Due to the coronavirus pandemic, the United States will not conduct face-to-face tests of new fighters. It has been replaced by a competition of AI algorithms that will control the aircraft.
Due to the coronavirus pandemic, the Defense Advanced Research Projects Agency (DARPA) will not conduct face-to-face tests of new fighters. They will be replaced by the AlphaDogfight Trials neural network algorithms competition. It will be held virtually – participants and spectators of the event will be able to watch them online. The fighters will be controlled by artificial intelligence (AI) algorithms that will simulate air combat.
The Department of Defense has launched AlphaDogfight Trials to expand its AI development base through the Air Combat Evolution (ACE) program. ACE’s goal is to automate aerial combat and build pilots’ confidence in artificial intelligence (AI) so that they can co-operate the aircraft in the future. Researchers are confident that the new system will be able to conduct air combat faster and more efficiently than humans, significantly reducing the workload on the pilot and giving him time to make tactical decisions in a larger combat mission.
In January of this year, DARPA conducted a second round of trials, where participants presented advanced AI algorithms. On it, neural network algorithms created by eight participants fought an air battle with an artificial intelligence system of a conventional enemy. The battles between the algorithms were fought one-on-one at a relatively low difficulty level.
Eight teams will take part in the third trial, which will present their algorithms in the tournament mode. On the final day, the top four AIs will compete for the singles championship title. The latter team will have the opportunity to test their algorithm with a live US Air Force pilot.