Luke Roberts, a 26-year-old musician with an ash-blond Mohawk and a scruffy chin, sits at a table littered with wires, spare parts, and half-assembled robots. Behind him stands a red and black human-sized bot with four-foot-long arms and a gray computer screen for a face. This is the University of Maryland’s Robotics Center, in College Park, a 35-minute drive from Washington, D.C., and where Roberts completed his master’s degree. Here students’ mechanical creations learn to walk, crawl, and even fly.
Unlike some of the other hulking bots standing about the room, Roberts’s consists of just a few carbon fiber sticks about 30 inches long, a whir of wires and sensors, and microchips. It weighs roughly half a pound. “The skeleton holds all the electronics,” he explains, pointing at the flight controller, which translates the human pilot’s commands into the computer code that activates the wings’ gears, along with other vital parts. “This is the battery, and this is the brain.”
Using rubber bands to tie on the wings made of Mylar (the material used to make the thin film in helium balloons), he demonstrates how to quickly assemble the bot. Suddenly a creature takes shape, looking like a cross between a dragonfly and a pterodactyl. Holding it in his outstretched hand the way a medieval hunter displays a falcon, he pushes a button on his remote control box and the “bird,” named Robo Raven, starts flapping its wings. After examining it closely to make sure nothing is loose, Roberts deems it ready for a test flight. Then he removes the wings and packs it up for transport to the airfield.
In the morning, Roberts arrives at the field, reassembles Robo Raven, and launches it into the air. It flies off quietly, with only the gentle sound of flapping wings. The gadget looks so real that local hawks have attacked and destroyed different versions of it more than once. Unlike all other human-built aerial machinery—the bulky, noisy metal monstrosities incongruent with nature—it blends into the landscape like a living, breathing thing. No existing manmade flying machinery comes close to achieving this feat. Helicopters maneuver well, but they’re inefficient because they burn fuel to generate lift and propulsion. Planes are better on energy use because they create lift with their wings, though they are less maneuverable. Rockets have the pitfalls of both, plus they take off with deafening blasts.
Modern micro air vehicles, better known as drones, inherited the same problems as helicopters and planes. Powered by propellers, certain kinds can use a lot of energy and stay airborne for only a short time—rarely more than 30 minutes. They don’t handle turbulence well; a gust of wind can send them tumbling.
Winged robots, like Robo Raven and several other so-called “bird bots” being developed at laboratories around the world, are far superior to the tiny, energy-hogging flying gizmos of today. They have a host of uses: gathering intelligence; delivering packages to hard-to-reach places; doing aerial surveillance—for example, tracking the spread of forest fires, which in 2015 cost the federal government $2 billion to control. When they’re built to resemble raptors, they can keep pest birds in check around airports, farmlands, and landfills. In 2011, in just five states (California, Michigan, New York, Oregon, and Washington), birds damaged $189 million worth of blueberries, cherries, honeycrisp apples, and wine grapes. At landfills, birds interfere with daily operations, and their droppings contain pathogens dangerous to humans and livestock. When birds get sucked into airplane engines, they can cause crashes. Bird robots promise to overcome such problems—saving both money and human lives.
Since the time of Leonardo da Vinci, who sketched and even possibly tested a flapping-winged glider, people have been trying—and failing—to make a birdlike aerial machine. There have been major hurdles to overcome. The biggest among them: weight. Engines that could generate sufficient lift and computers capable of matching birds’ brains’ ability to process large amounts of sensory info were too heavy to fly. Re-creating birds’ wings, which consist of dozens of bones and muscles, called for moving parts that could be both tiny and lightweight but also durable enough to withstand hours of flight.
Today it’s possible to overcome these challenges using tiny electronics and computer processors, as well as 3D printing for building intricate mechanical parts from various composite materials. Roberts’s adviser, Satyandra K. Gupta, who was a professor of mechanical engineering at the University of Maryland for more than 17 years (he recently accepted a new job at the University of Southern California in Los Angeles), says that the technological advances in this last decade were a boon to building winged bots. “A lot of motors and microcontrollers and all other miniature [electronics] are now available,” says Gupta. “These things were too bulky before to create a flying prototype.”
That doesn’t mean building bird robots is easy. Gupta, a stout man with a touch of gray at his temples, who has been fascinated with birds since his childhood in Mathura, near New Delhi, can’t count how many Robo Ravens crashed to the ground during the three and a half years his students worked on the bot. From the outset, the team designed it to use its two wings independently—like real birds do. To achieve that, the team powers each wing by a separate motor. “It’s like having two muscles,” Roberts says. “You can flap each wing at a different angle or at a different frequency.”
Gupta and his students see Robo Ravens being used for information gathering as well as for search-and-rescue operations. The bots could, for instance, stealthily circle a spot where enemy troops are hiding. Or they could hover over a hard-to-reach area, looking for lost hikers or mountain climbers—and do so much more cheaply than a helicopter search.
Bird bots still occasionally wreck, especially as scientists, always looking to improve their models, test the limits of what bots can do. “Unlike with building a ground vehicle, where if something gets loose, it will just come to a stop,” Gupta says. “In this case, if something goes slightly wrong, it crashes.”
Joining Roberts’s early morning test flight, the professor and two other students, John Gerdes and Alex Holness, along with intern Vishal Gupta, unload multiple generations of flying machines onto the dewy grass field owned by the College of Agriculture, which allows them to fly their bird bots on the condition that Robo Ravens stay out of the vegetable gardens.
Each bird in the five-generation collection has its own unique features. Robo Raven II carries sensors that monitor its ascent, descent, and wing speed. Raven III can partially recharge in flight, thanks to its solar batteries—little dark squares glued to its Mylar wings. Generation IV flies on autopilot. Roberts’s baby, Robo Raven V, has self-adjusting features—like real birds do—to manage changing flight conditions, like wind, which should further improve its stability. “[Birds] take advantage of different things that happen,” he says. “They end up dipping down and coming back up and hovering a little bit. And then occasionally they do some flapping.”
The ability to recharge in flight is unique and very valuable for intelligence gathering—to be useful, bots must stay airborne for long periods of time. “The advantage of using the energy-harvesting technology, like solar, is being able to use these platforms indefinitely,” says Hugh Bruck, a University of Maryland professor of mechanical engineering, who works with Gupta’s students. Adding solar cells required a redesign, because they increased the stiffness of the wings, affecting the Mylar’s natural flexibility, which is needed for flapping. Still, it was worth the effort. And while the bot can’t yet fly on solar power alone, the team is looking at high-efficiency cells, hoping to eventually remove the current lithium batteries entirely.
Other scientists are finding inspiration in birds’ perching and grasping talents. Justin Thomas, a Ph.D. student at the University of Pennsylvania’s GRASP Lab, studies aerial robotics, including the ability to seize objects at high speed. In 2011 his lab devised an experiment in which a trio of quadcopters built a small structure from Lego-like blocks. The quadcopters were programmed to find the blocks on the lab’s floor, pick them up, fly them over to the construction spot a few feet away, and place them on top of one another, forming a small tower. The blocks had retro-reflective markers detectable by cameras, which served as the machines’ eyes (except they were placed not on the quadcopters but around the room to minimize flying weight). The building algorithm was coded into the quadcopters’ “brains.” The drones accomplished the task, but Thomas noticed that they spent too much time positioning themselves to pick up the blocks. “They would have to hover over the object, descend, hover there for a bit, grab the object, ascend,” he says—which was so slow that sometimes their batteries died before they finished.
In contrast, a Bald Eagle, caught on video snatching a fish out of the water, was so quick it became the model for Thomas’s gripper invention. “The eagle could do it so fast, so elegantly, and just so magnificently,” Thomas says, that it inspired his clawing robot design.
Grabbing things at high speed—for example, trying to snatch a roadside package leaning out the window of a speeding car—is hard because you need to align your hand with the object without losing momentum, and then clutch quickly. But birds of prey evolved a way to do so: During the capture moment, they swing their legs backward so their claws move slower in relation to their prey. Thomas and his team borrowed that trick. They built a quadcopter with raptor-like legs, which it sweeps back when seizing its targets. That allows it to grab a parcel at 7 miles per hour, which in human terms is exactly like trying to nab an object while riding a roller coaster.
The clawing bot doesn’t need a person to steer it. Instead, a camera, still remote to minimize weight, supplies the target’s location, and the bot buzzes over and snatches it autonomously. The price tag was higher, though, because the team bought a pre-built quadcopter for about $3,500 and then outfitted it with talon-like grippers for an additional $100. Still, due to the nature of its task—grabbing an object at a known location—the bot was pretty successful.
Thomas sees many practical applications for his clawed creature, from law enforcement to environmental uses. For example, the bot could plunge into a burning forest and quickly place temperature sensors or smoke detectors that tell firefighters where the blaze is heading. In the case of oil spills, it could reach affected areas faster than most boats and swoop down to gather samples for oil spread assessments. A clawed bot may not be able to fly inside a tornado, but it could use its talons to perch on a tree near the storm’s path, supplying important information to meteorologists. And it could serve as the police’s eyes and ears in shootouts or terrorist attacks. Equipped with microphones and cameras, clawed bots could be inconspicuously positioned on trees or buildings and stream information about the criminals’ movements.
To do all this, the bot would have to carry a camera onboard, avoid obstacles, and make its own decisions based on the situation—like identifying the best spot to swoop down to place a sensor. Thomas’s invention hasn’t done this yet, partially due to limitations of weight and computational power. But as computer processors grow ever smaller and more powerful, he says, so will the bots’ brains.
Robots are being built not only to move like real birds but also to resemble them. For instance, “Robird,” designed as a thesis project by an enterprising graduate student, Nico Nijenhuis, in the Netherlands, is an exact replica of a Peregrine Falcon. In 2011 Nijenhuis, then 25 and working as a part-time chauffeur while studying applied physics at the Technical University of Twente, teamed up with three bird enthusiasts who had devised Robird’s early prototype. The bot looked more like an airplane with flapping wings than an actual bird, but Nijenhuis quickly recognized that it had practical potential, such as bird pest control and aerial surveillance. “I said, ‘This is so cool,’ ” he remembers. “I see so many opportunities. We should start working towards a commercially viable product.”
But there were some major design matters to resolve. First, to be a convincing menace, the bot had to look and move exactly like a falcon. If it was built too light, it would get tossed around in the wind, which wouldn’t look realistic. If it was too heavy, it wouldn’t fly at all. To preserve its looks, the electronics had to be hidden in the bot’s shell, which posed another problem, because the carbon fiber obstructs GPS and other signals. Fortunately, Nijenhuis found a 3D-printing company that could build the shell from composite materials—in this case, nylon mixed with glass fiber particles—which gave him the needed low-weight flexibility and strength without suppressing the signals. The falcon’s head was fashioned from the real bird’s and 3D-printed, too. “We took pictures of live falcons and translated them into computer drawings, and it has become the actual head of our model,” he says.
At the same time, his team worked on perfectly mimicking the falcon’s wing movements, so complex that no amount of math could describe them precisely. Resorting to trial and error, they watched dozens of Robirds smash on the ground during the three years it took to master the model. “It’s so incredibly difficult to create a perfect bird,” Nijenhuis says. He remembers, after months of trying to find the right balance between lift and drag, how thrilling it was when it finally worked. “You throw the bird into the air and it flies, and it flies really well, and it does everything you want it do,” he says. “It brings you close to tears!”
Nijenhuis’s company, Clear Flight Solutions, now uses a fleet of 12 hand-painted Peregrine Falcons and two Bald Eagles to serve several clients. The bots shoo starlings and crows away from blueberry plantations. At the Twente landfill, Robirds reduced pest crows by as much as 70 percent and gulls by roughly 95 percent. So many clients want to hire Robird that Nijenhuis is increasing his fleet. “We’re in the middle of building 10 new ones,” he says excitedly. “We’re going to go over 20 operational falcons.”
The bird models and applications are quite diverse, but there’s one thing the scientists agree on—how difficult it is to model a bird’s flight. For example, the birds’ wing movements are so complex that so far modern science hasn’t been able to precisely describe them using formulas and mathematical equations, according to David Lentink, a mechanical engineer who leads a bio-inspired engineering design lab at Stanford University in California. That’s the big reason it takes so many trials, errors, and crashes to build a bird. Lentink is working to solve this problem. Using various mechanisms that he has developed to watch and measure the forces that birds generate with their wings as they fly, he hopes to understand how the animals achieve lift and propulsion, change their speed and direction, and adjust to wind gusts. He wants to fill in the gaps in understanding to help builders design better bots. Once the math is pinned down, he says, constructing small, safe, agile, and long-lasting flying machines would become much easier.
And he predicts that it will happen within the next 10 to 20 years, ushering in an era in which lifelike aerial robots will routinely help humans accomplish tasks in dangerous or hard-to-reach places. Some million-dollar questions: Will they ever be able to coexist with us in cities? Will they be able to deliver packages to consumers or prescription meds to the elderly, coasting through turbulence, which often exists around big buildings, and avoiding trees, birds, and other bots, without falling on people’s heads?
Lentink believes that’s the future. Current avian models may be unsafe for flying in crowded spaces, but once engineers perfect them and bots soar as skillfully as birds, collisions will become less frequent. “No one worries about birds crashing, and billions of birds fly every day!” he says. What’s more, he maintains, we will learn to use flying robots just like we learned to use trains, planes, and automobiles—
despite the fact that early models of these inventions were prone to accidents. Urban planners of the future—like those before them—could pattern cities to accommodate the innovations that redesign the way we live. For example, by designating air corridors for robots to fly.
Asked whether we will one day receive Amazon packages dropped off on our doorsteps by flying machines, Lentink answers with a definitive “Yes.” But not just any robots, he adds. “Bird-size robots—yes!”