For years we have seen the challenges that autonomous driving still has ahead, a task that for the moment humans do better than machines. But this could have the days numbered, or at least when it comes to drones.
A group of researchers from the University of Zurich has created an autonomous driving algorithm capable of beating two professional drone pilots in a race. This new algorithm allows the drone to establish more efficient routes, calculating the angles, speed, and trajectories necessary so that it can overcome obstacles in the circuit at full speed.
There are two key elements that have enabled researchers to reach this milestone. The first has been to incorporate the limitations that the drone has (such as its maximum speed or its limited ability to turn) into the equation. And the second has been not to decompose the route of the circuit into smaller pieces, but to analyze it as a “whole”. This is how Philipp Foehn, the main author of the project, explained this.
So that humans would not be at a disadvantage against the drone - since the drone was already familiar with the circuit - the researchers allowed them to practice and study the circuit extensively beforehand. In addition, they all had the same type of drone. But there was no way: the drone beat the humans in all the laps of the race.
Although these advances may seem superfluous, this algorithm "can have enormous applications in the delivery of packages with drones, in search and rescue work, and much more," said the director of the Robotics Group at the University of Zurich, Davide Scaramuzza.