drones may spend most of their time up in the sky but as buildings rise and urban areas grow denser it would be handy if they could navigate their way through the streets too. well researchers from the university of zurich and the national centre of competence in research have come up with a control system which could offer this whilst simplifying drone learning altogether. by using a deep learning algorithm they’ve dubbed dronet, the system enables them to autonomously navigate these types of settings, by showing them how cyclists and cars do their thing.

drone learning
image courtesy of the university of zurich

 

 

 

the algorithm relies on a regular camera similar to that of a smartphone to guide the drone safely around obstacles that might appear in its way. to do so, researchers at the university of zurich gathered training data captured from bikes and cars traveling through real-world environments. by showing the drones how these examples navigate urban-settings, the drones eventually learned how to do it themselves including how to cross roads and avoid obstacles such as pedestrians and construction works. the video shows how the system works whilst a paper published online details its research.