Robotic Fly Takes to the Air, Briefly

Liz is a staff writer for Science.

Lift off. Finally, a robot the size of a fly can actually fly.
Credit: Kevin Ma

First there were drones, then there were quadcopters. Now there's RoboBee, which really looks more like a fly. After more than a decade of work, engineers have built an insect-sized robot that can take off, fly back and forth, land, and take off again.

A new fabrication process based on the principle of pop-up books made this microrobot possible, but don't expect to see it in stores any time soon. It takes 2 days to make a single RoboBee, and the tiny device still requires a tether to supply power and guidance for flight. "Getting all the sensors and power on board is definitely many years out," says Sarah Bergbreiter, a mechanical engineer at the University of Maryland, College Park, who was not involved in the work. Nonetheless, this new robot is "pretty fantastically cool," she adds.

Although thousands of insect species dart about with agility that puts stunt pilots to shame, most engineers have considered building a robotic fly an impossible task. You can't buy off-the-shelf parts for the body, and no existing power source, sensors, or controllers are small enough to fit on board. What's more, researchers don't even have a good grasp on how aerodynamic principles change on such small scales, so they can't precisely predict how delicate wing movements will alter flight. Yet 12 years ago, mechanical engineer Robert Wood, who is now at Harvard University, decided to embrace the challenge of building an insect-sized robot—in part to understand the flight mechanics of small flapping wings, and in part "because it was so hard," he recalls.

The method that Wood and his colleagues developed to make microrobots with movable parts is impressive in its own right, says Vijay Kumar, a mechanical engineer at the University of Pennsylvania who was not involved with the work. "It has applications well beyond the specific structures they are building," and could even lead to new kinds of medical devices.

Harvard graduate student Kevin Ma starts by layering carbon fiber material and polymer film into a flat sheet, and then cuts a design into the sheet using a laser. The design includes scaffolding that pops up, pulling the 2D pattern into a 3D shape in which the different parts fold together. For now, the finishing touches are made with tweezers under a microscope, but Ma hopes that eventually the whole robot will pop up all at once in an automated process.

Real insect wings are flexible in many directions, but that was too difficult to reproduce in the robot. So its wings can just flap and rotate. "But surprisingly, that's enough to get this thing to fly," Ma says. RoboBee's ceramic muscles bend when subjected to a voltage, causing the attached wings to flap. The voltage comes from a thin wire that connects the robot to a power source and a computer. The robot has no sensors, so to achieve controlled flight, Ma and his colleagues put the robot in a box where eight cameras track it as it moves. The cameras relay position information to the computer, which then activates the muscles to correctly fly the robot.

Ma spent many months trying to get the robot to fly. It turned out that if the wings were the least bit asymmetrical, he couldn't control its airborne activity. The pop-up technology helped increase precision, but it still took many rounds of tweaking the design before it finally worked. "It was an amazing feeling of having all of this hard work suddenly bear fruit," Ma says. Wood calls it their "Kitty Hawk moment."

"This is the smallest flapping wing aircraft that has ever been built and made fully functional," Kumar says. "It's a big accomplishment.

Right now, as Ma and his colleagues report today in Science, the robot takes just 20-second-long flights, as its material will fatigue and fail after a total of 15 minutes of use. (For comparison, Orville Wright's first airplane flight clocked in at 12 seconds.) "We don't want to wear out the system," Ma says. He and his colleagues are trying to come up with designs that will stress the materials less so that the robots will be more durable. Despite Bergbreiter's predictions, Ma hopes to equip the robotic fly with its own power and sensors so it can fly untethered by the time he graduates in 2 years. "The Holy Grail is to have everything on board," Kumar says.

Even tethered, the robotic fly can help Wood, Ma, and others learn more about how real insect wings interface with air flow, insights that will help simplify the control of robotic flight. When autonomous, the robots could be used in large numbers to go into disaster sites to help locate trapped people or gas leaks and other hazards. "When you scale things down, smaller is better," Kumar says. "You can go into nooks and crannies that you couldn't otherwise go into."

Posted in Technology