An AI ‘pilot’ developed by Heron Systems triumphed over a U.S.A.F top F-16 pilot in DARPA’s simulated aerial dogfight contest today.
“The AlphaDogfight Trials have concluded! Congratulations to Heron Systems whose AI agent won the championship among the systems competitors and then beat our F-16 pilot in five straight simulated dogfights in the man-vs-machine finale,” Defense Advanced Research Projects Agency said in a Twitter statement.
The three-day trials showed AI systems’ ability to maneuver an aircraft in a simple, one-on-one combat scenario and shoot its forward guns in a classic, WWII-style dogfight.
“It’s a giant leap,” said DARPA’s Justin Mock, who served as a commentator on the trials.
DARPA intends to take the simulator used in the trials, and the simulations, to Nellis AFB, where other Air Force pilots can take a stab at trying to beat AI pilots. Next steps will be to move on to testing the latter’s performance in other types of aerial combat missions.
Eight teams were selected last year to compete in the trials, which aim to demonstrate advanced AI algorithms capable of performing simulated within-visual-range air combat maneuvering, colloquially known as a dogfight. The eight teams chosen for the competition were Aurora Flight Sciences, EpiSys Science, Georgia Tech Research Institute, Heron Systems, Lockheed Martin, Perspecta Labs, PhysicsAI and SoarTech.
The series of trials were designed to energize and expand a base of AI developers for DARPA’s Air Combat Evolution (ACE) program. ACE seeks to automate air-to-air combat and build human trust in AI as a step toward improved human-machine teaming.
The three-day event was streamed live from the Johns Hopkins Applied Physics Laboratory (APL) in Laurel, Maryland, via a ZoomGov webinar featuring multiple “channels.” Dubbed AlphaDogfight Trials TV (ADT TV), this multi-view format afforded viewers comprehensive perspectives of the dogfights in real-time and featured experts and guests from the Control Zone, akin to a TV sports commentary desk.
Trials 1 and 2, held in November 2019 and January 2020, respectively, were hosted in person at APL. Teams conducted subsequent virtual “scrimmages” in May and July to hone their algorithms leading up to Trial 3.
Each day of Trial 3 had a different focus with activities broadcasted throughout the day on ADT TV:
Day 1: each of the eight teams flew their algorithms against five APL-developed adversary AI algorithms.
On Day 2: teams flew against each other in a round-robin tournament.
On Day 3: top four teams competed in a single-elimination tournament for the AlphaDogfight Trials Championship. The champion team then flew against an F-16 pilot to test the AI’s abilities against a human.
Clik here to direct you to the DARPA event.