The Berkeley Science Review: Articles

Robot Flea Circus
Berkeley engineers build bionic bugs (view PDF)
by Tracy Powell

By the time it has exploded into a Rorschach smear on the windshield, you’re more likely to turn on the wiper fluid than ponder the marvelous engineering that propelled a fly there in the first place. Similarly, anyone dancing a spastic limbo to avoid an oncoming bee is unlikely to notice, let alone admire, the sheer aerodynamic improbability of its onslaught. To most of us, these buzzing, careening little motes of life are disgusting, threatening, or simply beneath notice.

In contrast, a handful of researchers have come to see bees, fleas, flies, and their creeping, crawling kin as inspiration—minute miracles of engineering whose physical capabilities outstrip anything we humans have accomplished. Before getting intimate with your windshield, for example, that fly could take off backwards, land upside down, or turn at right angles in the blink of an eye—feats your average aeronautical engineer would kill to achieve. Similarly impressive, a flea will routinely jump 150 times its own body length, in the process generating g-forces that would kill a human.

How do they do it? Biologists are probing the mechanisms of movement that allow plump bugs to traverse highways and lumbering bumblebees to maraud passers-by. Engineers, in turn, are looking over biologists’ shoulders, translating newly elucidated principles of insect locomotion into mechanized modules. In doing so, they have spawned a veritable menagerie of tiny robotic creatures. Though still in their technological infancy, many of these mobile robots are taking their first halting steps, hops, and wingbeats in labs across the UC Berkeley campus.

Flight, in C# Major
On the computer screen, a single, disembodied wing anchored to a slab of plastic flaps frantically, its frenzied beating visible only in slow motion. Though the sheer mass of the plastic renders the wing’s exertions futile—that slab isn’t going anywhere, let alone flying off toward the ceiling—the flapping nonetheless generates a tiny amount of lift. According to the jeweler’s scale the plastic rests upon, the slab’s weight lessens very slightly as the wing labors to haul it into the air.

Experimenting with dismembered wings may sound like the province of scientific sociopaths, but in fact, no insects were harmed during the making of this motion picture. Rather, the video chronicles the test of a robotic wing pioneered by researchers in Professor Ron Fearing’s laboratory in the Department of Electrical Engineering. The wing was ultimately able to generate up to 1400 micro-Newtons of lift—which, while it won’t elevate a wallet-sized chunk of plastic, should provide ample force to propel a robot lighter than a paperclip into the air. To engineers working on the micromechanical flying insect (MFI) project, this was a significant advance.

MFI began in 1998 as a collaboration between Fearing’s research group and Michael Dickinson, a MacArthur Award recipient and erstwhile UC Berkeley professor (since relocated to CalTech). Dickinson’s inventions, which include RoboFly, Bride of RoboFly, and Fly-o-Rama, helped illuminate important aspects of insect flight mechanics and navigation. Taking inspiration from these insights, a long succession of Fearing’s graduate students and postdocs have since attempted to translate theory into practice, working to design and launch a robot insect.

After almost a decade, the culmination of their efforts is an airy black lattice of carbon-fiber beams sporting a tiny, translucent wing at either end; at 25 millimeters from wingtip to wingtip, the whole contraption is about the width of a quarter. “People always ask, ‘Why are the wings so small?’” says Erik Steltz, a graduate student who works on the project. He speaks with the determined good nature of someone repeatedly told that the miracle of cutting-edge engineering to which he has devoted four and a half years of work is, well…a little puny-looking. “They’re not too small! They may look small relative to the body of the robot, but because the body is mostly air, the wings are the right size to lift what the robot weighs.”

Steltz’s indignation is understandable; while the wings may look ineffectual, the engineering behind them is unquestionably powerful. Together, the two wings produce nearly three times more lift than necessary to get a 100 milligram robot off the ground. The researchers obtained such high forces by mimicking the stroke of a bee’s wing, flapping it at an amazing 275 wingbeats per second. (In musical terms, this frequency produces something close to a C#.) When activated, “It actually sounds like a fly that buzzes by your ear,” says Steltz.

The key to their robot’s flapping prowess is the motor, or actuator, an ingenious little device that sweeps the wings up and down. Only one centimeter long, it consists of a thin panel of carbon fiber sandwiched between two pieces of piezoelectric (PZT) material. According to Steltz, “The molecules in a PZT material scrunch up together when you apply a voltage across them—all the material gets a little shorter. Then, when you remove the voltage, they go back to their regular, relaxed position.” When you apply a charge to one of the actuator’s PZT layers, its contraction bends the whole sandwiched structure toward it. If you then switch the voltage to the other PZT layer, the structure bends back in the other direction. Thus, by alternately applying charge to opposite sides of the actuator, you cause it to flap back and forth.

Although the range of motion generated by the actuator itself isn’t large (each only moves about 0.5 millimeters at the tip), the motion is amplified by tiny carbon fiber bars that translate the initial mechanical stimulus into a larger sweep at the end of the wing. By attaching two actuators to each wing and operating them in sequence, the researchers are attempting to recreate the powerful, figure-eight motion of an insect wing.

The robot’s structural components are generated using techniques pioneered by the Fearing lab. First, the two-dimensional shape is cut with a laser out of a thin slab of carbon fiber—a very strong, lightweight material also used in snowboards and high-end bicycle frames. Two pieces of this carbon fiber are layered around a thin film of strong, flexible plastic that acts as a joint, and Steltz then assembles the robot by folding and gluing the two-dimensional scaffold into its final, three-dimensional form. “It’s like carbon-fiber origami,” he explains.

Though Steltz has not yet progressed beyond testing the robot’s individual flight components, Harvard professor Robert Wood, a collaborator and former graduate student of Fearing’s, has been more successful. In an exciting development, he recently published a paper detailing his robot’s maiden flight. (video footage can be found at technologyreview.com/Infotech/19068.) Wood’s robot sported a streamlined design, requiring only a quarter as many joints and a single actuator, rather than two per wing. With fewer parts to malfunction, the robot managed to hoist itself a few vertical centimeters along a pair of vertical guideposts. However, a more complicated structure will eventually be required for controlled, directional flight.

In the short term, Steltz hopes to soon launch his own, more intricate prototype. Longer-term goals for the project include replacing the wires with a lightweight onboard battery and, eventually, removing the guideposts currently required to hold the fluttering robot steady. “Only if the two wings are absolutely, perfectly matched will it go straight upwards,” he explains. “For a hand-assembled prototype, this just isn’t going to happen.” More precise assembly techniques may address this issue, but the underlying problem of controlled flight, as opposed to a simple guided launch, is more complicated.

After all, flies may be very stupid, but they have an incredibly sophisticated navigational system hardwired directly into their meager little brains (see sidebar, p. 26), and any robot that’s going to do more than trundle up and down guideposts on a lab bench had better have the same. To this end, Fearing’s lab has begun to develop small-scale sensors that could be deployed on an airborne robot to help control its flight. Simple photodiodes aimed at different angles, for instance, can mimic the fly’s ability to distinguish up from down. An early prototype for a robot gyroscope, developed by Wood during his time in the Fearing lab, may one day be refined into a navigational aid.

Though it may be tempting to pile batteries, gyroscopes, and all manner of flight-enhancing gadgetry onto the MFI, the added mass of any extra hardware must be offset by increased lift generated by the wing—clearly, engineering on such a small scale has its challenges. But while Steltz’s robot may seem diminutive at 100 milligrams, some researchers have been even more ambitious in their quest for miniaturization. Using a very different approach, Steltz’s fellow graduate student Sarah Bergbreiter has managed to develop a jumping robot that is four times smaller and ten times lighter.

Honey, I shrunk the robot
Just down the hall from the Fearing lab, the walls of Professor Kris Pister’s laboratory are lined with ranks of small plastic drawers, each one rattling with resistors, capacitors, and other mechanical viscera that, when assembled, sustain some of the lab’s larger robots. Bergbreiter, a student in the lab, has spent the last several years here amongst the electronic bric-a-brac, working to develop a tiny robot (about one-sixth the size of a postage stamp) capable of hopping around under its own power.

Like the MFI project, her work fits into a larger theme in modern robotics: mobile sensor networks, which aim to deploy swarms of tiny, sensor-carrying robots into hostile or unfamiliar environments. Between the discreet stature of the robots and the Department of Defense’s early sponsorship of the field, mobile sensor networks inevitably conjure up images of espionage—flocks of creeping automatons sidling up to eavesdrop on enemy strongholds. But beyond these predictable clandestine scenarios, researchers also envision applications in search-and-rescue efforts and environmental monitoring. Equipped with sensors for humidity, temperature, or toxic compounds, robots could scuttle about monitoring weather and pollutant levels across a geographical range, or flit through a burning building to assess its safety for firefighters.

With an eye to enabling this sort of roving sensor network, Bergbreiter has devoted years to developing various mobile robots. After encountering a 2003 Nature article that described the jumping prowess of the spittlebug, she determined that hopping could be a particularly useful form of movement. Based on the article’s measurements of physical forces produced by the insects while jumping, she explains, “It didn’t seem too far off from what we could accomplish from an engineering standpoint.”

Furthermore, the ability to jump was actually quite relevant to one of the main challenges in small-scale robot mobility. “If you’re talking about shrinking a robot down to millimeter size, being able to move is really a challenge because everything else around you proportionally starts to seem a lot larger,” says Bergbreiter. “You have a lot of obstacles to deal with, and if you can deal with them mechanically, by jumping over them, it makes your life a lot easier.” She therefore set out to use insect jumping as a model for this sort of obstacle-avoidance system.

Bergbreiter knew that the components necessary to faithfully recreate an insect’s jumping mechanism would be particularly difficult to fabricate on a small scale (see sidebar, this page). Besides, while nature has produced some astonishing mechanisms for movement, a spittlebug’s spring-loaded leg among them, she suggests that scientists and engineers should not feel compelled to precisely emulate natural systems. Steltz echoes this sentiment. “This is something we fight about, when people ask why we don’t do it the exact same way as the insect…they are kind of missing the point,” he says. “A cheetah is fast, but cars can go faster. Why didn’t nature put wheels on a cheetah? Nature can only use biologically compatible materials and processes; with an engineering solution, we are not constrained by those same limitations.” In this spirit of ingenuity, Bergbreiter improvised a novel strategy to launch her robots, one that exploits tension.

To understand the mechanism, imagine a middle-school miscreant shooting a garden-variety rubber band: He loops the band over a finger to anchor it, and stretches it back (taking aim at a suitably helpless target). As it extends, stored energy builds along the length of the orange rubber. Once the band is released, this stored energy launches it toward the aiming fingertip, speeding it inexorably toward its victim.

Bergbreiter’s jumping microbot will operate on a similar, though less malignant, principle. A tiny, figure-eight-shaped strip of rubber is looped at either end around two tiny hooks, each about the width of a human hair. One hook is held stationary (like the finger aiming the rubber band) and pointed at whatever surface the robot is sitting on. The other hook, which is attached to a stiff leg, is winched slowly backward to stretch the band. To jump, the mobile hook is released. The relapsing band yanks the hook and attached leg rapidly down, slamming the leg against the surface and propelling the robot into the air. (Wheee!)

Though simple in principle, fabricating this mechanism on such a small scale proved difficult. First, each hair-thin rubber band had to be individually made by hand. Though initially done by cutting bands out of a thin sheet of rubber with a hand-held laser pointer, the resulting uneven, scalloped edges weakened the bands, causing them to snap when stretched. Pouring liquid rubber into a band-shaped mold and allowing it to cool into the proper form eventually produced bands that could stretch without snapping.

Even more challenging was the design and fabrication of a motor capable of applying enough force, on a small enough scale, to stretch the rubber band to twice its original length—far enough to propel the robot 20 centimeters (about 30 times the robot’s length) into the air.

The difficulty of this task was due largely to scale. Because of their extremely small size, microelectromechanical systems (MEMS) like the jumping microbot are subject to strong physical forces that make their components stick together, a difficulty that must be circumvented in the design process. Furthermore, in order to achieve such mechanical intricacy on a scale of mere millimeters, these systems must be made via a complicated process in which silicon wafers are subjected to harsh chemical baths, bombarded with charged particles, and baked in ovens at temperatures up to 850°C (1,562°F).

In Bergbreiter’s case, the mechanism resulting from this process was called—appropriately enough for an insect-inspired project—an inchworm actuator. It takes advantage of electrostatic force, in which an electrical charge applied across two conductive plates generates an attraction between them. If one plate is held in place and the other allowed to slide, the mobile plate will scoot across a small gap toward its stationary partner. Essentially, the inchworm actuator is like a tiny, electrostatic bucket brigade: A series of electrostatic attractions drags the hook from one plate to the next along a linear array, ratcheting the attached rubber band to greater and greater lengths.

Even with the fabricated bands and actuators in hand, Bergbreiter still had to assemble each device. This involved hunkering down over a stereo microscope, tweezers in hand, threading a whisker-thin rubber band around two equally diminutive hooks—no simple task. “It turns out it’s a lot like playing the game ‘Operation’, which I was terrible at as a kid,” she quips. Caffeine was a definite no-no: “I had a Diet Coke at lunch one day, and then I tried assembling some rubber bands; it was a complete disaster. You can shear off the hooks if your hands shake.”

Persistence, practice, and abstention from caffeine eventually led to a collection of properly threaded jumping mechanisms. By stretching the rubber band with a force meter, Bergbreiter has shown that the mechanism can produce enough energy to propel her robot into a jump. She has also demonstrated that a small solar cell can power the inchworm motor and its control mechanism. Unfortunately, amidst all the solar cells, ricocheting bands, and ratcheting inchworms, one crucial task has yet to be done: Put it all together. As she moves on to an assistant professorship at the University of Maryland, Bergbreiter seems eager to finally assemble all these robotic bug bits into a functional unit, laughing, “I envision 50 little robots jumping around happily on my desk.”

But getting a little wafer of silicon to hop itself off the ground is only half of the problem. What about landing? It is discouraging to envision thousands of dollars’ worth of experimental robot jumping once and splattering at the end of a single, 20-centimeter bounce. This is no trivial consideration. Silicon is notoriously brittle, and the froghopper that inspired Bergbreiter’s work experiences as much as 400 times the force of gravity (gs) while jumping. (Human fighter pilots, even swaddled in pressure suits to help withstand the force of high-velocity flight, struggle to remain conscious after only a few seconds at 7-10 gs; 15 gs or more can be lethal.) Bergbreiter estimates that her robots, which should accelerate to a speed of 1–2 meters per second over mere milliseconds, will only experience tens of gs, but this is still a considerable strain.

“Right now, I’m not even worrying about the whole landing thing, or subsequent jumps, but that’s obviously the next challenge,” acknowledges Bergbreiter. Having pioneered the combination of a rubber component (the jumping band) with hard robotic elements, she speculates that further use of rubber could help cushion the brittle silicon components of her microbot, enabling a softer landing.

But must a microbot that goes up necessarily come down? As Bergbreiter points out, her jumping mechanism would be a great launch strategy for flight, perhaps helping a robot like Fearing’s get airborne. Jumping could also be combined with other forms of mobility—for instance, allowing a crawling robot to leap over an obstacle before scuttling on its way. Given the variety of mobility research conducted at UC Berkeley, such combinations may be just around the corner.

The coming robot army
The fly and the froghopper may already have robotic knockoffs, but many strategies of biological motility studied on campus have yet to be mined for inspiration. Professor Robert Full, for instance, runs the Department of Integrative Biology’s PolyPEDAL lab, aptly named for the many forms of movement studied by its members. From cockroaches and centipedes to lobsters and geckos, Full investigates the many tactics biological organisms use to get from point A to point B. (One of their research organisms, a stomatopod, even travels by somersaulting!)

As Full’s group sheds light on these biological mechanisms, engineers are eagerly moving in to reinterpret them in silicon and carbon-fiber. A small, fiberglass model of a robot cockroach in the Fearing lab—skin-crawlingly apt in its sprawling, six-legged posture—represents one nascent collaboration between the two groups.

Could bionic bees, fleas, and centipedes one day rub elbows with their biological brethren? Will sensor-laden robot cockroaches soon be scuttling through war zones, burning buildings, and the environment at large? Fearing’s and Pister’s small-scale robots may be just the leading edge of a future swarm of mechanized bugs, and as Bergbreiter points out, “Whether it’s crawling critters, or jumping critters, or flying critters, this is where a lot of this microbot research is happening.” It may be time to hunker down and invest in a good fly swatter, as campus researchers come closer to realizing their vision of a miniature, mechanized insect world.

Tracy Powell is a graduate student in plant and microbial biology.

Want to Know More?
Check out the Micromechanical Flying Insect: robotics.eecs.berkeley.edu/~ronf/MFI

Jumping Microbots:
www-bsac.eecs.berkeley.edu/~sbergbre/microrobots

POLY-PEDAL lab:
polypedal.berkeley.edu/twiki/bin/view/PolyPEDAL/WebHome

Comments on this article? Drop us a line at with 'letter to the editor' in the subject!