Skip to main content

‘Morphing wing’ drone capable of landing just like a bird

Learning to perch a UAV on the ground using deep reinforcement learning
Drones can look pretty darn graceful up in the sky, but that effect is often ruined when you see the clumsy ways that they return to Earth.

Fortunately, this could be about to change, courtesy of work coming out of the U.K.’s University of Bristol and BMT Defence Services. In a new piece of research, investigators were able to get a “morphing wing” drone to perform a birdlike “perched landing.”

“We were very inspired by the way that birds fly, because they manipulate in a specific way in order to change their direction quite rapidly,” Antony Waldock, Principal Systems Analyst at BMT Defence Services, told Digital Trends.

The unmanned aerial vehicle (UAV) used in the experiment boasted a uniquely shaped wing, capable of reshaping itself while the craft is in flight. This is in contrast to current UAVs, which have fixed and rigid wings, thereby reducing the flexibility in how they travel through the air.

At the end of the flight, the wings then allow the drone to come in for an altogether smarter landing: lifting its wings as it does so, much as a bird would swoop in to land. This may look pretty, but — perhaps unsurprisingly — aesthetics aren’t the only thing at play here. As UAVs are increasingly used in intelligence gathering and humanitarian missions, the ability to land in small or confined spaces could prove incredibly useful.

No human pilot was directly responsible for the landing. Instead, as Waldock explained, it involves some clever machine-learning algorithms, which let the craft work out how best to make its descent.

“What we wanted to do with the machine-learning technology was to be able to tell the UAV what we wanted it to achieve, not how to achieve it,” he continued. “Using a type of machine learning called reinforcement learning, the UAV was able to treat this task as a game, with its goal being to maximize its score by landing at a certain point, velocity, and orientation. By exploring this space in an intelligent way, it’s able to work out how best to make its landing.”

Next task? To pull off a landing on a telephone wire, while muttering rude things under its breath about the neighborhood cat. (Okay, we made up that bit!)

Editors' Recommendations