General Atomics reported successful tests of the Avenger MQ-20A autonomous unmanned aerial vehicle equipped with artificial intelligence. GA-ASI’s original reinforcement learning (RL) architecture was used in flight to develop and validate an RL agent in an operational environment.

The team flew in “follow and avoid” mode, where the flight path was updated in real time to avoid opponents.

Real-time monitoring was provided to the system using a network of Infrared Search and Tracking (IRST) sensors provided by Lockheed Martin.

Avenger’s fully autonomous flight used an AI pilot for 30 minutes as part of a live, virtual and constructive swarm of collaborative drones. The flight was carried out as part of GA-ASI’s ongoing commitment to the development of advanced autonomous artificial intelligence and machine learning.

“The flight was successful and demonstrated a host of innovative capabilities in the race to provide autonomy for Collaborative Combat Aircraft (CCA). It is exciting to see how AI can be used to improve how and where we fly unmanned systems as the battlefield becomes more complex. The ability to “follow and dodge” dynamically update its path is the first step towards an ecosystem of cooperative autonomous warplanes.”Michael Atwood, Senior Director of Advanced Programs at GA-ASI.

The Avenger is a remote-controlled aircraft for flights at medium and high altitudes. The unmanned aerial vehicle can perform large-scale surveillance and emergency strike missions on land or at sea.

Unlike previous MQ-1 Predator and MQ-9 Reaper (Predator B) drones, the Avenger is equipped with a turbofan engine. Its design also includes stealth features – an internal gun store and an S-shaped exhaust to reduce infrared and radar signatures.