Autonomous Drone Races Against World-Class Human Pilots, Wins Big

A drone flying through smoke 6 photos
Photo: Robotics and Perception Group, UZH
A drone racing along a time-optimal trajectoryA quadrotor races against two human pilotsA quadrotor races against two human pilotsA quadrotor races against two human pilotsA quadrotor flying through smoke
For the first time, an autonomous drone has beat two human pilots in a drone race. Thanks to a new algorithm created by a team of researchers from the University of Zurich, which determines the time-optimal trajectories, the drone outperformed the pilots.
Because of their limited battery life, drones must fulfill whatever duty they are given as quickly as possible. Not only that, but they need to be fast enough in order to be used in different applications such as search and rescue operations, inspecting a building, or delivering cargo.

Moreover, in recent years the vehicles have not only been used for aerial photography, but they also became the subject of one of the fastest-growing sports in the world: drone racing. What began as a hobby has now evolved into a professional international sport, where pilots race with custom-built drones at speeds over 80 mph (129 kph) through 3D courses.

Human drone pilots are incredibly skilled at this and have consistently outperformed autonomous drones in drone races – until now. A team of researchers at the University of Zurich has developed an algorithm that can tell an autonomous drone the fastest trajectory through a series of waypoints in a circuit.

"Our drone beat the fastest lap of two world-class human pilots on an experimental race track," says Davide Scaramuzza, head of the Robotics and Perception Group at the University of Zurich.

To see how the algorithm performed, researchers used it on a quadrotor (drone with four propellers) around a race track. The quadrotor had to race another identical drone which was remote-controlled by two human pilots. The pilots were given the opportunity to practice on the circuit ahead of the race. Even if they got the odds on their side, the quadrotor won.

It zipped through all of the laps and had a more consistent performance. Unlike human pilots, once the algorithm has determined the ideal route, it can replicate it multiple times in the same way.

The team says that the algorithm still needs some improvement before it gets used in commercial applications. Currently, the computer can take up to an hour to calculate the drone's time-optimal trajectory. In addition, the quadrotor now relies on external cameras to determine its location.

Scientists plan to reduce the time it takes to calculate the path and to use onboard cameras. But the algorithm still has the potential to be helpful in a variety of applications.

For those who want to learn more about the methods applied, the scientists' extensive work was published in the journal Science Robotics. You can see the drones in action, too, in the video released by UZH Robotics and Perception Group.

If you liked the article, please follow us:  Google News icon Google News Youtube Instagram
About the author: Florina Spînu
Florina Spînu profile photo

Florina taught herself how to drive in a Daewoo Tico (a rebadged Suzuki Alto kei car) but her first "real car" was a VW Golf. When she’s not writing about cars, drones or aircraft, Florina likes to read anything related to space exploration and take pictures in the middle of nature.
Full profile


Would you like AUTOEVOLUTION to send you notifications?

You will only receive our top stories