Engineers at Stanford University have been working on a robot that is part quadrotor flying drone and part peregrine falcon. The upper half of the robot looks like a common quadrotor drone designed to fly through the air. The lower half of the drone has a pair of clawed feet – significantly different from anything we’ve seen in the past.
Image via Stanford
On the lower half of the drone, engineers created legs inspired by the peregrine falcon. The goal of creating the robot was to allow the flying machine to perch on a limb like a bird. The ability for a drone to fly and land in a tree is a significant improvement on current drones that typically require flat ground to take off and land. One significant challenge in the design is the infinite variability of tree branches the robot would need to be able to land on while operating in the real world.
Birds make flying and landing look easy, but perching on a limb is particularly difficult to engineer into a robot capable of flight. The challenge is that no two branches are alike. Branches differ in size, shape, and texture. Some limbs will be covered with tiny sticks growing out of their surface and leaves, while others might be covered in moss. Birds have no problem perching on any limb they choose, but designing a flying robot capable of doing the same thing was a significant challenge.
The system the Stanford engineers created is called a “stereotyped nature-inspired aerial grasper” shortened to SNAG. The bird-like legs allow the flying robot to move through the sky like a typical drone but gives it the ability to carry objects and perch on various surfaces like a bird. In their research, the scientists had previously studied parrotlets, which is the second smallest parrot species. In that research, the small birds flew back and forth between special perches made of varying materials and in varying sizes.
Perches were made from wood, foam, sandpaper, and Teflon. All were embedded with sensors that allowed the team to record the physical grasping force as the parrotlets landed and took off from the material. Five high-speed cameras recorded the motions of the bird during flight and landing. Scientists learned something surprising from their research, discovering that the bird performed the same maneuvers no matter what the perch was made of.
During landing, the feet handled the variability and complexity of the surface texture. Every bird uses similar formulaic behavior, which is why the S in SNAG is for stereotyped. When designing the SNAG robot, engineers followed a similar approach as the parrotlet by having the flying robot approach every lending the same way. However, the legs of the small parrotlet wouldn’t work for a large quadrotor drone, so the team settled on the leg structure of the peregrine falcon.
The drones’ 3D printed leg structure was perfected over the course of 20 different iterations. While the bird has muscles and tendons controlling its legs, the flying robot’s legs are controlled by motors and fishing line. Each leg in the system has a motor allowing it to move back and forth and a second motor to handle grasping capability.
The motors and fishing line were routed similarly to how tendons move around the ankles of a bird. Like in the bird, the robot’s legs were designed to absorb landing impact energy and convert it into grasping force. The bio-inspired design of the robot legs resulted in a surprisingly strong and high-speed clutching action for the feet that can be closed in 20 milliseconds.
SNAG also has ankles able to lock and an accelerometer on the right foot that knows when the robot lands and triggers an algorithm responsible for balancing the robot on the landing surface. In testing, the grasping system of SNAG was able to catch objects thrown by hand, including a prey dummy, a corn hole beanbag, and a tennis ball. Eventually, the flying robot was tested in a forest, and the engineers found that SNAG performed extremely well.
In the real world, it was able to perform so well that the team decided the next step in developing the robot would focus on what happens before landing to help improve flight control and situational awareness. Before settling on the final design for the robot’s feet, the team tested two different toe arrangements. One arrangement is called zygodactyl, characterized by a foot with two toes in the front and two in the back, which is the foot arrangement of the parrotlet.
The second toe arrangement is called anisodactyl, with three toes in the front and one in the back, which is the toe arrangement of the peregrine falcon. During testing, the team found there was little performance difference between the two arrangements.
After designing a quadrotor drone capable of perching on limbs, project engineers thought of some potential uses for similar production drones in the future. One of the most likely scenarios for drones of this type will be conducting environmental research. Another potential use for this drone and its unique structure is in search and rescue operations.
To make snag more capable of performing environmental research, the team attached a temperature and humidity sensor and used it to record microclimate details in Oregon. Part of the motivation for designing the drone was to create improved tools to study the world around us.
The drone the Stanford researchers created is significantly different from drones that most of us are familiar with, such as the DJI Mavic 3 and Mavic 3 Cine drones that linked in late October. Researchers from multiple institutions are working to improve drones in large part because they have such potential in various fields.
Another significant challenge drone researchers are trying to solve is giving drones the ability to operate at high speeds in unknown environments – a skill at which birds excel. Combining a drone capable of perching on a limb in any forest around the world with an AI system capable of allowing that drone to operate in a cluttered and unknown dense forest would result in an impressive machine indeed.