New Technology Keeps Drone in Flight Even with Rotor Failure
Researchers at the University of Zurich show how a damaged quadcopter can continue to fly using only onboard vision sensors.
Millions of people fly on airplanes around the world on a daily basis, secure in the knowledge that there is virtually no chance the plane will drop out of the sky, even if it loses an engine. The same can’t be said for multi-rotor drones, which will begin to spin on its axis and fall to the ground if one of its rotors becomes inoperative. Now, researchers from the University of Zurich have solved that problem by providing the drone with a sense of position using a pair of image sensors.
“When one rotor fails, the drone begins to spin on itself like a ballerina,” states Davide Scaramuzza, head of the Robotics and Perception Group at UZH. “This high-speed rotational motion causes standard controllers to fail unless the drone has access to very accurate position measurements.” Think of it like a helicopter that loses function in its tail rotor. With no opposing force, the helicopter will begin to spin in the same direction as its top rotor and will no longer maintain lift.
Quadcopters and other multi-rotor drones rely on position to maintain flight, and if it loses function in one of its rotors, it’s difficult for it to maintain position unless it has access to position measurements. To overcome that problem, the researchers outfitted a quadcopter with two different cameras to provide a visual reference- a standard camera that records images at several times per second at a fixed rate and an event camera that activates when a change in light is detected.
Algorithms combine the information gathered from the two cameras and use it to track its position relative to its surroundings. The drone’s onboard computer can then use that information to maintain control of its flight. While this is a good step forward, it still has its problems, specifically when the drone is flying in lowlight conditions, which causes the event camera to experience motion blur, resulting in disorientation and eventually a crash.
More details can be found in the team's paper and the code is available on GitHub.