Microspies

“We will make our thrust here,” the colonel says. “The rocket launchers will be along the ridges. We will release the gas from the warehouse by the field—here. They will never think it is coming from there because the lights will be on. They will expect the gas to come from the water tower. All units must be in position before dawn. We will take them completely by surprise. Aah, these flies!”

He slaps his hand down so hard that coffee splashes from his cup, then inspects his palm with a satisfied expression before wiping it on his thigh.

Six miles away in a subterranean bunker a video monitor goes blank. A bespectacled lieutenant in a camouflage shirt sighs and gets up from her chair. She stretches, then crosses the room to an open door. “Frank,” she says, leaning in, “can you send some more flies?”

In his cluttered office in the Los Angeles suburb of Simi, Matt Keennon tosses a diaphanous creature into the air. Tremulous but purposeful, it flaps its way across the room, where waiting hands catch it. Cradled in them, it flutters a moment longer, then subsides when its captor’s fingers—huge, clumsy things beside the ethereal flier—click off its master switch.

The little creature is called the Microbat. It was built at the California Institute of Technology’s Micromachining Laboratory by a team of graduate students overseen by Yu-Chong Tai. The effort also involved AeroVironment, which is headed by Paul MacCready, a multi-disciplinary engineer famous for, among other things, the Gossamer series of human-powered aircraft. The youthful Keennon is AeroVironment’s project manager for micro air vehicles (MAVs), a new class of aircraft being funded by the Defense Advanced Research Projects Agency (DARPA). The toys in this game are small, but the players are big.

The Microbat’s thorax and wing-flapping mechanism consist of tiny sticks of carbon fiber. Its wings are gossamer plastic webs supported by a network of stiffeners that were not built up of separate components but etched from single sheets of titanium alloy by the same photolithography techniques that are used for the mass production of computer microcircuits. The Microbat carries no payload, and it serves no purpose other than to demonstrate the feasibility of a small electric ornithopter that can operate only at low speeds and indoors, where a drop of rain or a puff of wind will not immediately destroy it. More important, it demonstrates the possibility of building parts of flight vehicle structures by chemical micro-machining. Both demonstrations are prophetic.

Slow-moving, moth-like airplanes (as well as crawling robotic cockroaches and other sci-fi stuff) are where reconnaissance seems to be headed now. So-called MEMS (micro-electromechanical systems) manufacturing techniques, derived from the tools of computer chip manufacture, will get it there.

 DARPA’s involvement with toy-size airplanes began at a workshop entitled “Future Technology-Driven Revolutions in Military Operations,” conducted at the RAND Corporation in Santa Monica, California, in 1992. Bruno Augenstein, a RAND scientist, chaired a panel discussion on power supplies for “mobile microrobots,” then a completely hypothetical class of military vehicle. Despite initial skepticism, the idea that an airplane that would fit in the palm of your hand might be a useful reconnaissance device gradually took hold.

In 1995 DARPA put out a specification for a small camera-carrying aircraft. Six inches—an arbitrary value, but one that has turned out to make practical sense—was the basic constraint: The entire aircraft had to fit within a six-inch sphere.

DARPA also specified a typical mission. The midget spyplane would fly one kilometer, just over half a mile, to a target; loiter there for half an hour in turbulent winds of up to 25 mph, perhaps maneuvering among obstacles such as buildings while repeatedly climbing to 350 feet and descending again; then return to its base. It had to be quiet and inconspicuous, its launching and control system had to be easily portable and operable by an unskilled person, and the whole system had to be both robust and cheap.

In 1997 DARPA gave grants totaling several million dollars to several organizations to develop MAVs; AeroVironment, which had already begun attacking the problem on its own, was one of them. The company’s Simi Valley facility has produced a number of flying models, most of them of roughly circular planform, six inches in diameter, and powered by a single tractor propeller spinning at 20,000 rpm.

The most successful of AeroVironment’s models, nicknamed Black Widow, has remained aloft for more than 20 minutes flying at 35 mph. The ground operator launches it by compressed air from a telescoping rail, then controls it in flight by radio, like a model airplane—which, after all, it is. Unlike the typical radio-controlled flier, however, the Black Widow’s operator watches not the airplane itself, which is a mere speck darting in the sky, but the video picture sent back by its tiny television camera. The whole apparatus—airplane plus launch and control mechanisms—fits in a briefcase.

Fortuitously, Keennon says, various pieces of “COTS”—commercial off-the-shelf—hardware are available in the right size to fit on a six-inch flying disk. Flight controls, for example, are operated by tiny Swiss-made electric motors an eighth of an inch in diameter and 0.01 ounce in weight. The airplane’s “eye,” also an inexpensive item, is a 510- by 492-pixel color array like the ones used in home video cameras but stripped down to the size of a bean and the weight of 0.05 ounce.

AeroVironment’s current MAVs are skittish creatures, with high roll rates and low natural stability. They require skilled radio control operators. The next step in the program, which the company is currently pursuing with its own funds, independent of DARPA’s support, is to add electronic gyroscopes and autopilots that will keep the airplanes stable and upright. The operator would then need no special skill to fly one, and would be free to concentrate on the mission rather than on controlling the aircraft.

After adding stability, the next improvement will be GPS navigation, which would permit the MAV to fly a programmed mission without assistance from a human operator. The icing on the cake would be some kind of system using acoustic or optical sensing that would let it maneuver in an urban environment, avoiding obstacles on its own, just like a bird. That level of autonomy, however, is still far off.

The requirement that it send back usable video images puts an important lower limit on the size of a MAV, because each pixel in the imaging array must be considerably larger than the longest wavelength of visible light. This means that a video camera capable of sending back useful detail can’t be much smaller than the one Keennon’s team is now using. Another non-scalable item is the radio antenna. An antenna that fits within a six-inch space works efficiently only with short-wavelength, high-frequency radio waves. Unfortunately, high-frequency radio signals travel by line of sight—both antennas have to be able to “see” each other—and do not readily penetrate walls or travel around hills. A longer retractable trailing wire, however fine, would impose a severe drag penalty. Antenna size will also pose a problem for GPS reception, especially if future MAVs became significantly smaller than the current ones.

The peculiar configuration of AeroVironment’s MAVs is the logical outcome of the six-inch size restraint. If you merely scaled a conventionally proportioned airplane down to a six-inch wingspan, its wings would have an area of only about .04 square foot. Flying at 30 mph—a higher speed would require too much power—such a wing could support only about three-quarters of an ounce at most, with no margin for maneuvering or gust response. But the weight of the entire aircraft, including powerplant and all the electronic and sensing equipment it is supposed to carry, would in reality be around two or three ounces.

It turns out that the best solution is simply to make the wing area as large as possible—essentially, to fill the entire six-inch DARPA circle with wing. This approach has other advantages as well: It provides a simple, stiff, voluminous structure with ample interior space for systems and payload. True, the circular planform lacks the characteristic usually associated with efficient airplanes: a fairly high aspect ratio. The most efficient airplanes have wings whose span from tip to tip is much greater than their chord—the distance from leading to trailing edges—and you don’t see a lot of airliners with circular wings.

But for an airplane of this size or smaller, a low aspect ratio may not be a hindrance. The very wingtip vortices that produce drag on conventional airplanes help produce lift instead on small, short-span wings operating at low Reynolds numbers (see “Mr. Reynolds, We’ve Got Your Number,” next page). In fact, recent research on insect flight suggests that the judicious use of tip and leading edge vortices keeps those notoriously small-winged bumblebees—the ones that, according to legend, myopic scientists have pronounced flightless—aloft. This is only one of the differences, fundamental to the creation of miniaturized aircraft, between full-scale and micro-scale aerodynamics. The behavior of air on micro-scale wings is only beginning to be understood.

Although most of the systems of a MAV are electronic and AeroVironment has concentrated on electrically powered airplanes, not everyone agrees that an electric motor is the best choice for a powerplant. Batteries have a low “power density”—that is, they pack little punch for their weight. (This is a problem for electric cars as well.) For some tasks, such as peering into upper-story windows or loitering inconspicuously, an aircraft that can hover is essential; at present, battery-powered electric motors don’t have the power to hover for long.

Those traveling the all-electric route look to future improvements in batteries, motors, and propellers, as well as to further miniaturization, for increases in power-to-weight ratio and efficiency—the fraction of the available power that goes into useful work—of tiny power plants. But gram for gram, chemical fuels like gasoline are much more energetic than batteries, and even though extremely tiny internal combustion engines, unlike tiny electric motors, are not available off the shelf, several programs are taking the internal combustion route instead. 

MLB Company of Palo Alto, California, has flown several designs powered by small Cox model airplane engines. One of them takes off vertically. Stephen Moore of MLB says that at this scale the power requirement for vertical takeoff and hovering is not terribly different from that for agile maneuvering. Given the tremendous energy content of chemical fuel, a multi-mode tail-sitter craft that can both fly and hover becomes an attractive possibility.

A startling solution to the power problem is in the offing at the Massachusetts Institute of Technology in Cambridge, Massachusetts: a jet engine the size of a shirt button. Components of such engines have actually run in test beds. The baseline design involves a single centrifugal-flow compressor spinning at 2.5 million rpm on gas bearings. Combustion takes place in a doughnut-shaped chamber surrounding the engine, and the exhaust gas flows back inward toward the center through a turbine. A starter-generator is built into the case; if needed, the engine could serve as a tiny electrical generator, putting out 10 to 20, or perhaps as much as 100 watts, or it can be used as a jet engine with a thrust of up to a third of a pound.

The key to making such a device cheaply and in large numbers is a version of the same photolithographic manufacturing technique used at Caltech to make the wings of the Microbat. Engine parts would be etched in sheets of silicon, like microchips. (By the early 1990s, electric motors smaller than the point of a pin, invisible to the naked eye, had already been made by this technique.) Just one micro-engine would be sufficient to supply both the thrust and the electrical requirements of a present-day MAV.

At the Georgia Institute of Technology Aerospace Laboratory, Robert Michelson leads a project to develop and refine an entomopter, a machine that will not only fly like a bug but, if need be, crawl like one too. The entomopter has a “chemical nose” and other features to permit it to home in on certain kinds of targets. Its builders expect to provide it with navigation and obstacle-avoidance skills as well. But the present centerpiece of the project is its power plant, a device called a reciprocating chemical muscle.

The RCM is something like the piston and cylinder of a steam engine, except that the gas that drives it comes not from combustion but from a chemical reaction. The energy available from the chemical fuel is much greater than that available from current batteries. And the chemical reaction also has the advantage of versatility: Its waste heat can be converted into electricity to operate onboard sensors and transmitters, and spent gas can be vented over the wings to provide differential lift and, therefore, flight control.

By calling their prime mover a “muscle,” the Georgia researchers underscore their reliance on the guidance of Mother Nature. “Nothing in nature achieves sustained flight with fixed wings or with propellers,” observes Michelson. “All tiny creatures flap their wings continuously. Flies don’t glide.”

A similar project, called the Micromechanical Flying Insect, is under way at the University of California at Berkeley, where a team headed by biologist Michael Dickinson has shed light on how insects use their wings. To simulate the Reynolds number of insect flight, Dickinson and co-workers built and instrumented a pair of 10-inch wings driven by six separate actuators, and have observed them flapping in a tank of mineral oil. In addition to a new understanding of very-low-Reynolds-number aerodynamics, such work has spawned a new vocabulary for talking about flight phenomena, with terms like “delayed stall,” “rotational circulation,” and “wake capture.”

Wing flapping works in several ways to provide insects with a flying ability that would be the envy of any fighter pilot. To start with, the flapping of wings plays the same role as the spin of a helicopter’s rotor: It creates a relative wind over the lifting surface even while the vehicle—or bug—is standing still. But flapping also sets up tiny vortices that take the place of the cambered flying surfaces, high-lift devices, and moveable flight controls of fixed airplane wings. The eddies set up by their wings not only keep bugs aloft but also allow them to hover, fly backward or sideways, and turn on a dime (or the corresponding currency of the bug world).

Putting the new understanding to practical use is the next step, and not an easy one. The Berkeley team, with some sponsorship from DARPA, proposes to duplicate, in a mechanism about the size of a quarter, at least some of the abilities of a large, repulsive, carrion-eating fly called Calliphora. “You can’t build [robot insects] now based on known principles,” Dickinson has said. “You have to fundamentally rethink the problem.”

Most of the proposed uses for MAVs are military; the funding, after all, is coming from the Department of Defense. But some workers in the field propose broader applications for tiny flying robots. Georgia Tech’s Michelson has suggested sending robot “terminators” after real-life insect pests, but suspects that the largest potential outlet for small aerial robots might be the toy market. Stanford’s Ilan Kroo leads a team developing a “mesicopter,” a multi-rotor electric helicopter. Currently of centimeter size but potentially much smaller, the mesicopter is shaped like a thin, square wafer with a little rotor at each corner. Essentially a flying microchip, a mesicopter’s motors, sensors, guidance, and telemetry systems would be etched in place in a single completely automatic manufacturing operation. Kroo’s team envisions swarms of mesicopters investigating the interiors of storms or the atmosphere of Mars.

The word “swarm” is particularly significant. Of course, it suggests insects, and much of the more startling MAV research is headed in the direction of emulating those successful products of natural selection. But it also alludes to nature’s profligacy. Many creatures that live in hazardous environments reproduce in huge numbers so that just a few may survive to maturity. MEMS manufacturing techniques imply a similar approach to machines. Rather than launch a single costly, sophisticated, man-carrying device to do a job, you would launch hundreds of simple, cheap robots. If most of them fail, no matter—they are expendable. Only one needs to complete the task.

A case in point, reminiscent of the wholesale egg-laying habits of marine creatures, is what Kris Pister of the Berkeley program calls “smart dust.” Consisting of various kinds of motion or chemical sensors, a power supply, a microprocessor, and a system of communication, all packed into the volume of a grain of coarse sand, these “motes” would be sprinkled randomly over a wide area to report back what they find. What would do the sprinkling and receive the reports? A MAV, of course.

The use of MEMS in the construction of miniature aircraft is likely to bring about, in the next decade, innovations that seem incredible today. The button turbojet could revolutionize propulsion, even if only for model aircraft builders. Soldiers fighting in blasted cities in central Asia will be grateful for the ability to look around corners with tiny airborne cameras. Children will shriek with delight as their robot wasps attack a neighbor’s action figure.

But will we have to wonder, even in civilian life, whether every persistent fly we encounter is carrying a listening device?