AI enables drones to fly into the unknown

Researchers at the University of Zurich have developed a new approach with which autonomous quadrocopters can fly at high speed through unknown, confusing surroundings. This is done with the help of the sensors and calculations on board the drone.

When it comes to exploring complex and unfamiliar environments such as forests, buildings or caves, drones are hard to beat. They are fast, manoeuvrable and small, transport payloads and can go practically anywhere with sensors. But so far, autonomous drones have hardly been able to find their way around an unknown environment without a map. Experienced human pilots are currently required to achieve their full potential.

“When maneuvering a drone, you have to understand the environment in fractions of a second in order to quickly steer the drone onto collision-free paths,” says Davide Scaramuzza, professor who heads the Robotics and Perception Group at the University of Zurich. “This is very difficult for both people and machines. Experienced pilots can reach this level after years of training. But machines still have a hard time doing that. “

AI algorithm learns from a simulated expert

In a recent study, Scaramuzza and his team trained an autonomous quadrocopter to fly at speeds of up to 40 kilometers per hour through previously unknown environments such as forests, buildings, ruins or trains without colliding with trees, walls or other obstacles. The drone only relies on the built-in cameras and the calculations of the quadrocopter.

The drone’s neural network – its brain, so to speak – learns to fly around obstacles by observing a kind of “simulated teacher”: an algorithm that flew a computer-assisted drone through a simulated environment full of complex obstacles. The algorithm was always informed about the position of the quadrotor and the measured values ​​of its sensors and had enough time and computing power to calculate the best flight path in a fraction of a second.

This “simulated teacher” cannot be used outside of the simulation, but its data is used to teach the neural network how to predict the best flight path based on the data transmitted by the sensors. This is a great advantage over existing systems that first create a map of the environment based on sensor data and then plan flight paths within this map – two steps that take a lot of time and make it almost impossible to fly at high speed.

Source: com! professional by

*The article has been translated based on the content of com! professional by If there is any problem regarding the content, copyright, please leave a report below the article. We will try to process as quickly as possible to protect the rights of the author. Thank you very much!

*We just want readers to access information more quickly and easily with other multilingual content, instead of information only available in a certain language.

*We always respect the copyright of the content of the author and always include the original link of the source article.If the author disagrees, just leave the report below the article, the article will be edited or deleted at the request of the author. Thanks very much! Best regards!