Popular Science, December 5, 2002
In the basement of the Valley Life Sciences Building at the University of California, Berkeley, biologist Michael Dickinson walks down a cinderblock hallway to an anonymous steel door. Beyond it lies a small, windowless room crammed with high-speed video cameras and lasers and computer cables draped as thick as cobwebs. In the center of the room is a glass tank big enough to hold a vending machine. This is Robofly.
The tank looks empty, except for a piece of plastic shaped like an insect wing, which dangles from a mechanical arm. But when Dickinson turns on a filter, the tank sprays a creamy mist of bubbles. The tank, it turns out, is filled with 2 tons of mineral oil. Dickinson taps on a keyboard, and slowly the wing begins to move through the oil. It swings forward and back, transforming the bubbles from a disorganized cloud into slow-motion swirls and falling curtains of diamonds.
Dickinson stands before Robofly, carefully watching vortex after vortex. A real fruit fly-an escapee from some other experiment, no doubt-drifts by, its wings beating 200 times a second and stirring up tiny, invisible vortices of their own. But Dickinson ignores it. Robofly has been his obsession for more than a decade, ever since he built the first prototype in Germany. Back then, he used sugar syrup in the tank. “There was sugar everywhere in the lab,” Dickinson recalls, “and the maids went on strike. They refused to clean our lab until my advisor did something about the sticky American.”
Stickiness is a good way to sum up Dickinson’s approach to science. Once he attaches himself to a problem he doesn’t let go until he’s solved it. For almost his entire career, he has been stuck on one seemingly simple question: How do flies fly?
Though engineers figured out decades ago how to build airplanes that cross oceans, the aerodynamics of insects continue to baffle them. The way an airplane generates lift can be accounted for by a straightforward concept: The air streaming over the top of the wings exerts less pressure than the air below, and that imbalance keeps the wings aloft. But insects, though equipped with barely a brain to speak of, make complex maneuvers far beyond an airplane’s capabilities. They turn quicker than any fighter jet and land upside-down on ceilings. “These animals can move perfectly sideways, they can move backward and forward, they can rotate in place,” says Dickinson. “Every time we do an experiment, we wonder how the hell do these little sesame-seed-size nervous systems do this?”
That question led him to build not just Robofly-which is 100 times larger and 1,000 times slower than a fruit fly-but a growing family of strange and wonderful machines with names like Bride of Robofly, Fly-o-rama, and the Rock-and-Roll Fly Arena, all designed to help reveal the secrets of insect flight. The answers that Dickinson, whose e-mail handle is “flyman,” gleans from those devices might someday enable engineers to build automated craft the size of rice grains that explore other planets, fly into burning buildings to search for victims, or spy on military opponents-literally, a fly on the wall. “He’s got a great engineer underneath that biologist exterior,” says Ronald Fearing, a Berkeley electrical engineer who has collaborated with Dickinson on building robotic flying devices.
Last year, Dickinson’s research earned him a MacArthur “genius” fellowship, a grant of $500,000 over five years, with no strings attached. MacArthur fellows are selected for their exceptional creativity, and for track records that promise important future advances. Dickinson was one of the last fellows to learn about his award. He had been backpacking with his fiancee in Hawaii, 18 miles into a Kauai forest. When he returned to civilization, he checked his voicemail and found frantic messages to call his lab, but his cellphone conked out in the rain before he could return any calls. He finally found a pay phone at a state forest where he had gone to watch fruit flies swarming in guava trees.
Dickinson is a small 39-year-old who wears glasses and sometimes sports a beard thick enough to pass for a mountain man on his frequent hiking trips. He was introduced to flies in the mid-1980s, when he was at the University of Washington getting his Ph.D. in neurobiology. Fruit flies are the guinea pig of this field, because they have only about 500,000 neurons-compared with 100 billion in a human brain. A fly uses most of those neurons to gather sensory information, including light with its eyes, smells with odor-sensitive hairs, and balance with club-shaped gyroscopes behind the wings. Those signals get funneled through the nervous system, which then sends commands to the wings. The commands have to be simple yet exquisitely precise, because the time between wing beats is only a few thousandths of a second.
As a graduate student, Dickinson studied the natural strain gauges on flies’ wings that help them sense how much their wings are bending. But along the way, something began to bother him. “How can I understand what the sensors on a fly wing are used for if I don’t understand the forces a wing encounters?” he wondered.
For decades, scientists had been attempting to do just that by testing models of insect wings in wind tunnels. But in 1984, biologist Charles Ellington of the University of Cambridge examined the measurements that had been amassed and found that the numbers didn’t add up. No one had managed to explain even half of the lift that insects actually create. Since then, researchers have proposed a host of ways in which insects could be invisibly adding lift to their flight. Evaluating these theories has been impossible, though, because there are simply too many variables to consider: “No computer on the planet could tell us what the forces are,” says Dickinson. He decided to take a different approach: “Let’s just measure it.”
Dickinson knew he couldn’t measure the forces at work on a living insect in flight: Even the smallest man-made sensors couldn’t fit on a fly’s wing. So he decided to use a mechanical fly instead. The tricky part would be to make it experience the same forces a fly does. Air behaves very differently around an insect than around a large animal. To us, it’s delicate and slippery. But on the scale of fruit flies, it’s thick and sticky.
Working with Karl Gtz at the Max Planck Institute for Biological Cybernetics in Germany, Dickinson created his first robotic wing. He found that a 2-inch-wide wing flapping in sugar syrup would experience the same forces that a much smaller fruit fly wing does in the air. Dickinson and Gtz built the wing and designed a simple computer-controlled motor to make it flap back and forth. As the sensor-laden wing waved through the syrup, they flooded the tank with aluminum shavings. They then took video footage of the swirling shavings and compared it with the forces recorded by the sensors.
A fly, Dickinson discovered, uses a number of tricks to generate lift. One trick involves keeping its wings at steep angles. Instead of gliding smoothly over the wing, the upper stream of air forms a swirling vortex along the wing’s leading edge. This vortex lowers the air pressure above the wing, providing an extra upward boost.
Any pilot can tell you that tilting your wings at a steep angle is a risky strategy. The more steeply an airplane climbs, the harder it is for the stream of air traveling over the top of a wing to stay attached to the wing’s edge. When the stream pulls away altogether, the plane loses its lift and stalls. But flies have an advantage over planes: They don’t have to hold their wings in a fixed position. A fly flaps its wings back and forth so quickly that the vortex at the wing’s leading edge doesn’t have time to detach before the wings finish a stroke. At the end of each stroke, the fly rotates its wings so that it can flap them in the reverse direction. This creates a new vortex, while the old one slips off harmlessly-without causing a stall.
Later experiments with Robofly, whose wingspan measures 25 inches, revealed a second source of lift, which comes as a fly rotates its wings between strokes. A rotating object (such as a tennis ball hit with backspin) pulls air over its top surface, which reduces air pressure above the object, and pushes air in the opposite direction below, increasing air pressure there. This rotational force, which is also generated as a fly flaps its wings, can supply the insect with up to a third of its entire lift.
A fly can also generate lift from its own wake. When it sheds a vortex from its wings at the end of each stroke, the vortex drifts slowly away, still spinning. As the insect brings its wings back through the next stroke, Dickinson found, the wake pushes against the wings and lifts them up.
As Dickinson decrypted the physics of insect flight, he found himself drawn into the company of people who build robots. Recently he helped his Berkeley colleague Fearing create the Micromechanical Flying Insect, a blowfly-like device less than an inch long that’s being developed with funding from the Office of Naval Research and the Defense Advanced Research Projects Agency. So far, though, the robotic blowfly has only flown on a tether, and with just one wing. The main limiting factor, says Dickinson, is the battery, which is currently too large and too weak to make extended flight a reality.
Dickinson’s interest in robotics is as a biologist; he doesn’t aspire to be the next Orville Wright. He looks at flying devices as an opportunity to evaluate his notions about how animals function. “In biology, it’s rare that you have the opportunity to test your ideas by building something,” he says.
Dickinson recently moved his lab to the California Institute of Technology in Pasadena, where he and his students are continuing to develop machines to study insect flight, such as the Rock-and-Roll Fly Arena, a flight simulator for fruit flies. The arena is a hollow cylinder 6 inches in diameter; 12,000 light-emitting diodes line its inner wall. Dickinson’s team glues a fly to the tip of a steel rod at the center of the arena, leaving its wings unhampered. The wall surrounding the fly lights up in a shifting pattern of bars and boxes that tricks the fly into believing it is flying freely in an arena.
As the fly tries to turn, a camera detects the changes in wing motion and feeds that information to a computer, which quickly alters the lights on the wall. The fly “thinks” that it’s really turning, to steer clear of an obstruction. The “Rock-and-Roll” part of the arena’s name comes from the way the entire simulator can pitch and yaw, letting Dickinson and his students study how flies use their gyroscopes to steer.
The arena has helped Dickinson’s team develop a set of rules that they think govern how flies maneuver. As a fly moves toward an object, the object grows larger in its eyes. If it grows bigger in one eye than another, the fly turns to avoid it; if it expands directly in front, the fly stretches out its legs to land. To test that theory, Dickinson and his students have built a machine called Flyball (think eyeball).
Flyball consists of a video camera mounted on a system of tracks. It travels around an arena decorated with a random collection of white and black squares, sending what it sees to a computer. The computer uses Dickinson’s rules to choose where to go next. Dickinson hopes that it will make decisions to turn the way flies do. “It’s fine to make a diagram, but to actually build something you really have to put your money where your mouth is.” If he’s right, the camera will take the same flight path as a real fly. If not, it may crash into a wall.
Ultimately Dickinson hopes to take everything he’s learned about how flies fly and apply it to an even bigger question: how insect flight evolved. Insects probably developed wings from body scales more than 300 million years ago. Wings were the secret to their success-flying insects make up the vast majority of all known animal species on Earth. Since the origin of flight, insects have fine-tuned their anatomy to accommodate a range of flying styles, from the broad gliding wings of dragonflies to the fierce fighter-jet attacks of wasps. “You have to understand the mechanism before you can understand how behavior evolves,” says Dickinson.
But it is hard for Dickinson to imagine the day when he will have answered all his questions. “The more we study fly flight, the more we realize how little we understand,” he says.
A case in point: After making the rounds to see how his students are doing, Dickinson comes back to his office, where Mark Frye, a post-doctoral fellow in the lab, wants his help. Frye is writing a computer program to simulate an airborne fruit fly-essentially a dry run for Flyball. The simulated fly takes a path through a simulated arena, using the decision rules the team has put together. But it’s not working.
“If I only let them turn in one direction, it works great,” Frye says. But if he lets the simulated flies choose which way to turn, the results go haywire.
Dickinson offers some suggestions-perhaps make the insect respond not just to what’s right in front of it but also to what it saw a few moments earlier. They try to work out how Frye might turn that into a program, and after a while they fall silent, both a little exasperated. Then Frye utters three words that are something of a motto of Dickinson’s team, and often can be found scribbled on blackboards around the lab:
“Little (expletive) robots.”
Dickinson smiles and repeats the sentiment: “Little (expletive) robots.” These apparently simple creatures that, neurologically speaking, amount to little more than a few circuits, still manage to confound him at every pass. But in truth Dickinson’s motto is a term of affection. “Flies are wondrous things,” he says. “Probably every human on the planet sees at least one fly a day, and yet we don’t even notice them. Right under our noses are these extraordinary little machines.”
Copyright 2002 Popular Science. Reprinted with permission.