The battle between drones and anti-drone technology is heating up. Since drones are now very maneuverable, special technologies are being developed to neutralize them for security reasons and even for military applications. And even the public has been known to take down drones by shooting them.
Most drones just aren’t very good at avoiding incoming objects, especially if those objects are moving fast. Drones can often avoid static obstacles like buildings or trees, but even the most advanced drones struggle to react to dynamic objects and adjust their trajectories accordingly. This makes them vulnerable to fire or interference, but it also means they are not good at avoiding collisions with other drones or with people or aircraft.
But now a team from the University of Zurich has developed a drone that can dodge, dive and dive to avoid an incoming object. The drone is shown dodging a soccer ball that researchers throw at it from different angles, moving at speeds of nearly 32 feet per second. He is able to dynamically and autonomously avoid the ball, which means he can react very quickly to nearby moving objects and avoid collisions.
And the drone doesn’t use an external position tracking system as you might imagine. It uses its own on-board sensors, using a technique called visual-inertial odometry to estimate the location of the incoming bullet. It works by using a type of extremely low latency camera, called an event camera, which tracks the movement of the ball and can relay that information to the drone’s motors to control its movements.
“We really wanted to push the boundaries and see what these robots are capable of,” University of Zurich researcher Davide Falanga told The Verge. Even though these types of cameras are still rare due to their high price and early stage of development, researchers believe that they will eventually become available to the general public once they become more affordable: “Absolutely , I think in the long term, I think we’re going to see more and more use of these cameras,” Falanga said.
The research is published in the journal IEEE Robotics and Automation Letters [pdf link].