Drones learn autonomous navigation by copying bikes, cars

"With this algorithm we have taken a step forward towards integrating autonomously navigating drones into our everyday life," said researcher Davide Scaramuzza.

Brooks Hays

Jan. 23 (UPI) -- Autonomous flight is no problem in open airspace, but today's self-flying drones are ill-equipped to navigate busy city streets and tree-filled neighborhoods at low altitudes. Researchers at the University of Zurich are trying to change that.

Scientists have developed a new algorithm called DroNet. The software helps drones safely navigate urban obstacle courses.


"DroNet recognises static and dynamic obstacles and can slow down to avoid crashing into them," Davide Scaramuzza, professor of robotics and perception at Zurich, said in a news release. "With this algorithm we have taken a step forward towards integrating autonomously navigating drones into our everyday life."

The new technology is powered, not by advanced sensors, but a sophisticated algorithm designed to analyze the drone's surroundings and respond quickly and efficiently.

RELATED Lefties and righties: Flights of honeybees reveal individual directional tendencies

"This is a computer algorithm that learns to solve complex tasks from a set of 'training examples' that show the drone how to do certain things and cope with some difficult situations, much like children learn from their parents or teachers," said Scaramuzza.

Scientists provided the software with training examples of bikes and cars navigating urban traffic. The algorithm can identify navigational patterns and glean basic traffic laws by watching the movements of bikes and cars through a crowded environment.

Tests showed the algorithm allowed the drones to not only use their training examples to navigate city streets, but also safely utilize the lessons learned in other environs, like a parking lot or indoor office.

RELATED Study weighs risks of human-drone impacts

Researchers detailed their efforts in the journal IEEE Robotics and Automation Letters.

The technology could eventually be used to power drones working in search and rescue situations or for parcel delivery services.

"Many technological issues must still be overcome before the most ambitious applications can become reality," said Zurich Ph.D. student Antonio Loquercio.

RELATED Study finds drones accurate for relay of RFID signals for inventory control

RELATED Scientists train drones to land on moving targets

Latest Headlines