Nutty Professors Built a Squirrel Drone

It is a well-established and indisputable scientific fact that squirrels are the most interesting animal in the known universe, and anyone that disagrees is wrong. I mean, have you ever seen them chasing each other up trees and zooming between branches faster than an IndyCar driver staring down the checkered flag? These high-wire antics may seem reckless, but squirrels are full of surprises. They can (in theory, at least) safely fall from any height by stretching out and giving themselves a very large size-to-mass ratio, which greatly reduces their terminal velocity.

Suffice it to say, when any squirrel-related tech news arises we clamor to cover it here at Hackster News. The latest example comes to us from the Pohang University of Science and Technology, where a team of researchers is building a highly-maneuverable drone modeled after the behavior of the flying squirrel. It incorporates foldable wings into its design that allow it to perform tighter, more responsive maneuvers than conventional drones. Considering how agile the furry variety of flying squirrel is in the air, that sounds like a recipe for amazing.

The entire system weighs just 548 grams, with the foldable, silicone wings themselves adding only 24 grams. It is powered by four high-speed motors, two servos for wing deployment, and is guided by an Arduino Portenta H7 with a dual-core STM32 microcontroller. Sensors include dual IMUs, a barometer, and GNSS — all of which help the drone stay oriented and on-course mid-flight.

What makes this drone so special is not just the addition of wings, but how those wings are used. The team designed a Thrust-Wing Coordination Control (TWCC) framework to intelligently manage when and how to deploy the foldable wings. You can think of it as the drone’s built-in instinctual reflexes, like how a real squirrel would instinctively adjust its glide to avoid an owl or make a daring leap. The TWCC algorithm coordinates the wing deployment with propeller thrust to expand the drone’s set of controllable movements, particularly the pesky vertical ones that are typically challenging for standard quadcopters.

To accurately predict how air will behave around the wings in real-world flight, the team developed a physics-assisted recurrent neural network. This machine learning model was trained on real flight data to dynamically adjust the angle of attack — that is, how the wing slices through the air — based on actual aerodynamic behavior. By doing this, the drone can make smarter in-flight decisions and better anticipate how folding or unfolding its wings will affect performance.

In outdoor tests involving dynamic trajectories and virtual obstacles, the flying squirrel drone demonstrated a 13.1% improvement in tracking accuracy over conventional drones.

By blending a bio-inspired design with a cutting-edge control system and machine learning, the researchers have created a drone that does more than just fly — it adapts, reacts, and performs with a level of agility that brings it one step closer to what is seen in the natural world. If the future of drones looks anything like the flying squirrel, then it would seem that we are on the right path.