While many drones already support straight-line take-offand, scientists at the University of Zurich have developed a more magical navigation algorithm that allows the four-axis fleet to perform some impressive aerial acrobatics. The team hopes the new algorithm will help improve the speed of drones and their efficiency in areas such as search and rescue.
Screenshot from the video (from: University of Zurich)
Previously, many human pilots on YouTube have demonstrated their masterful remote control skills. In addition to its own skills, there are various adjustments to the hardware and software of the device.
To apply this idea to automated control systems, the University of Zurich team developed an artificial neural network to perform acrobatic movements through flight simulation software.
Based on this, researchers can easily simulate different trajectories, paths, and acrobatic shipping stakes without having to worry about physical damage to the drone.
After several hours of training in the simulator, the quadcopter equipped with special navigation algorithms can use on-board cameras and sensors to implement a variety of control instructions for a series of acrobatic actions, such as roller flip, foldback, high thrust climbing, angular acceleration maneuvering and so on.
Deep Drone Acrobatics – RSS 2020 (via)
Davide Scaramuzza, a professor of robotics at the University of Zurich, says that by learning algorithms, drones can master acrobatics that can also be challenging for the best human pilots.
In addition, the team hopes the new algorithm will extend the drone’s range and flight time in order to play a more effective role in areas such as search and rescue and drone delivery.
Details of the study have been published in the recently published journal Robotics Science and Systems.