NASA scientists pit humans against AI in high-speed drone racing

Drone racing is a sport that is demanding of pilots. The person flying the drone must be paying attention and have fast reflexes to avoid crashing into obstacles or other drones. NASA's Jet Propulsion Laboratory in Pasadena, California is racing drones pitting human pilots against an AI to see who's better.

The race happened back on October 12 and was the culmination of two years of research into drone autonomy that was funded by Google. The reason Google funded the work was because it was interested in JPL's work on vision-based navigation for spacecraft. The race saw a timed trial held between the new JPL AI and a drone pilot called Ken Loo, who is said to be world class.

There were three custom drones built for the event called Batman, Joker, and Nightwing and the scientists also built the complicated algorithms that were needed for the drones to operate at high speeds. These algorithms are reportedly integrated with Google Tango technology, which was worked on by researchers at JPL.

The drones were racing units and could travel at speeds up to 80mph in a straight line. The racing course JPL designer didn't support that speed with the drones only able to fly at up to 40 mph. JPL says that compared to the human pilot Loo, the AI drones flew more cautiously, but more consistently.

NASA says that sometimes the drones moved so fast that the motion blur caused them to lose track of surroundings. During the race, Loo reached higher speeds but was forced to face fatigue, something the AI piloted drones didn't face. The official results of the race saw Loo average 11.1 seconds around the course while the AI averaged 13.9 seconds. The AI was much more consistent with its lap times, Loo's times varied more.

SOURCE: NASA