Insect Drones

May 11, 2015, 8:23 a.m.


Travis Rathbone

Five years ago, Richard Guiler and Tom Vaneck were sitting at a bar a few blocks from their office, trying to take their minds off work. For nearly a year, the two engineers had been struggling to develop a durable drone that could dodge objects, navigate inside buildings, and fly in stormy weather. They’d tried fixed-wing models, but adding enough sensors to effectively detect obstacles made them too heavy to fly. They’d tried helicopters, but the rotors kept getting tangled in branches and electrical wires. They’d even built a motorized balloon; all it took was a gentle gust of wind to blow it off course.

As they sat nursing their beers, Guiler and Vaneck watched as a fly appeared to slam into a window. Instead of breaking apart on contact as their drones did, the insect bounced off the glass and recovered. Then it did it again.

“It was an epiphany,” says Vaneck, who works for the Massachusetts research and development company Physical Sciences Inc. (PSI). “We realized if we could make a manmade system that could hit things, recover, and continue on, that’s a revolution.”

The idea of borrowing designs from nature is far from new, particularly when it comes to flight. The ancient Greeks dreamed up Daedalus, who fashioned wings for his son (which unfortunately worked a little too well). Leonardo da Vinci sketched a human-powered ornithopter. But until recently, inventors lacked the aerodynamics expertise to turn diagrams into mechanical versions of something as quotidian as a fly or a bee. As technology has advanced, scientists have decoded many of nature’s secrets. And engineers have developed the first flying, insect-inspired vehicles, opening the door to an entirely new class of machine: the microdrone.

“Nature has a several-hundred-million-year lead time on us when it comes to great design,” says Peter Singer, a fellow at the Washington, D.C.–based Brookings Institution. “The robots you know tomorrow are going to look like nothing you know today. More likely, they will look like the animals around you.”

Unraveling The Mystery Of Flight

Although insects and their relatives represent roughly 80 percent of the world’s animal species—some 900,000 known types—the mechanics of their flight had long been an enigma. Traditional fixed-wing aircraft rely on a steady flow of air over the wings. The same is true of helicopters and rotors. But as the wings of insects flap back and forth, the air around them is constantly changing. And the stubby wings of bees and other insects lift far more weight than can be explained using conventional steady-state aerodynamics principles.

Engineers have developed the first insect-inspired vehicles, opening the door to an entirely new class of machine: the microdrone.

Before scientists could understand flapping flight, they first had to see it in the minutest of detail. In the 1970s, Torkel Weis-Fogh, a Danish zoologist at the University of Cambridge, used high-speed photography to analyze the exact wing motions of hovering insects and compare them to the insects’ morphological features. From this, he formulated a general theory of insect flight, which included what he called the “clap-and-fling effect.” When insect wings clap together and then peel apart between the up and down strokes, the motion flings air away and creates a low-pressure pocket. Air then rushes back into the pocket, forming a swirling vortex. This vortex creates the force necessary to lift the insect between wing flaps. Similar vortices might be generated by the angle and rotation of the wings, Weis-Fogh posited, providing additional lift.

Two decades later, computational techniques caught up with theory, and scientists began to apply these principles to manmade systems. Charles Ellington, a Cambridge zoologist and former Weis-Fogh student, built a robotic wing that could precisely mimic the movements of a hawk moth. He placed it in a wind tunnel filled with smoke so that as it flapped, he could analyze the fluid dynamics. At the University of California at Berkeley, neurobiologist Michael Dickinson built a robotic fruit-fly wing that likewise mimicked a fly’s natural motion, and he submerged it in a two-ton tank of mineral oil. Working independently, the researchers
characterized the aerodynamics of flight with unprecedented specificity.

Dickinson and electrical engineer Ron Fearing won a $2.5-million DARPA grant in 1998 to apply these principles to a fly-size robot. They assigned a graduate student named Rob Wood, among others, to help develop techniques to fabricate the tiny parts and painstakingly assemble them with a pair of tweezers. Dickinson and Fearing also communicated which aerodynamics insights the students should try to reproduce. “Flies have really complex wing trajectories. There are a whole bunch of subtle things that happen,” Wood says. “Michael told us the most important features to generate vortices and other aerodynamic effects.”

By the time Wood graduated in 2004 and opened his own lab at Harvard University, he had helped pioneer a way to use extremely energy-efficient, exotic materials to replicate the motion of a fly’s wing; he had built a gyroscope that could mimic the sensors insects use to detect body rotation; and he had invented methods to manufacture complicated systems on a miniature scale. What remained was to put it all together into a working insect-size flying machine.


Kevin Ma and Pakpong Chirarattananon, Harvard Microbotics Lab

RoboBees, built in a Harvard robotics laboratory, are actually modeled after flies. Piezoelectric actuators that expand and contract with electricity flap the wings at 120 times per second. The wings can also be controlled independently.

Turning Insights Into Robots

On a freezing day in 2006, Wood arrived at his Oxford Street laboratory at Harvard. On the workbench sat a 60-milligram robot with a three-centimeter wingspan and a thorax roughly the size of a housefly. It was tethered to a six-foot-tall computer rack crammed full of high-voltage amplifiers and data-acquisition equipment. Wood carefully checked the connections and signals.

Then he flipped on the power and watched as the wings of his tiny creation began to vibrate, lifting the robot into the air for several seconds. Wood jumped in jubilation. It had taken him seven years to get to this point, and it would take another five to reach his next breakthrough: sustained flight along a preprogrammed path. An e-mail with proof of that milestone arrived in his inbox at 3 a.m. in the summer of 2012. An ecstatic graduate student had sent a video update on the lab’s latest prototype, now named RoboBee. It showed the delicate machine rising into the air and demonstrating, for the first time, stable hovering and controlled flight maneuvers in an insect-scale vehicle.

“I didn’t end up sleeping the rest of that night,” Wood says. “The next morning, we had champagne and all that, but it was more of a relief. If we couldn’t do this, we would have realized we were doing something wrong the whole time.”

Wood has pioneered microscale robotic flight; other researchers have used flapping-wing dynamics to reduce the size of aerial vehicles capable of carrying payloads. In 2011, California-based AeroVironment demoed its Nano Hummingbird. The aircraft has a 16.5-centimeter wingspan; it can fly vertically and horizontally and hover in place against gusting wind. It weighs 19 grams—lighter than some AA batteries—but it carries a camera, communications systems, and an energy source.

Insect Aerodynamics

Trevor Johnston

1. As a wing flaps, a tornado-like spiral of air forms along its leading edge. This vortex causes air pressure to drop briefly above the wing; higher pressure pushes the wing from below. 2. The wing rotates in preparation to flap in the opposite direction. This rotation creates forces similar to backspin on a tennis ball, pulling a faster stream of air over the top surface. 3. As the wing moves in the opposite direction, it collides with the swirling vortex of air created by the previous stroke, a principle called wake capture. Depending on the angle of the wing as it hits the wake, it can generate additional upward or downward force.

TechJect, a company that spun off from work done at the Georgia Institute of Technology, recently unveiled a robotic dragonfly with a six-inch wingspan. It weighs in at 5.5 grams (lighter than a quarter) and can be outfitted with modular electronics packages enabling high-definition video and wireless communication. The TechJect Dragonfly takes advantage of an aerodynamics principle called resonance. When wings flap at their most efficient frequency—which happens when air density, wing speed, and an organism’s weight are perfectly balanced-—they create waves of vortices that merge and build. The audible result is the hum of a hummingbird or buzz of a bee, says Jayant Ratti, TechJect’s president. A flapping-wing drone utilizing resonance generates significant improvements in energy efficiency, creating optimal lift with minimal effort.

Ratti and his team made the product commercially available to hobbyists and early adopters last year, and they plan to release another version by the end of 2014 for other markets. “The acceptance has been phenomenal,” Ratti says. “It is not yet a mature technology, but it’s getting there. We are still getting feedback and making improvements.”


Physical Sciences Inc.

The InstantEye, built by Physical Sciences Inc., has shock absorbers that mimic those on a fly’s body.

Building A Tougher Drone

Small, fragile drones don’t solve the problem of damage caused by unexpected impacts, and so Guiler and Vaneck have focused on durability. After observing the fly at the bar, the two engineers searched for someone with experience replicating insect flight. They teamed up with Wood, whose lab had since joined Harvard’s Wyss Institute for Biologically Inspired Engineering, and together they applied for an Air Force grant. Wood’s group then used an image-capture system to record and analyze fly behavior before, during, and after collisions with glass. By closely observing the positions of the flies’ body parts, they could measure the exact flip and twist of wings and legs.

When Guiler and Vaneck slowed down the film, they were amazed at what they saw. “I thought the fly would tumble a bit and lose a lot of altitude,” Vaneck says. “But the fly recovery was elegant. It happened so rapidly; it was breathtaking.”

Guiler and Vaneck homed in on the idiosyncratic geometry of the fly’s body. Its exoskeleton had accordion-like parts that acted as shock absorbers. It also seemed to sense impending collisions. Just before the moment of impact, the fly flew at an angle that ensured its legs touched the glass first. At that instant, the wings froze. Every time the fly slammed into the window, it reflexively surrendered to the crash momentum and fell. But within milliseconds, the fly’s center of gravity appeared to pull the fly back into a stable position. Then its wings flapped again, propelling the insect into a controlled hover. “It can hit and recover in two or three wing beats, which is phenomenal,” Vaneck says. “There is no manmade system that can do that.”

The Coming Swarm

Courtesy KMEL Robotics/YouTube

Single bio-inspired drones are useful, but dozens can work together to accomplish a complex task. Vijay Kumar, an engineer at the University of Pennsylvania, teamed up with Arizona State University biologist Stephen Pratt to apply three lessons learned from ant swarms to fleets of quadrotors. 1) In nature, ants act autonomously. Engineers traditionally use a centralized system to choreograph movement in swarms, Kumar says. As a swarm grows larger, the control algorithms become increasingly complex. Instead, Kumar tries to program his aerial vehicles with a common set of instructions; the quadrotors divide up tasks and assume complementary roles. 2) Individual ants are interchangeable. “If I want to scale up my swarm, maintain the predictability of its behavior, and make it robust, the gang has to be able to perform the task if an individual is knocked out,” Kumar says. So he makes his aerial vehicles identical to one another. 3) Ants sense their neighbors and act on local information. Kumar outfitted his vehicles with motion-capture systems, cameras, and lasers that enable them to avoid obstacles and maintain a set distance from each other. As a result, they can fly in tight formations, work together to pick up heavy objects, and collaboratively create a map of their environment.

The two engineers used those insights to guide the development of a resilient flying machine. The body needed to be shockproof, and the wings needed to be controlled independently. So they designed a shell for a quadrotor that incorporated shock absorbers—rubber dampers in between sections made from carbon fiber and plastic. They gave each of the four rotors its own motor in order to mimic the alternating wing speed that provides four-winged insects with exceptional control. When the vehicle is blown out of position or clips an obstacle, its computer detects the discrepancy between its current position and its programmed flight path, and an autopilot reflexively kicks in to recover stability.

Last February, the engineers sent their drone, called the InstantEye, to Fort Benning near Columbus, Georgia, for its annual Army Expeditionary Warrior experiments, where an infantry platoon used it to help complete a set of assigned missions. The soldiers gave it a “green” rating, one of the highest available.

An early prototype of the RoboBee

Pratheev Sreetharan, Harvard Microrobotics Lab

Overcoming Future Hurdles

As the first generation of microdrones reaches the market, significant engineering challenges still remain. For Wood, the big hurdle is power. Unlike the much larger InstantEye, Nano Hummingbird, and Dragonfly drones, RoboBees must be connected to an external power source. Wood is using microfabrication to try to shrink onboard batteries, and he’s collaborating with researchers at Harvard, the University of Washington, and the Massachusetts Institute of Technology to pursue novel batteries, micro fuel cells, and wireless power transfer. He estimates he is only one or two years away from his first autonomous-power demonstration.

Guiler and Vaneck aim to replace the propellers on their quadrotor with flapping wings. The InstantEye is far better at recovering from wind gusts and minor collisions than other drones are, but its propellers can still get tangled in branches or power lines. “We wanted to bring something to the field fast,” Guiler says. “But what we discovered was flapping-wing birds and insects are perfectly suited for environments where you have dynamic obstructions—the trees are moving, the branches are moving. If they do get stuck, by their very motion they get unstuck. They kind of beat their way through. We realized a flapping wing was the only thing that would work.”

RoboBees could search disaster sites for survivors, monitor traffic, or pollinate crops.

And then there’s Dickinson, who initiated the project to build the robotic fly. Today he runs a lab at the University of Washington and works with advanced imaging systems to study insect flight. Early high-speed cameras captured about 3,000 frames per second. “Fifteen years ago, the flies looked like little fuzzy UFOs,” he says. Now the biologists use cameras that can run at 7,500 frames per second, significantly higher than what was once available to researchers, and that work in infrared light. (Update: This sentence has been rewritten for clarification. Other cameras can reach much higher fps.) Dickinson has also gone beyond analyzing flight; he’s using electrodes to record the activity of neurons in insects’ brains. He links them to a flight-simulation system and presents them with visual stimuli—a picture of a predator, for instance—that cause them to react. “We can begin to learn how neurons in the brain are processing information in flight and how sensory information is transformed into action,” Dickinson says. “The stuff that made Rob’s work possible was just the basic mechanisms by which animals keep themselves in the air. Now we are going beyond that to understand how flies steer and maneuver.”

Learning how nature creates superior sensors could lead to lighter, smarter drones. And as that happens, their range of applications will grow. Guiler and Vaneck plan to sell the InstantEye to the military and law enforcement. The British Forces have recently begun using a microdrone, a hand-launched helicopter called the Black Hornet, to scout for insurgents in Afghanistan. Microdrones may also have uses closer to home. They could allow police and SWAT teams to gather footage inside office buildings or banks and between skyscrapers, where winds typically gust.

Wood visualizes an even more diverse array of uses for RoboBees. A box of about 1,000, he notes, would weigh one pound. They could easily be shipped to a disaster site and deployed to search for survivors. They could also monitor traffic or the environment and help pollinate crops. Research scientists could use them to gather data in the field.

Whatever their application, microdrones are no longer a da Vinci–like dream of engineers. They’re taking off—agile, resilient, and under their own power.

Adam Piore’s last feature for Popular Science_, about savants, appeared in the March 2013 issue._

This article originally appeared in the January 2014 issue of _Popular Science.

<< News Archive