Fixed-wing drone learns to land like a bird
We've all seen how birds touch down on a ledge. They swoop in, tilt up, and then open their wings wide at the last moment, flaring out like a parachutist to drop at almost zero speed onto the desired spot. It all seems very natural and easy for a bird. But for any sort of aircraft, especially a fixed-wing one, such a feat would be exceptionally difficult to say the least. Nevertheless, researchers at the University of Bristol claim to have achieved this with an unmanned aerial vehicle that has learned to land like a perching bird with the help of some clever algorithms and morphing wings.
Ordinarily, when a fixed-wing plane flies, air resistance against its body in flight is a measure of its drag coefficient. An aircraft is designed to minimize its drag coefficient so that it stays in the air, but when it needs to slow to land, it must increase that drag by tilting back so that its wings present more surface area to the oncoming air. Tilting back too far at too slow a speed, however, and drag can quickly exceed lift, leading the plane to stall, lose control, and possibly crash.
The University of Bristol UAV, on the other hand, is able to tilt up its wings to create more drag for landing, while morphing a section of those wings to provide just enough lifting force to maintain stable flight control. To control these complex wing structures, University of Bristol researchers (in conjunction with BMT Defence Services) employed machine learning algorithms to develop a flight controller inspired by birds.
In this way, the researchers have shown how the combination of a morphing wing UAV and machine learning can be combined to perform a bird-style perched landing on the ground. The UAV has also been trialed at altitude to prove the current methods and the team is in the throes of developing the system to the point that it can perform on-going and repeatable ground landings.
It's not the first UAV to employ morphing wing technology, UC San Diego demonstrated a version of their own quite a while ago, and NASA has been testing variable-geometry wings as part of its Adaptive Compliant Trailing Edge (ACTE) project. But the University of Bristol UAV is the first to effectively use such technology to bring in a UAV to land like a bird.
"The application of these new machine learning methods to nonlinear flight dynamics and control will allow us to create highly maneuverable and agile unmanned vehicles," said Dr Tom Richardson, Senior Lecturer in Flight Mechanics in the Department of Aerospace Engineering at the University of Bristol. "I am really excited about the potential safety and operational performance benefits that these new methods offer."
Delivered as part of an 18-month project for the Defence Science and Technology Laboratory's Autonomous Systems Underpinning Research program, the research will help add to the development, production, deployment and operational use of intelligent unmanned systems by UK Armed Forces.
The short video below shows the UAV in action.
Source: University of Bristol