Dr William Roderick, 20, said: "Mimicking the way birds fly and roost is not easy. Over millions of years they have evolved to make take-off and landing look so easy, even among all the complex and changeable branches you find in the forest."

Bird robots in the forest

Like parrots, the avibots treat each landing the same way. But, to illustrate the size of a quadrotor, the Avibot is based on the legs of a peregrine falcon. It has a 3D-printed structure that took 20 iterations to perfect, replacing bones, motors and fishing line in place of muscles and tendons.

Roderick moved equipment, including a 3D printer, from Stanford's Lentink lab to rural Oregon, where he set up a basement lab for controlled testing. He sent the bird robot along an orbital system, firing the robot at different surfaces at predetermined speeds and directions to see how it behaved in various situations. Finally, Roderick ventured into a nearby forest for some real-world test runs.

Learning has a unique perspective. For example, the researchers ran the robot with two different toe arrangements, with three toes in the front and one in the back, like a peregrine falcon. Two toes in front, two in back, like a parrot. They were surprised to find that the performance difference between the two was small.

Lentink praised Roderick for his persistence in a years-long project.