WASHINGTON — Space companies are finding new ways to exploit artificial intelligence for commercial and national security applications, executives said Feb. 8 at the SmallSat Symposium in Mountain View, California.
“Whether you’re running a constellation or operating any space system, AI is paramount,” said Alvaro Alonso Ruiz, co-founder of Leanspace, a French startup building a cloud infrastructure to run space missions.
AI and machine learning technologies, for example, allow operators of remote-sensing constellations to plan satellite operations, run simulations and identify what resources are needed where, he said. “Optimizing resources is the real breakthrough … for how can we apply these technologies to build more sustainable businesses in space.”
Kameron Baumgardner, chief technology officer of New Mexico-based startup RS21, said there is growing interest in using machine learning for autonomous space operations. The company last year won a U.S. Space Force contract to research the use of AI to predict satellite failures in orbit.
The technology will be tested in an upcoming Space Test Program experiment, STPSat-7, projected to launch in 2023 to low Earth orbit. RS21 is developing software that will be integrated in the STP ground station at NASA’s Johnson Space Center, Houston, prior to launch. The data will be used to train the AI model in preparation for the STPSat-7 experiment.
A key challenge in AI applications is to develop a “robust data analytics architecture,” Baumgardner said. In military programs, information security is paramount. In model-based decisions, whether it’s on orbit or on the ground, “we have to make sure that that model is robust against perturbations, especially in environments where we’re dealing with adaptive adversaries.”
RS21 is expanding the use of AI for prognostics and predictive logistics, “so you know 24 or 48 hours ahead of time whether or not you’ll have a satellite failure.”
AI also is helping startups that are designing satellites and planning their businesses, said Jeromy Grimmett, founder and CEO of Rogue Space Systems, based in New Hampshire. The company is developing small satellites for in-space servicing.
“We use AI and simulations quite a bit in our development, as we try to refine our designs,” he said. “I think it’s going to be really interesting to apply some of that technology to the economics of space and doing analysis, so we can better engineer the delivery of services and capabilities.”
Wargaming and training
Colorado-based startup True Anomaly is using AI to build models for space wargaming and training, said the company co-founder and CEO Even Rogers
True Anomaly, which recently came out of stealth mode, is developing an “autonomous orbital pursuit vehicle” aimed at military applications. According to the company’s website, “AI pilots help human operators plan complex no-fail missions and execute them with multiple spacecraft.”
Rogers said the U.S. Space Force faces an “interesting historical conundrum” because it has to prepare to defend assets in space, a type of conflict that has never previously been fought
That’s where AI becomes really useful, he said, “to help build the knowledge that’s necessary to make good design decisions … and understand how our systems perform in space operations environments.”
AI also can help develop realistic warfighting simulations, Rogers said. The U.S. Space Force has to train to fight against a “thinking adversary, and one of the ways you can do that at scale is by training adversarial agents.”
“It’s a critical problem the Space Force and the space defense community face when it comes to developing systems and tactics to respond to a conflict that we’ve never seen,” said Rogers. “The Ukrainian conflict has blown that open and made it very clear that space assets are now a legitimate target.”
In the private sector that supports national security, he said, “we have a responsibility to develop technologies that allow us to defend ourselves in the event that our spacecraft are targeted.”