NASA checked who manages the drone better: man or AI

NASA engineers developed several quadcopters of their own design, and also created the AI ​​that manages them. As a test of the capabilities of the system, representatives of the agency held a competition of their AI with a professional drone operator

Researchers at NASA's Jet Propulsion Laboratory in Pasadena, California, recently held very peculiar competitions. They were attended by man and artificial intelligence ( its weak form ). Both rivals drove drones at high speed. In this case, the human operator was far from working for the first time with a drone, it was a professional “pilot”.

About the competition, which took place on October 12, it became known only now. And the competition itself is part of a large two-year project sponsored by Google. The company was interested in the technology of NASA in the field of visual navigation unmanned systems. In order to demonstrate the merits of its development, the agency and conducted this kind of experiment.

The project team has built three custom drons, called Batman, Joker and Nightwing. Complex algorithms were also developed that allowed Copters to fly at high speed while avoiding obstacles. These algorithms were connected to the Google Tango system, in the development of which NASA was also involved.

All three drone were created by speed - the maximum speed of their flight is 129 kilometers per hour. The maximum speed of the device can be achieved only in the case of flight in a straight line. Due to the difficult flight path, which was due to the configuration of the "track", drones could gain a maximum of about 64 kilometers per hour. Higher speed can not be gained, otherwise inevitably there was a collision with an obstacle.

“We opposed the algorithms against the human operator, who usually controls the drone using intuition and sensations,” says Rob Reid of JPL. “You can see how the copters under AI control gently pass obstacles and change course, while the human operator manages the UAVs much more aggressively.”

During the competition, Ken Lu, one of the most famous drones operators in the world, “performed” on the side of people. He says that during the competition he encountered such a feeling as fatigue, which, of course, was not experienced by the car, which almost unmistakably drove its quadcopters. A person gets tired quickly enough, and the coordination of movements, the reaction, the accounting of mistakes - all this worsens with each round.

Lou was able to gain more speed, his drones flew faster than the drones running the machine. He also performed spectacular stunts in the air. Well, machine intelligence controlled its copter gently, smoothly, non-aggressively. But very confidently, and this confidence, so to speak, was felt throughout the flight.

"This is one of the most difficult tracks that I have ever passed," - said Lou after the end of the competition. “One of the problems is fatigue, I quickly got tired. My feelings were dulled, I started to lose, despite the fact that the track was well known to me. ”

Initially, the human operator had the advantage of a few seconds. He passed the circle in 11.1 seconds, while the AI ​​traveled the same distance in 13.9 seconds. But as time went on, Lou needed more time, but the AI ​​went round and round for about the same time.

The developers of the drones and the system itself said later that the copters could fly much faster. "One day you will see them at professional competitions," NASA representatives said. AI used GPS as a navigation tool. Without this, it would be impossible to navigate the computer.

As for the practical use of technology, this, of course, is far from being only a competition. Now many companies are developing systems for the automatic delivery of goods to customers. A similar system is trying to create, for example, Amazon. The same technologies can be used in the warehouse, during the liquidation of the consequences of disasters (for reconnaissance of the area). In the future, according to NASA, drones can even fly through a space station, deftly maneuvering to deliver cargo to astronauts or “transport” things from one place to another.


All Articles