On the NRO’s wish list: AI technologies to manage satellites and data

Responsive image

| 2022-11-16 19:57:24

RESTON, Va. — The National Reconnaissance Office continues to borrow pages from the space industry’s playbook as it seeks to accelerate deployments of spy satellites, the agency’s director Chris Scolese said Nov. 15. 

“We are taking seriously the need to move faster in all the things that we do,” Scolese told a large audience of executives at an Intelligence and National Security Alliance dinner event. 

The NRO designs, builds and operates the nation’s spy satellites.

Scolese said the NRO today can take satellites from the drawing board to the launch pad in less than three years. It is launching satellites on multiple commercial rockets in the U.S. and overseas

The next innovation the agency has set its sights on is the use of artificial intelligence and machine learning to orchestrate the operation of imaging satellites and to analyze data in orbit. 

The ability to process information aboard satellites instead of having to send it back to Earth is an emerging capability in the geospatial imaging industry that Scolese said could be a game changer and wants the NRO to incorporate in its satellite architecture. 

Scolese said the NRO also is looking at the use of machine learning to command satellites and respond rapidly to mission requests, a technology that is being pursued by the U.S. Air Force and the Defense Advanced Research Projects Agency.

In a fireside chat with former director of the National Geospatial-Intelligence Agency Letitia Long, Scolese said AI and machine learning are key areas where he could use more help from the private sector. 

U.S. intelligence agencies are very adept at downloading imagery from satellites and analyzing it on cloud computing systems. “The challenge now is moving those capabilities into space where you can do automatic feature recognition and automatic target recognition,” Scolese said.

In-space analytics would allow the NRO to deliver critical intelligence to users on the ground faster “by reducing the amount of data that comes down to just what’s needed,” he said. “And of course, figuring out just what’s needed is a challenge that requires us to work with our partners to figure that out as you move that into space.” 

For example, if military units in the field need imagery over a particular area of the world, AI and machine learning could be used to instantly determine which satellite is located over the target area and whether that satellite has the right type of sensor. 

“If you want an image of Kherson [Ukraine], if it’s cloudy, it’s gotta be radar … So we’ll be using AI and ML to address those kinds of things,” said Scolese.

“We need help with that,” he added. “Those are hard problems, and it requires a change in our mindset.”

SpaceNews