AI fighter pilot wins human veteran navy pilot in simulated battle, what is the advantage of AI over humans?
On August 18, 2020, the
Navy F/A-18 Squadron Commander's Take On AI Repeatedly Beating Real Pilot In Dogfight-The Drive
https://www.thedrive.com/the-war-zone/35947/navy-fa-18-squadron-commanders-take-on-ai-repeatedly-beating-real-pilot-in-dogfight
The state of simulated battle between AI pilot and Navy pilot by DARPA is live-streamed on YouTube, and you can see it from around 4 hours 39 minutes 50 seconds in the following archive.
AlphaDogfight Trials Final Event-YouTube
In the US Navy, pilots receive training for air combat called 'Basic Fighter Manuva (BFM)' every day. BFM is for the pilot to understand the positional relationship and situation with the enemy in a physically harsh three-dimensional space, not only to understand how to actively operate a fighter plane, but also to understand the weapon system. With an accurate understanding, we will also train the ability to judge 'I need to pull the trigger to fire the weapon'. “A dynamic and stressful environment creates better fighter pilots,” said Farba.
However, “There are some situations where pilots are in a pit of enemies. If they encounter an enemy on the battlefield, they have only one goal:'defeat the enemy with the first shot before the enemy shoots. Training” and the actual battle are very different.
For example, in simulated combat training, if two aircraft come into contact and shoot each other immediately, dogfighting will continue until either is clearly shown to be the winner. Only after they settled down, and when the pilots saw the video of the training results, they realized that at the first meeting, the bullets had struck each other's wings and engine. Mr. Farba points out that the conclusion of the simulated battle is not necessarily the same as the actual battle because the wings and the engine may be lost due to the impact of the missile in the actual battle even if it is faint in the simulated battle.
On the other hand, 'the real sport for fighter pilots is BFM engagement training using only machine guns,' Farba argues. In the BFM engagement training only with machine guns, the pilot will test whether various movements can be utilized while considering the fuel consumption of the fighter. It seems that the pilot always thinks about two to three movements according to the movements of the opponent, and BFM engagement training only with machine guns can be said to be exactly chess in the air.
The following is a video of a pilot who actually boards the
FA 18E Super Hornet BFM in Portland, OR-YouTube
The simulated battle between the AI pilot and the human pilot held in the DARPA contest was also a form of BFM engagement training for this machine gun only. Mr. Furba has judged that 'AI was superior to humans' in the following three points in this simulated battle.
・Accurate machine gun shooting
In order to increase the accuracy of shooting, the pilot must always consider the plane of movement, the range and the distance to the opponent while operating the fighter. According to Mr. Furba, the shooting time is only 3 seconds when the opponent is caught in front, and the pilot will track and evaluate the opponent within 3 seconds, focus on the opponent, and do not collide with the opponent. It is necessary to pull the trigger while calculating the avoidance movement. However, the AI of defense company Heron Systems , who won the DARPA contest, ends this calculation in an instant. In the 5 mock battles conducted by DARPA, there were about 15 effective shots by AI, but none by human pilots.
・Efficient flight
For a fighter, fuel is the lifeline itself. In order to fly while reducing fuel consumption as much as possible, the pilot must check the aircraft's angle, airspeed, and gravitational acceleration from the equipment in the cockpit, and even a veteran pilot can always maintain the highest fuel economy It is a difficult task. However, processing such data is a tricky task for AI, and it enables efficient flight while fine-tuning the speed and direction of the fighter plane as needed.
・Release from physiological restrictions
In the simulated battle of AI and human beings this time, in the last battle of the fifth round, a dogfight was developed in which each person chased each other and flies like a circle and a circle. At this time, in reality, the gravitational acceleration of 9G is applied to the pilot's body. To withstand the gravitational acceleration of 9G, pilots need to stay focused on the maneuver, while exerting considerable force on the leg and abdominal muscles. 'Humans are always the limiting factor in aircraft performance,' said Farba. However, AI can continue to control a fighter plane, no matter how much G it takes, and can fly in a trajectory that is impossible for the human body.
'We're ready to get the AI into the cockpit,' Farba said, expecting AI to be deployed as a fighter pilot in the not too distant future. 'Let's prepare a dream story that 'I was good at that time' and 'Grandpa was the greatest pilot in the Navy ever' in order to talk while holding my grandson.'
In addition, this time it was a simulated battle with a simulator, but a real dogfight using an autonomous unmanned fighter is scheduled to be held in 2021.
An aerial duel of manned fighter and autonomous unmanned fighter will be held in 2021-GIGAZINE
Related Posts:
