.
Our engineering team showed off this system of AI-orchestrated agents to USSF commanders and software integrators at Space Systems Command’s SDA TAP Lab in Colorado Springs at Demo Day for Cohort 7 of the Apollo Accelerator.
These tools are clustered around a new agentic artificial intelligence system that melds multiple data sources into a clear, realtime picture of spacecraft activity – not by merely categorizing maneuvers, but also by revealing them in geopolitical and historical context.
Simply put, PSAI’s system can identify sensitive points of interest or geopolitical threat that adversary spacecraft leverage for surveillance to provide insights and intelligence.
By building a record of these observation targets, our system can then spot similar maneuvers in future and identify when and where spacecraft might be looking earthward ahead of potential armed conflict or other coordinated orbital maneuvers.
The system is designed to dispatch a fleet of software agents that fetch, process, and analyze various data sources whenever the lab’s prototype battle management system – known as Welder’s Arc – detects an unexpected maneuver by an adversary’s spacecraft.
The agents:
- identify the spacecraft’s capabilities from publicly-available information
- calculate from its earth-facing camera specifications and orbit path what it might be able to see
- fetch a list of likely observation targets in the view path, including:
- lat/long data on sensitive installations such as dams, military bases, or power plants; or
- news articles about current or recent geopolitical events
- suggest theories about the adversary’s intent – and recommend strategies for allied response
The system is meant to function as a sort of co-pilot for U.S. Space Force operators, who can modify the agents’ behavior to investigate other facets of each maneuver and then validate or correct the system’s recommendations which trains the prediction model to constantly improve its performance. And it can transform the sequential behavior of Welder’s Arc’s dozens of analytical applications into a living, responsive system for identifying, predicting, and responding to hidden, unknown, and hostile activity by adversary spacecraft.
Architecture like this can be configured to address many more space-domain awareness problems that require swift and complex analysis of multiple varied data sources to feed rapid decision recommendations to USSF Guardians – shortening analytical processes that often take hours and sometimes days, and reducing the risk of operational surprise.
In another demonstration, we showed how our AI agents can monitor news sources for indications of pre-launch activities and launch of satellites by U.S. adversaries.
This news-analysis agent is connected to our existing capability-analysis tool, which aggregates publicly-available information from many sources into specification profiles of just-launched satellites. This lets us link launch events to specific satellites before they enter orbit, and can help reduce the workload on other systems that labor to associate uncorrelated satellite tracks after launch to specific vehicles.
The team also demonstrated an expansion of our computer vision model for identifying satellite capabilities from orbital images and clean-room photos.
We are working to overcome a significant challenge to using AI to identify adversary spacecraft: Very few images of spacecraft exist that can be used to train computer-vision systems. Instead, we have begun training our system on synthetic visual data provided by a major defense industry partner within the lab. This allows us to train the model on thousands of synthetic satellite images for better accuracy, where only a few dozen publicly-available photographs of orbiting satellites exist.
Our work is timely: In an effort to keep the U.S. competitive in the global space industry, the White House this week issued an executive order meant to reform contract regulations and streamline the review process for space companies vying for federal business. At its core:
It is the policy of the United States to enhance American greatness in space by enabling a competitive launch marketplace and substantially increasing commercial space launch cadence and novel space activities by 2030. To accomplish this, the Federal Government will streamline commercial license and permit approvals for United States-based operators.
During remarks at the start of Demo Day, USSF Colonel Jason West confirmed this new direction, saying that the Space Force is working to overhaul the existing defense-contracting environment – which usually sees software developed at a slow and deliberate pace by a limited number of major contractors – with a more agile process that will favor fast-moving innovation from smaller companies.
With our success at the SDA TAP Lab and our growing number of small and large space- and defense-industry partners, PSAI sits in a great position to ride this next wave of innovation. We welcome the change, and look forward to what comes next.
Watch this space.