Ranked as the 58th best university globally and the number one school in Switzerland, the University of Zurich has become a leader in the fields of robotics, especially drones. Led by Professor Davide Scaramuzza, the Robotics and Perception Group (a sub-branch of the Departments of Informatics and Neuroinformatics) has been making some incredible advancements in the complex, autonomous flight of micro drones. Recently the Robotics and Perception Group partnered up with American technology company Intel who has become well known for their micro drone displays.

In 2016, Intel set a world record for the Most Unmanned Aerial Vehicles (UAVs) airborne simultaneously when they sent 500 Shooting Star drones into the air to perform for the company’s 50th anniversary celebration. The Shooting Star is a micro drone with LED lights that can be pre-programmed to arrange into an array of patterns. Drones such as Intel’s Shooting Star have gone on to be featured at the Olympics, the Super Bowl, countless concerts and arena performance, Disney light shows, in place of fireworks, and much more. A programmer carefully choreographs the flight plan of the drones, each drone knowing the plan of action before being released into the sky. Professor Scaramuzza and his students wanted to develop a drone that could accomplish similar tasks autonomously.

With the creation of a complex algorithm, standard, and neuromorphic cameras the Robotics and Perception Group designed a drone that can maneuver in ways that can take human pilots years to practice. At accelerations up to 3g the drones can do barrel rolls, flips, and loops with complete autonomy. These are all moves that require a lot of skill and are often used to prove a drone pilot’s acumen in competition. Rather than being controlled by a human pilot or through a predetermined course,
Scaramuzza and his team have drones complete these complex acrobatics with only onboard sensing and computation.

The drones have onboard sensing that can compute it’s spacial awareness and translate the aerobatic movements. Before testing on an actual drone the team ran computer simulations to predict the drone’s autonomy. Within the simulator program the drone’s trajectory, flight paths, and thrust could be acutely monitored. The team was able to collect enough data to support the theory that the drone could successfully maneuver in a real world scenario without risk of failure while still completing tasks that no single drone had ever done on it’s own before. “Our algorithm,” Scaramuzza said, “learns how to perform acrobatic maneuvers that are challenging even for the best human pilots.”

While the research done by the Robotics and Perception Group could lead to an incredible Intel drone light spectacle, both parties see this as a way of advancing drone capabilities on a broader scale. Drones that are able to complete complex acrobatic moves autonomously open up a lot of advantages in drone technology. It could lead to a more efficient use of battery power and longer flight times. But perhaps the biggest advantage is that it would pave the way for higher drone safety protocols. As the FAA continues to find ways to allow drones into shared airspace, one of the biggest concerns is how drones will be able to detect obstacles like other aircraft. However, as Professor Scaramuzza said, “This navigation is another step towards integrating autonomous drones in our daily lives.”