AI pilots used to control MQ-20 drone in test
NewsApril 12, 2023
SAN DIEGO, California. General Atomics Aeronautical Systems (GA-ASI) used artificial intelligence (AI) to control an MQ-20 Avenger Unmanned Aircraft System in a recent test that involved maneuvers between human and AI pilots using Live, Virtual, Constructive (LVC) and a Low Earth Orbit (LEO) satellite communication datalink, the company announced in a statement.
The LEO SATCOM connection allowed for rapid retraining and redeployment of AI pilots while the aircraft was airborne, the company says. The test aircraft used a Beyond Line of Sight (BLOS) Active Electronically Scanned Array (AESA).
The demonstration was important because of the integration of AI pilots and LVC ecosystem using a BLOS datalink, the company said, adding that the test fused third-party capabilities, human-on-the-loop control, and autonomy to make a teaming of human and machine pilots possible. Operator commands were captured via hands on throttle-and-stick controls and sent via LEO SATCOM to AI pilots running Reinforcement Learning (RL) algorithms. Data from agent performance was collected, sent to the ground for retraining to improve performance, and then redeployed via LEO SATCOM, the statement reads.