Why AI Is a Critical Capability for the US Space Force
The Pentagon is looking for ways to leverage artificial intelligence in all branches of the military. But for the newly christened US Space Force, AI will be core to its mission from day one, says Chris Bogdan, a retired three-star USAF general who’s now a senior vice president with Booz Allen Hamilton.
“The US believed space was a sanctuary for decades,” Bogdan told Datanami in a recent interview. “We didn’t think of space necessarily as a warfighting domain. Unfortunately, our adversaries didn’t think that way, and over the last 10 to 15 years, they have create offensive and defensive capabilities that really have put at risk many of our space assets, on the military, on the intel side” and in the commercial realm.
The transition from space as a neutral zone into a contested one has spurred investments in developing new technology and new modes of operating in space to secure US interests, Bogdan explained. This is spurring change not only within the newly created US Space Force, but also the existing US Space Command and Air Force Space Command, which operates the National Space Defense Center (NSDC) at Schriever Air Force Base in Colorado.
AI and machine learning will be critical capabilities that enable commanders in US Space Force and other entities to aggregate and analyze large amounts of sensor data and make optimal decisions within short timeframes, Bogdan said.
“Let’s say an adversary has launched an anti-satellite missile from the earth,” Bogdan said. “We can detect that now. We can also pretty much tell early on what orbit it’s going into. But the next set of steps — what’s at risk, what’s the threat assessment, and what do you do about that in terms of courses of action and how you maneuver — all of that has to be done in a very, very tight timeline.”
A US space commander would have about seven to 10 minutes to react to the missile launch and issue the command to move satellites to protect them, Bogdan said. That doesn’t give them the luxury to allow analysts to pore over the data and run multiple scenarios to come up with the best decision.
“In order to do those kinds of things in that kind of timeline, you absolutely have to have data analytics, machine learning, [and] artificial intelligence to sort through all that data … to give that warfighter the best decision space that we can,” he said.
There are multiple challenges at play in the space theater. For starters, there’s the physics challenge of tracking all 21,000 or so objects orbiting the earth and predicting where they will be at any one time. Ten percent of those objects are satellites, and the rest is either debris or other objects (such as the car-size “mini moon” that telescopes have picked up recently).
No single branch of the military or government office has a perfect view of all those 21,000 objects, so one of the first things that the Department of Defense is doing is piecing together their satellite orbit “libraries” to create a unified view.
“Once you’ve decided that you’re going to have to fight combat in and from space, you really need a clear picture of what’s going on everywhere,” Bogdan said.
The problem will get even bigger in the coming years, as commercial entities plan to launch many more satellites, including large number of low-earth orbit satellites to create constellations of satellites. Bogdan said the number of objects in orbit could triple in the coming years. Tracking 60,000 objects and predicting their orbits days out poses a significant computational challenge, which is why Booz Allen Hamilton and other defense contractors are exploring the use of quantum computing to give them edge.
BAH is in the midst of building a new platform that can not only track these objects and predict when the objects are in “conjunction” (or slated to collide), but also provide a set of recommended actions as part of an end-to-end system for decision-makers operating in the new space theater.
The company has a contract with the NSDC to build a next-generation computer platform for space domain awareness, but the Pentagon has not selected BAH to provide some of the higher order applications. BAH is investing its own money in building out those capabilities in the hopes that it wins those contracts, which Bogdan said could be awarded in the next year.
AI for Space Warfighters
The advanced applications that BAH is currently building for the space warfighter make extensive use of AI and ML to assist the decision-maker in making good and fast decisions atop huge amounts of data. The approach that the space warfighter takes to making decisions isn’t dramatically different than the air, land, or sea domains, except for the physics, Bogdan said. But those physics provide for some interesting challenges.
“We do orbital analysis on these propagation vectors, and we can figure out from five to seven days ahead of time, if this satellite is maneuvering, what will now be at risk for us,” he said. “The system will show the warfighter that, and it will show him or her over that 5-7 days what things this other satellite is going to come close to, close enough to potentially affect its operations.”
Next, the system will determine what threat the other satellite poses. Is the satellite moving through its own propulsion, or is the satellite moving because it’s falling out of orbit? If it’s not falling out of orbit, then what threat does that satellite pose? Is it a “hunter killer” satellite? These questions must be answered quickly to give the decision-maker the information he or she needs to make a good decision.
“The final step in this chain is the system will be able to tell you, based on a series of collision avoidance models that we’ve built – another set of machine learning algorithms – what to do about it,” Bogdan said. “This is a very complex problem. Because, if you know an adversary satellite or debris or whatever is going to conjunct with your satellite in the next two to three days, you obviously want to protect your asset and do something. But there are a lot of constraints on what you can or what you should do.”
Just hitting the thrusters and moving the satellite comes with a set of tradeoffs. For starters, the fuel on the satellite is limited, and any non-planned use of fuel limits the lifespan of that asset. The satellite was also put in that particular orbit for a particular reason and a particular mission, Bogdan said, and if you move it out of its orbit, it may not able to complete its mission as well as it could have otherwise.
“Last if you maneuver that satellite, you may put other things downstream at risk because they’re so many different things maneuvering,” he said. “So coming up with the right course of action and how to maneuver that satellite requires a tremendous amount of data analytics and a tremendous amount of AI and ML to make sure that, one you can still protect your asset, but at the same time limit or minimize the disruption to the mission, the life of the satellite, or any downstream impacts.”
BAH is hoping to build this system and put it into action on the platform that it’s building at the NSDC. According to Bogdan, AI will be instrumental in giving the United States the capability to fulfill its new space mission.
AI Sensor Fusion
Keeping track of objects in space and where they’re going is important. But what’s more important for the decision-maker is how the information is presented, and what kinds of options are available. In BAH’s system, the decisions are presented with impact scores that weigh the downstream affects of moving the satellite and the capability of the satellite to continue its mission.
“One of the things that the warfighter cares most about when he or she is maneuvering a satellite is how much degradation in the mission for that particular satellite is going to occur if I pick the course of action 1, 2, or 3,” Bogdan said. “With our system, the warfighter can actually click on course of action one and get another set of screens that will actually show him or her what the change in orbit of satellite is and what the degradation in mission performance of that satellite is in terms of coverage, in terms of RF [radio frequency] it can pick up, or whatever that satellite does.”
It’s all about using AI and ML to enhance the decision-making capability of the human, not replacing the human’s role in making that decision, Bogdan said. It’s not unlike the “sensor fusion” capability in the F-35 Joint Strike Fighter, the huge project that Bogdan oversaw while in the Air Force. In the F-35, the computers will automatically handle some tasks, like choosing a radar mode and a direction, that the pilot was previously responsible for. Freeing pilots from these mundane tasks enable them to focus their attention on more important things.
In the space theater, decision-makers will have the ability to “see” what a satellite can see in certain cases, Bogdan said. Advances in virtual reality technology are also giving decision-makers a 3D view of the satellite’s neighborhood, much like the displays in modern jet fighters provide the pilot with beyond-visual-range capabilities.
Eventually, much of the intelligence could be pushed to the satellite itself. “The next step in the evolution of this kind of space warfare,” Bogdan said, “is actually to make your satellites smarter [by] putting some of these analytics and some of the decision making at the edge, on that satellite itself.”
NASA Upgrades Search Capabilities
Booz Allen Gives Government a Deep Learning Edge
From Big Guns to Big Data: US Navy Looks to Tech for Advantages