NASA Drone Race faces a human pilot against artificial intelligence


Of the many, many projects in which NASA is currently working, a more terrestrial task involves the investigation of autonomous drones, with the aim of making a bot capable of navigating difficult terrain on our planet of origin, before possibly moving to space missions later on the road.

To test their last years of research on artificially intelligent drones, NASA's Jet Propulsion Laboratory (JPL) in Pasadena, California, decided a race was in order. With their unmanned AI plane on the one hand, they invited professional drone runner Ken Loo to see if a human could compete faster. Then they set up a course in the lab and gave their drone AI a preprogrammed path to follow: Loo, of course, would trust his own abilities.

We will avoid spoiling the results until you have had the opportunity. to see. Check out the race below:

According to JPL, the AI ​​drones are designed according to racing specifications and capable of reaching speeds of up to 80 miles per hour, but the race of obstacles forced them to move at half that speed to navigate effectively. It turned out that the drones were too cautious, and could have used that extra boost, because Loo finally won the day despite being increasingly fatigued towards the end: he averaged 11.1 seconds per round, while the AI ​​averaged 13.9 seconds.

The whole thing fell apart last month, but JPL just released the images of their custom drones, called Batman, Joker and Nightwing (they seem to be functionally identical). Loo may have won the race by a close margin, but the researchers were even more than happy with his performance Batman who made use of the Google Tango program that JPL helped to work with.

Because even if the AI ​​drone sailed slowly along the racetrack, it still did it with extreme precision, driving much more smoothly than Loo's sudden movements. JPL Rob Reid, the task manager of the project, said the following in a press release that came along with the footage:

  Opening quote

"We put our algorithms against a human, who flies much more because of the feeling I can actually see that AI flies gently on the course, while human pilots tend to accelerate aggressively, so their path is more irregular. "

  Closing date

Therefore, this was not exactly a "slow and steady win the race" situation since the drone AI lost decisively, but the fact that it could navigate those courses consistently it means that soon it could be ready for larger field tests. JPL wants to see it working in warehouses or helping in disaster situations soon, but being at NASA could give it an even brighter future than that.

NASA has quite a few drone projects underway at this time, with potential plans to send them to Mars, Saturn's moon, Titan, or someplace else. It may not be "Batman" who goes to Mars (as great as it would be to hear NASA say "we have sent Batman to Mars"), but its technology is moving in the right direction at this time.

Source link

Leave a Reply

Your email address will not be published.