SALT LAKE CITY—Senior DoD space officials are looking to add more autonomy to their missions, but the lack of in-orbit exercises makes it difficult to know where it would be the most valuable.
Autonomy is becoming more and more necessary for space, as a limited number of operators must manage a growing number of sats in an increasingly contested environment. Delegating some tasks to computing is also critical to some next-gen space missions—including rendezvous and proximity operations, space domain awareness, and deep-space missions—where a communications lag makes it impractical to keep a human in the loop.
Fly high: Just look at aircraft, according to Col. Owen Stephens, the director of contracting at the Space Force’s Space Rapid Capabilities Office. By flying real-world exercises and scenarios, pilots have been able to figure out how their planes respond to a certain set of conditions—and determine which tasks can and should be automated.
Jet fuel, however, is abundant. For spacecraft operators who are limited by what they can carry to orbit, that’s not the case—leaving Guardians conserving fuel instead of practicing on-orbit.
“We have a dynamic going on with the Space Force in space right now in that we’re scared to use our satellites,” Stephens said at a panel on the sidelines of SmallSat, hosted by Redwire. “We really need to enable the ability for the Space Force to get relieved of that psychological burden and actually start using their machinery so they can learn from it…and figure out ‘Oh, this would be a really good thing to automate.’”
Pros and cons: Officials spent much of the panel expounding on the pros of autonomy in orbit—including freeing up people to take on other tasks, and enabling AI to respond faster than humans can react.
However, autonomy also presents a few challenges:
- Scaring operators with warnings, in some cases prompting people to fight against the autonomy or turn it off (think about when your car dashboard flashes red and steps on the brakes, Stephens said).
- Ensuring people don’t abandon autonomy completely, when it inevitably makes mistakes.
- A lack of data for things like contested space operations, making it impossible to train an autonomous model.
