Heron Systems, a small company based out of Maryland, took first place in August during DARPA’s AlphaDogfight Trials Final event, a three-day competition designed to demonstrate advanced algorithms capable of performing simulated, within-visual-range air combat maneuvering, or commonly referred to by people in the cockpit as dogfighting. Before defeating a flesh-and-blood pilot, Heron Systems’ F-16 AI (artificial intelligence) agent first had to defeat seven other companies’ F-16 AI agents, with some of the agents being far from what anyone would describe as a small company.
The participating companies ranged from major defense contractors to four-person firms, with each spending less than a year developing and teaching their AI agents how to fly and fight to win in simulated aerial combat. The teams that Heron beat before getting to go up against a human were Aurora Flight Sciences, EpiSys Science, Georgia Tech Research Institute, Lockheed Martin, Perspecta Labs, Physics AI, and SoarTech. AI agents developed by Lockheed Martin, Aurora Flight Sciences, and PhysicsAI rounded out the top four teams.
“The AlphaDogfight Trials accomplished exactly what we’d set out to do,” said Colonel Dan Javorsek, Program Manager in DARPA’s Strategic Technology Office. “The goal was to earn the respect of a fighter pilot—and the broader fighter-pilot community—by demonstrating that an AI agent can quickly and effectively learn basic fighter maneuvers and successfully employ them in a simulated dogfight.”
The trials were designed to energize and expand a base of AI developers for DARPA’s Air Combat Evolution (ACE) program. According to DARPA, the ACE program was established to increase trust in combat autonomy by using human-machine collaborative dogfighting as its challenge problem.
ACE applies existing AI technologies to the dogfight problem in experiments of increasing realism. In parallel, ACE goals are to implement methods to measure, calibrate, increase, and predict human trust in combat autonomy performance. Finally, the program will scale the tactical application of autonomous dogfighting to more complex, heterogeneous, multi-aircraft, operational-level simulated scenarios informed by live data, laying the groundwork for future live, campaign-level Mosaic Warfare experimentation.
That is, in many envisioned scenarios, in an air domain contested by adversaries, a single human pilot will be able to increase lethality by effectively orchestrating multiple autonomous unmanned platforms from within a manned aircraft. This will shift the human role from single platform operator to essentially mission commander.
In particular, ACE is expected to create a hierarchical framework for autonomy in which higher-level cognitive functions (e.g., developing an overall engagement strategy, selecting and prioritizing targets, determining best weapon or effect, etc.) may be performed by a human, while lower-level functions (i.e., details of aircraft maneuver and engagement tactics) is left to the autonomous system. For this to be possible, the pilot must be able to trust the autonomy to conduct complex combat behaviors in multiple scenarios.
The bottom line is: an algorithm can withstand a lot more gs than a pilot can. The only limiting factor then is whether the airframe itself can withstand the more extreme maneuvers that a pilot cannot.
