MIT为何研发仿生机器人昆虫？微型机器人的挑战与未来

There are robots the size of bees, others that can jump on water, and some that are powered by tiny combustion engines the size of a penny.

One day they could work in swarms, they could save your life, or even spy on you.

We got access to the best micro robotics labs in the world. To learn how these robots work and what are they for.

This is a tiny yellow submarine.

Under water, it can move around by flapping these miniature wings nine times per second, but you can use those same wings when the submarine is out of the water, only now you have to flap them 250 times a second to make it fly.

So, this robot can do both. It can fly and swim, but since it weighs only 175 milligrams, about the mass of two Cheerios, surface tension is a problem.

- That's a consequence of physics at a smaller scale. The surface tension is like a wall that blocks the transition process.

This happens because water molecules are slightly polar.

Groups of these molecules pull in all directions, but at the surface there's no water above, so the pull is only sideways and downwards.

This imbalance creates strong, cohesive forces that compress the surface into a tightly packed layer, making it difficult to break.

This is the same effect that lets water striders walk on the surfaces of ponds and lakes.

This other robot weighs only 68 milligrams and by using a spring mechanism that mimics a flea's leg, it can jump without breaking the water's surface, just like a water strider.

It's like there's solid ground below.

Now that's great if you wanna stay on top of the water, but this barrier can also be a problem if you want to go underwater.

To escape this trap, the submarine splits water into hydrogen and oxygen and then stores these gases in a buoyancy chamber.

It does this because the wings are super fragile.

If they started flapping while the robot was still trapped under water, they would break right off.

So, the buoyancy from the gas helps bring the fragile parts of the robot out of the water, but the robot is still stuck in that top layer of surface tension.

So, a sparker inside the chamber ignites the gas, and the explosion breaks the surface tension and shoots the robot 30 centimeters into the air.

And once it's free, this robot can fly.

This robot found a different way to break through the surface tension.

It uses these large water repellent copper pads on its feet to walk on the water, but when it needs to dive beneath, it applies 600 volts to those pads, which creates a positive charge that attracts water molecules to it and breaks the hydrophobic barrier, and that allows it to sink on command.

Then, once submerged, it can walk underwater.

Both of these robots were made by Dr. Kevin Chen at MIT.

- We're looking at the flight room, and this is where we do all of our flight experiment. As you can see, it has motion capture cameras.

This lab is one of the only places in the world where robots this small attempt flight.

- Okay, so because this robot's so small, it is such low inertia, right? So, you're saying that it could flip faster than any other drone in the world?

- Beyond 7,000 degrees per second.

Getting these robots flying is tough.

I mean, they're the size of bees, so the internal mechanisms have to be even smaller, like the parts of a watch.

Components have to be precise to within five microns. That's a 10th the width of a human hair.

So these bots fly but not like birds. I mean, they don't soar.

Instead, they have to use a whole lot more energy, flapping their wings hundreds of times per second.

Well, it comes down to this scale phenomenon.

Larger objects typically have less surface area relative to their volume, and that's important.

Let's just approximate a flyer by a cube. Let's say it's 10 centimeters on a side. Well then that would have a volume of 10 by 10 by 10, or a thousand cubic centimeters, and it would have an area of 10 by 10 by six sides, 600 square centimeters. So the surface area to volume ratio would be 0.6 to one.

But now imagine we have a much smaller flyer that is just one cubic centimeter in volume. Well, its surface area is going to be one by one times six. That is six square centimeters. So that's gonna be 10 times the surface area to volume ratio. It's gonna have a surface area to volume ratio of six to one.

Now why is that so important? Well, it's because drag depends on surface area.

So if you have more surface area to volume, well you're gonna have a lot more drag, and also, at that small scale, you'll be much lighter relative to that drag.

So you're not gonna have as much inertia, so you'll get pushed around more by the air, so you can't just soar through it like a bird.

And that's why bees and other insects flap their wings a lot.

What they're doing is generating swirls of air above the top of the wing, and those vortices create low-pressure zones. When combined with the high pressure below the wing, that is what generates lift.

This robot was inspired by seeds from a maple tree.

Their unique shape creates the same swirling vortices above the seed's leading edge, and as they fall, they spin and generate surprisingly high lift.

These seeds are still just falling, but if you add miniature electric rotors to the ends of each wingtip on this robot, then it can generate enough lift to fly.

But this robot isn't quite insect scale. It actually weighs about 50 times more than the RoboBees in Kevin's lab.

So to power something that small, you can't just use electric motors.

I mean, the magnets and coils don't scale down effectively to such a small size.

So to power the first RoboBees, they had wings driven by special crystals called piezoelectric crystals.

By applying a voltage across the crystal, they contract slightly, but only around 0.1%, not nearly enough of a deflection to make a robot fly.

So roboticists designed a chassis that mechanically amplifies the motion 30 times.

If you then turn the voltage on and off 120 times per second, the RoboBee flaps its wings and flies.

But there is a downside to piezoelectric crystals, which is, they're fragile.

Even a small impact to the wings and the crystal cracks, and the RoboBee stops working.

So at MIT, they are building their RoboBees differently.

Well, these robots have a secret ingredient.

Instead of using piezos to drive the wings, these bees use soft polymers.

They effectively work like tiny muscles.

They take a polymer and they coat each side with carbon nanotubes that creates two effective conducting plates.

So if you apply opposite charges to these plates, that pulls them together, stretching out the polymer.

But if like charges are applied to both plates, they repel, and so the polymer shrinks.

And if we roll up layers like this into a tube, we can amplify the force they generate.

It stretches up to 25% of its length.

By cycling the voltage hundreds of times per second, these muscles drive the RoboBees wings.

- When you shrink down to smaller scale, your fly wing frequency goes up higher. So we are at the 400 hertz range. - Which is right in between a honeybee and a mosquito. - Yes.

This flexible muscle can take bumps and scrapes and keep working, but if it's pierced by a needle, the carbon nanotubes get pulled in and then the plates touch, causing a short circuit that renders the muscle useless.

But the scientists have even found a way around this.

When high current is cycled, the carbon nanotubes that are touching burn off and so the muscle self-heals.

Kevin and his team even invented a process to perform laser surgery on the robot.

- You're creating smaller defect around a very, very big defect and then by isolating the small defect, you're using the small defect to isolate the big defect. So that was what we call the laser-assisted clearing process.

One robot was really tested to its limits. Its artificial muscle was pierced by cactus needles and hit by a laser beam and it could still fly.

But these muscles are energy intensive and for robots at this scale that have to be so light, there's no room for extra batteries.

Luckily, there is another way to get around.

This RoboBee conserves energy by hopping.

This tech was used on another drone at the City University of Hong Kong.

Normally this drone can only fly continuously for 6.3 minutes, but with the hopping attachment, it can keep moving for 50 minutes, nearly 10 times longer.

Scientists believe this could be even more effective in low-gravity, low-air resistance environments like Mars.

So it would be perfect for an ingenuity version 2.0.

But microrobots are already being used today.

Every day planes complete hundreds of thousands of flights and most of them have multiple turbine engines.

Now a crack in a turbine can be catastrophic, so manufacturers inspect them every 3000 flight cycles or 180 days.

But inspections cost tens of thousands of dollars and can take a whole day.

That's where this cockroach-inspired robot from earlier, HAMR, comes in.

It's incredibly fast, it can run 10.5 body lengths per second.

Speaking in relative terms, that's faster than a horse, and it's versatile.

Its special foot pads can apply a voltage to polarized metal surfaces, creating an opposite charge underneath its feet and that's how it's able to stick to metal surfaces, similar to a balloon sticking to a wall after you rub it on your hair.

Rolls-Royce and Harvard are working to put HAMR inside of engines to inspect for turbine cracks even upside down.

And since its mass is so small, adhesion forces are much stronger relative to its weight.

So HAMR can get into some tight spaces and that can be pretty useful.

One of the first times that robots were deployed in an emergency situation was during the 9/11 search for survivors at Ground Zero.

Unfortunately, they didn't turn out to be that helpful.

They were big and expensive and they'd get stuck.

Three different types inspected eight sections of rubble, but none found survivors.

So an ideal rescue robot should be able to navigate tight spaces, withstand damage and debris, operate across varied environments, and be inexpensive enough to be replaced if destroyed.

- The material cost is actually quite low for making the robot. The human labor is high, but in terms of the material, right, couple of dollars per robot, but it's really not that much.

So the idea is to deploy swarms of insect-sized microrobots to search for survivors in disaster zones.

But I understand when I say swarm, you might get a little worried.

I mean, swarms of miniature killer robots are straight out of dystopian sci-fi. Think the hunter-seeker from "Dune" or the killer robot bees from "Black Mirror."

But this idea isn't so farfetched.

In the early 2000s, bees were dying off. It's called colony collapse disorder.

Congress is holding hearings, even the Vice President has been briefed.

In fact, the whole RoboBee project started with the goal of replacing the bees.

Thankfully that idea didn't last long.

- Bees can do much better jobs in terms of pollination than those robots much more cheaply.

To pollinate, you need a huge colony of bees to do those effectively.

Also, from an environmental protection perspective, I think it doesn't make sense to replace bees with robotics bees from a cost-effective perspective and also from the perspective of, you know, if you have so much money, why you making those bees than protecting the real bees.

Okay, so they won't replace the bees, but I can still easily imagine a world where these same robots that are supposed to help in a disaster are secretly being used to spy on me.

I mean, it's a bug that would literally look like a bug.

- We really focus on the fundamental science and solving the fun technical problems.

And as a society in general, we all should think about collectively how to prevent those new technology from doing harm.

But we're getting a bit ahead of ourselves.

I mean, most of the robots we've seen aren't able to spy on us. In fact, they're not even fully autonomous.

- We have offboard sensing from those cameras, you have offboard power from those, and offboard computation. What you see today is everything is offboard.

But hopefully in five years, then we can combine both sensing autonomy and power autonomy, and that's the longer-term goal.

Harvard's RoboBee has managed short bursts of untethered autonomous flight.

So it's fair to say we aren't that far off from robot insects operating freely around us.

Still, there is a limit to how far these robots can go on just batteries.

Batteries need shielding to prevent damage, short circuits, and leaks.

And the thing is, as batteries are scaled down, this shielding has to stay about the same thickness.

So that means smaller batteries become increasingly inefficient.

And that's ignoring that the energy-to-weight ratio of batteries is just fundamentally lower than that of chemical fuels.

At the insect scale, every milligram matters.

- We just said, like let's just sail past all of that and just use a video game cheat code and just power our robot with the smallest explosions possible and put two tiny internal combustion engines on board it, and it works.

And it sounds like a combustion engine, which is probably my favorite part of it.

Cameron's penny-sized engine runs on a constant stream of methane and oxygen.

This is fed into a chamber where it's ignited by a spark and so it combusts, releasing a burst of energy.

The hot gases rapidly expand, pushing against a flexible polymer membrane that acts like a piston.

- So the membrane moves as the piston and then instead of having to, like, have any sort of elaborate system that brings it back down because it just naturally is elastic, it sort of has its own restoring force. That was our clever little innovation.

As the membrane shrinks back, it vents the exhaust gases, allowing the cycle to repeat.

Despite the continuous flow of methane and oxygen, the fuel line never catches fire.

That's because as explosions get smaller, their volume shrinks much faster than their surface area.

This causes them to lose heat more quickly to their surroundings.

In Cameron's robot, only a small amount of gas burns at a time, so heat quickly escapes into the fuel line, cooling the gas and stopping the flame from traveling back up the line.

With just two of these combustion chambers on a little robot, one for the front legs and one for the back, Cameron can control its heading.

- You can actuate just one of the two sides at a given time if you want because both sides are operational.

So if we spark in both sides, it'll move straight. But if we just do one or the other, it'll pivot.

And this robot is super powerful for its size.

It weighs 1.6 grams, which is about as much as a gummy bear weighs.

It can jump like two feet in the air approximately. It can carry 22 times its body weight, which is about what a cockroach or a lot of beetles can do.

We'll be able to put a fuel tank, you know, microelectronics sensors, a camera battery, and still have weight left over to go, and this thing will still chug along.

That's the future. That's the goal.

Scientists have created robots that can do what some insects do and there are clear applications for this work.

But for these roboticists, these self-labeled misfits, that's not what it's all about.

- If it's about application, we should all like make a startup and try to like think about what we can do to make money, right?

We think there are nice applications, like inspection and search and rescue, but I would say as a research lab, we are mostly driven by curiosity. I think that's a very honest answer.

If you are someone who designs hardware like these robots, mini or mighty, you know that you need to create a CAD model first so you can prototype and bring your ideas to life.

But the problem with most mainstream CAD software is you need a very powerful computer and a mighty budget.

That's why we reached out to today's sponsor, Onshape.

Onshape is a modern CAD and PDM system designed for businesses.

And unlike any other professional CAD system, Onshape is built entirely in the cloud, powered by Amazon Web Services.

That means you don't need expensive computer hardware, it never crashes, and you never lose your work.

It just runs in your browser on any system, be it Windows, Mac, Linux, or even on your phone.

It's also built to be collaborative so you can work alongside friends or colleagues like never before on the same design in real time.

For businesses, you can trial Onshape for six months and for students' hobbyists, it's completely free to use.

Onshape also uses tried and tested Git-style version control methods from software design so you don't have to worry about sending multiple large files around and keeping track of all the different iterations. It is all done for you.

If you work for a government agency or have contracts with the government, I have good news. Onshape now has a plan that enables teams to comply with federal regulations like ITAR and EAR, so nothing is holding you back.