The AlphaDogfight Trials carried out by the U.S. military looked to demonstrate the “feasibility of developing effective, intelligent autonomous agents capable of defeating adversary aircraft in a dogfight.”
Heron Systems is the defense contractor that developed the AI system. The AI that won is named Falcon that was able to defeat a human pilot that used the callsign ‘Banger’. The human pilot was an Air National Guardsmen pilot in Washington D.C. who also recently graduated from the Air Force Weapons School F-16 Weapons Instructor Course and has more than 2,000 hours of live F-16 flight experience. The AI won 5-0 in virtual combat one-vs-one.
Before Falcon faced off against a human opponent, it defeated seven other team’s AI opponents which include Lockheed Martin, Aurora Flight Sciences, Georgia Tech Research Institute, Perspecta Labs, EpiSys Science, SoarTech, and PhysicsAI.
“We’ve gotten an opportunity to watch AI come of age [against] a very credible adversary in the human pilot,” Program Manager in DARPA’s Strategic Technology Office Colonel Dan Javorsek said.
“The AlphaDogfight Trials is all about increasing trust in AI. If the champion AI earns the respect of an F-16 pilot, we’ll have come one step closer to achieving effective human-machine teaming in air combat.”
The Heron aircraft relies on a technique named deep reinforcement learning that allows the program to run test-runs repeatedly that calculates multiple solutions to a problem it receives and then it studies what will work and what won’t.
The AI pilot Falcon adopted twisting methods that made it difficult for Banger to keep up with. After Banger has been shot down the first four times, he reached speeds of over 500 mph and then dove the jet down to 13,000 feet. However, the human F-16 pilot wasn’t able to land a single hit because of the AI algorithm’s “superhuman aiming ability.”
The human pilot talked about the fight and said, “Some of the rules and constraints that we normally apply to our human training environment weren’t there. So you saw the AI maneuvering to a position of advantage where it was able to use its more refined aiming technique with perfect information… Outside of that, it was very similar to what we see in our ordinary training in simulators”.
The human pilot saw the potential of AI algorithms. “If I were to walk away from today saying I don’t trust the AI’s ability to perform fine motor movement and achieve kills, I’d have a lack of integrity”, he added.
DARPA has plans to convert the AlphaDogfight Trials training into the real world.
“There’s a long way to go. This was a far cry from going out in an F-16 and flying actual basic fighter maneuvers. But I think we made a really large step, a giant leap if you will, in the direction we’re going,’ DARPA’s Justin Mock said.
Kyle James Lee – The AEGIS Alliance – This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.