Pentagon’s Golden Dome puts AI at the center of missile defense architecture

 

 SpaceNews

“The human is the weak point in the loop,” Biltgen said. In some scenarios, “full auto mode” may be necessary…

As the Department of Defense works to define the architecture of Golden Dome, industry leaders are emphasizing the essential role artificial intelligence will play in managing its complexity and speed.

Speaking during a SpaceNews web event, Patrick Biltgen, vice president of space AI at Booz Allen Hamilton, said the system’s requirements remain undefined. “What is the point of Golden Dome? Is it to protect the homeland or is it to provide tactical capabilities or does it also do hypersonics over the ocean?” he asked. The answer to that question, he said, has implications for scope, cost, timelines and technical demands.

Rob Mitrevski, president of Golden Dome strategy and integration at L3Harris Technologies, said command and control is “certainly the most complex part” of the architecture. “We’ve moved from sort of the modeling and sim discussion to the decision support discussion in real operations.” He said AI will be critical for rapidly matching threats to interceptors. “Particularly in trying to prioritize which targets with which interceptors at which sequence of events need to happen in a very short time frame.” He added that AI’s role becomes even more important when dealing with hypersonic weapons. “When you’re talking hypersonics, it becomes increasingly important because of the lower latency, because drones are slow and intercontinental ballistic missiles are fast, but hypersonics in their trajectory are very difficult targets.”

Biltgen noted that integrating legacy and new systems poses major challenges. “It’s very hard to prescribe data and integration standards and demand that everyone integrate with them,” he said. AI models, he added, are now capable of translating between standards and even generating their own services to enable interoperability.

He said operational tempo is driving the need for greater automation. “The human is the weak point in the loop,” Biltgen said. In some scenarios, “full auto mode” may be necessary, provided it undergoes proper verification and validation.

Security is another concern. He described Golden Dome as a system built on ambiguity. “Are the interceptors and sensors in the right place where they can actually intercept? And if you give the adversary enough ambiguity about their success, then they can’t attack you.”

He also pointed to workforce challenges. “The government may need to train their own unique secret models,” he said, which will require new pipelines for AI talent.

Dan Knight, vice president of sensors and data integration at Arcfield, said that many of the needed data streams already exist. “We have the information or we have the data that we need. It’s just not in the right places.” He said model-based systems engineering will be essential to organize existing infrastructure and build from there.

Mitrevski added that prioritization from government stakeholders is critical. “It is a very complex system, and the things that are ready and easy should be done, because we do need to focus on the things that are difficult to see.”

Biltgen acknowledged that some remain uneasy with the idea of autonomous missile interceptors. “Some people are really weirded out by the idea that these interceptors might be AI enabled, like you’re going to have these angry space robots that are just going to chase things and intercept them.”