Posted 4 октября 2022, 06:05

Published 4 октября 2022, 06:05

Modified 24 декабря 2022, 22:38

Updated 24 декабря 2022, 22:38

"Avenger" does not need an operator: the drone itself thinks for people

"Avenger" does not need an operator: the drone itself thinks for people

4 октября 2022, 06:05
Сюжет
Aircrafts
General Atomics Aeronautical Systems has tested a new artificial intelligence vehicle that controlled the company's prototype large drone, the MQ-20A Avenger.

Alexander Sychev

The drone appeared in 2009 as a further development of the MQ-1 and MQ-9 unmanned aircraft systems created by the company and remained at the General Atomics test base. The drone is able to stay in the air for up to 20 hours, fly at speeds up to 740 kilometers per hour and carry about one and a half tons of combat load inside the fuselage or on hangers. The UAV flight ceiling is 18 thousand meters.

All currently existing unmanned aerial vehicles (UAVs) in service with the United States are controlled from the ground using communications satellites. During the current tests, the drone flew autonomously for half an hour, assessing the surrounding combat situation and making decisions on evading the detected threats. Moreover, the Avenger interacted with the drones of other systems, creating a swarm of drones.

The autonomous flight was controlled by the CODE (Collaborative Operations in Denied Environment) system and the standard OMS messaging protocol, which are part of the US Department of Defense's large-scale Skyborg program. Its goal is to create autonomous and semi-autonomous reconnaissance and combat aircraft that could perform independent combat operations or support guided aircraft.

This time, the flight also used a new Reinforced Learning (RL) computer program developed by the company. This program is an innovative tool for next-generation military platforms that enables decision making in dynamic and uncertain real-world environments.

The aircraft flew in a "pursuit and avoid" mode, updating the flight path in real time to avoid the enemy. All the data needed to make decisions came from a new system of infrared sensors for search and tracking (TacIRST), provided by Lockheed Martin.

“Our drone’s ability to ‘hunt and avoid’ dynamically updating its flight path as threats are identified is the first step towards building an ecosystem of collaborative autonomous warplanes”, - said Michael Atwood, senior director of advanced programs at General Atomics.

The current flight is the latest in a series of autonomous flights carried out by the company using internal funding. A few years ago, the leadership of General Atomics came out with a sharp criticism of the actions of the Pentagon, tired of receiving constantly changing requirements for the developed systems, which were funded from the state budget. Sometimes the wishes of the generals were so contradictory that almost finished projects had to be abandoned.

One such sufferer was the artificial intelligence program for UAVs. The Pentagon does not officially recognize the viciousness of its inconstancy, but the self-financed work of General Atomics began to show genuine interest.

Subscribe
Яндекс.Метрика