Moth-like drone navigates autonomously without AI

Researchers at the University of Cincinnati are developing a drone with flapping wings that can locate and hover around a moving light like a moth to a flame.

UC College of Engineering and Applied Science Assistant Professor Sameh Eisa and his aerospace engineering students are interested in these unusual drones because of their highly efficient flight, which can be scaled down for use in covert surveillance.

Moths have the incredible ability to hover in place or even fly backward. They automatically make fine adjustments to compensate for wind or obstacles to remain stationary or follow a moving object. Likewise, Eisa’s mothlike drone makes fine adjustments to maintain a desired attitude and distance from a light, even when the light moves, Eisa said.

“The reason we’re interested is size. It’s a more optimal design. These small robots would have to fly like a moth,” he said.

In his Modeling, Dynamics and Control Lab, Eisa explores animal-inspired engineering. Previously, he examined drones that could harness the power of dynamic soaring to cover vast distances more efficiently, much like an albatross.

For his latest project published in the journal Physical Review E, Eisa and doctoral student Ahmed Elgohary theorized that hovering insects are able to fly so adroitly because they employ the equivalent of extremum-seeking feedback systems.

These systems allow for drone navigation in real time without complex calculations, global-positioning equipment or artificial intelligence by simply making constant adjustments to control inputs such as the number of flaps per second.

Flapper drones control roll, pitch and yaw by flapping the wings independently. But this independent flapping is too fast for the naked eye to observe. Instead, the wings look like the blur of a hummingbird’s wings.

“Our simulations show that extremum-seeking control can naturally reproduce the stable hovering behavior seen in insects—without AI or complex models,” said Elgohary, the study’s lead author.

“It’s a simple feedback, model-free and real-time principle that could explain how these small creatures achieve such agility with very limited brainpower.”

The drone simultaneously measures the performance of whatever function it is programmed to optimize, like finding a light source, to correct its course in a constant feedback loop that allows for remarkably consistent and stable flight.

How stable? The drone was able to match the subtle but unique back-and-forth sway of each of the hovering insects it was designed to mimic: moths, bumblebees, dragonflies, hoverflies, craneflies, along with hummingbirds.

“Moths make it look easy,” Eisa said. “The reason we use extremum-seeking techniques is because they seem to be biologically plausible.”

Hovering insects like the nectar-loving hummingbird clearwing moth move their wings in a unique figure-of-eight motion that allows them to get lift on both the downstroke and upstroke of their wings. The flexible wings deform during each wing beat to maximize lift and maneuverability.

Elgohary and UC graduate student Rohan Palanikumar used a remote controller to demonstrate how the flapper drone flies in Eisa’s flight lab, which is surrounded by soft netting to protect both drones and people from inadvertent crashes. The drone has four “wings” made of wire and fabric.

Controlling the sensitive drone manually is much harder and less reliable than using its own extremum-seeking system, Elgohary said. But once activated, the flapper drone lifted into the air and hovered in place, if a bit wobbly. This wobble is intentional and provides the perturbations the system needs to evaluate changes in performance so it can constantly course correct to optimize its flight.

Eisa said the research is interesting not only for what it might mean for new autonomous unmanned aerial vehicles but also how these tiny insects manage their miraculous aerobatics with brains the size of a grain of pollen.

“It could change a lot of things about biophysics. If it is the case that hovering insects like moths use the equivalent of our extremum-seeking feedback, it probably evolved in other creatures as well,” he said.