General Atomics claims fully autonomous AI-piloted flight with Avenger MQ-20A
General Atomics claims that the latest Avenger MQ-20A test flight demonstrated capabilities in the race to operationalise autonomy for collaborative combat aircraft.
A company-owned Avenger MQ-20A UAS was used on 12 September by General Atomics Aeronautical Systems to fly a military aircraft with an AI pilot and operationally relevant mission systems software.
The company revealed on 19 September that the ‘completely autonomous flight’ with the AI pilot lasted for about 30 minutes as part of a ‘cooperating live, virtual, and constructive UAS swarm’.
General Atomics employed its Reinforcement Learning (RL) architecture to develop and validate an RL agent in an operationally relevant environment — specifically, enabling ‘chase and avoid’ behaviour in which real-time updates were made to the flight path in order to avoid adversaries using live fused tracks.
The live tracks were provided to the system using a TacIRST embeddable infrared search and track sensor system supplied by Lockheed Martin.
General Dynamics Mission Systems supplied the mission computer for the flight, which is part of a series by General Atomics to demonstrate AI and machine learning concepts for advanced UAS.