In 2013, some folks from Rob Wood’s lab at Harvard, including then-postdoc Sawyer Buckminster Fuller, published a paper in Science introducing a (mostly) controllable version RoboBee, an insect-size flying robot that could lift itself, hover, and move around a bit using two flapping wings. Since then, there have been several more generations of RoboBee, including this nutty explosive diving one.
The problem with robots at this scale, and especially flying robots at this scale, is energy storage. It takes a lot of oomph to lift off of the ground and stay there, which means that high power is necessary, which means a relatively big battery to provide that power for a significant amount of time, which means a heavier robot over, which means more power is required to lift off, and you can see what the problem is.
Fuller has since moved on to a professorship at the University of Washington, where he’s been working on ways of solving this problem of power autonomy. Last year, we wrote about a laser-powered flapping-wing microrobot, and in the April issue of IEEE Robotics and Automation Letters, Fuller has a paper on a new flapping-wing microrobot that might be able to be both power autonomous as well as getting airborne with the necessary sensing and computing to do something useful. The secret is not much of a secret— instead of two wings, why not use four?
This robot uses the same sort of piezoelectric actuators as Harvard’s RoboBee, just rotated sideways. At 143 milligrams, it weighs just about as much as a real honeybee, but the key statistic is that it’s capable of lifting an additional 260 mg (at least), which ought to be enough for both sensors and a battery or supercapacitor. The extra power comes from the extra wings, of course, and while you can’t simply double payload capacity by doubling the number of wings, you can, hopefully, go from “not quite enough payload” to “just barely enough payload.”
Payload is just one problem that earlier generations of two-winged robotic insects have struggled with. The other big issue is control—specifically, winged microrobots find it very difficult to keep control over their rotation, or yaw. Having four wings makes control over three axes very straightforward, with yaw, pitch, and roll all responding to different combinations of the speed and amplitude of wing beats. The frequency of the wing beats stays constant at about 160 hertz, and stroke amplitude refers to how far back and forth the wing moves with each stroke. Changing wing speed means a slower than normal stroke in one direction followed by a faster than normal stroke in the other direction, which keeps frequency the same but allows for yaw.
In the four-winged robot design, roll (x) and pitch (y) axes are actuated by varying the stroke amplitude across opposing wings; actuating the yaw (z) axis (“steering”) is performed by varying the speed of one direction of the stroke relative to the other. The approximate location of the center of pressure is denoted by a dot in the upper left diagram; the distance from it to the robot’s center of mass is given by r cp. Arrows denote approximate stroke amplitude of each of the four wings. Image: University of Washington
Despite having three-axis control over the robot, getting it to fly in a stable and controllable manner is still not easy. There’s some suggestion that it might, fundamentally, be passively stable in a hover, meaning that if properly calibrated it would just sit quietly in the air, flapping away, without needing constant attention. But what we’re seeing in the video are just the very first experiments showing that the robot can be controlled in flight, and that it can hover pretty well with the assistance of feedback from a motion-capture system.
The version of the robot in the video above was able to lift itself carrying a mass of 262 mg, and in the paper, Fuller talks about what that payload implies for eventual power and sensor autonomy. The first step will be to do one or the other rather than both at the same time. Sensor autonomy would mean a gyro, accelerometer, optic-flow sensor, laser rangefinder, and microcontroller, while power autonomy would require a microcontroller, a boost converter and a half dozen supercapacitors.
Fuller estimates that a complete sensor package would weigh about 200 mg, while the power system would weigh about 260 mg, either one of which is a mass the robot can currently lift. Eventually, the hope is that the robot will be able to fly with both at once, and as a start, Fuller suggests that reorienting the actuators to allow the wings to take advantage of the same kind of aerodynamics that larger beetles leverage to fly could increase lift by up to 20 percent.
For more details, we spoke with Fuller via email.
IEEE Spectrum: You presented a two-wing laser-powered insect at ICRA last year. How did that design evolve into this one?
Sawyer Buckminster Fuller: I basically used the same basic wing unit from that design, but added two more for more payload. Now we can fly some sensors in addition to just the power system!
Why have previous small robotic insects focused on two-wing designs, when they have such significant challenges to autonomous, tether-free operation?
There is one good reason to start out using two wings: it is a simpler, and requires fewer parts. When you’re just getting going figuring out how to make these tiny things that are hard to build, starting with two wings is exactly the right thing to do. But knowing what I know about yaw torque, I just couldn’t think of a way to make one with two wings that could also steer itself.
Can you summarize the advantages of using four wings rather than two, and in this orientation rather than a vertical one?
The reason for this redesign is to get around two limitations of the RoboBee design that first appeared in Science in 2013. That design consists of two wings flapping close together, with its piezoelectric actuators hanging vertically.
The first limitation is something that is not obvious by watching videos of it hovering, but you may notice if you take a close look: it is not able to steer itself (that is, rotate itself around a vertical axis). As it flies, you can see that it is continually rotating back and forth, which is because it is not able to reject disturbances caused by wind and the wire tether that supplies power and control signals. We actually had to write a pretty sophisticated nonlinear controller so that it could operate in the presence of these large rotations. But being able to face in a desired direction is something we’d like to do for a variety of reasons, such as to point a sensor in a certain direction or aim landing gear.
This new design can steer itself, and the reason it can is that the wings have been moved farther away from the center of mass so that they have a longer moment arm, for greater torque. The way this robot steers is similar to how a quadrotor does, which is to spin the two props that are spinning in one direction faster than the other two. But instead of changing the rotor speed, we actually change the speed of the flapping motion in one direction compared to the other. When the wing moves faster, the drag is higher, causing a force. The problem is that the force is very small, so we had to move the wing farther out to overcome the disturbances.
The second limitation is that it just doesn’t have enough payload capacity to do interesting things on its own. The suite sensors we’d need to fly fully autonomously without external motion capture cameras— a gyroscope, an optic flow sensor, and a small laser rangefinder— are getting smaller but still too heavy for the RoboBee. The extra two wings on this new design give it enough lift to carry these sensors. I also think it has enough payload capacity to carry its own tiny supercapacitor or battery and boost converter, to perform power-autonomous flight.
Can you talk about the challenges of making a robot like this fully autonomous?
The big ones are getting sensors small enough, and getting enough computation onboard. Thanks to the consumer electronics industry, there is immense pressure to make things small and light, and we are definitely taking advantage of that. Some parts, like the gyroscope, would be hard to make as good with research money as an industry that sells billions of them in every phone. There are also now tiny, wafer-level chip packages that can do floating point math at nearly 200 MHz (the ARM M4), so that really helps too. Eventually, I think we will want to run a real OS and maybe ROS onboard, but we are not there yet, we need somebody to make a tiny computer on a single chip that has the peripherals we need, and so far they’re all a little too big. I am actively looking for device collaborators who have the resources to build custom parts tailored to small robots. Hopefully, in the process of solving problems at this small scale, we’re going to come up with sensor systems that can be used on robots at all scales and in other application domains, like in medicine.
What are you working on next?
We’re thinking about ways to achieve power autonomy, sensor autonomy, and both combined, how to improve our actuators, and new modes of locomotion in addition to walking and flying, like jumping. I’m really excited about the potential for tiny robots in outer space: at a launch cost of about $10k per KG to low earth orbit, we’re talking just a few dollars to put up a sub-1 gram robot!
I’d totally pay a few dollars to see tiny flying robots zipping around the ISS. And that would handily solve the problem of payload mass, wouldn’t it?