AI Eye in the Sky Improves Artillery Fire Missions

In artillery fire missions, binocular-equipped forward observers request multiple shots to close in on a target’s coordinates. In the minutes between these bracketing rounds, the target may move and the observer remains in harm’s way. An SEI artificial intelligence prototype is giving U.S. Army artillery observers a smart eye in the sky.

The Army has been experimenting with off-the-shelf unmanned aerial systems (UASs) to provide combat units a birds’ eye view of enemy positions. The Army Artificial Intelligence Integration Center (AI2C) last year tasked the SEI with developing Shrike, a complementary prototype for recognizing and geolocating threats.

From its elevated vantage, the Shrike system uses computer vision and machine learning (ML) to identify potential threats. The system sends its aerial views, marked with potential targets and locations, to the forward observer’s device. If the soldiers fire, Shrike’s ML model detects the round’s hit location and recommends targeting corrections. Shrike’s sensors feed new data to its tactical ML pipeline for model retraining so the system gets smarter with every sortie.

“Shrike is about improving small-unit capability on the front line,” said Jeff Mattson, the SEI’s Shrike project lead. The project draws on areas of SEI expertise: ML in edge environments with memory and connectivity constraints, as well as integration into a package that is easily deployable on most UAS platforms. These research areas can inform Army fires modernization efforts.

Shrike enables any soldier to accurately find, fix, and finish a target through digital fires with just minutes of training.

Captain Jarek Ingros

Perception Team Lead for Robotics and Autonomous Systems, U.S. Army Artificial Intelligence Integration Center

Mattson conducted early field trials of Shrike with the 198th Infantry Brigade, 1st-19th Infantry Battalion’s Infantry Mortar Leaders Course (IMLC) and the Army Test and Evaluation Center (ATEC). Using Shrike, operators decreased the time from the first round to the effective round by 80 percent and improved accuracy by 50 percent, when compared to the Expert Infantryman’s Badge (EIB) call-for-fire standards.

“Shrike enables any soldier to accurately find, fix, and finish a target through digital fires with just minutes of training,” said Captain Jarek Ingros, AI2C’s perception team lead for robotics and autonomous systems. “This capability drastically increases the lethality of U.S. Army formations.”

Photos: (left) U.S. Army, SPC Eric Cerami; (bottom right) U.S. Army, CPT Avery Austin