AlphaDogfight Trials: Bringing Autonomy to Air Combat

Abstract

The Defense Advanced Research Projects Agency (DARPA) Air Combat Evolution (ACE) program “seeks to increase trust in combat autonomy by using human–machine collaborative dogfighting as its challenge problem. This also serves as an entry point into complex human–machine collaboration” (https://www.darpa.mil/program/air-combat-evolution). To set the stage for ACE, the AlphaDogfight Trials program was created to explore whether artificial intelligence (AI) agents could effectively learn basic fighter maneuvers. DARPA contracted the Johns Hopkins University Applied Physics Laboratory (APL) to create an arena to host simulated dogfights— close-range aerial battles between fighter aircraft—where autonomous agents could be trained to defeat adversary aircraft. During the dogfight trials, AI agents competed against each other and the winner competed against a human pilot. By the end of the trials, the program demonstrated that AI agents could surpass the performance of human experts. APL was critical to the success of this program: the Lab created the simulation infrastructure, developed the adversary AI agents, and evaluated the competitors’ AI solutions. This article details APL’s role in advancing combat autonomy through this program.