Hacker News new | past | comments | ask | show | jobs | submit login
Origami-inspired soft artificial muscles (wyss.harvard.edu)
200 points by lithander on Nov 28, 2017 | hide | past | favorite | 52 comments



I did my PhD with this group (Rob Wood). When I was there, these kind of actuators, and the robots you would put them in, were just getting started. It's great to see how far they've come.

One of the great benefits of these soft actuators is that you can embed them in soft structures and then get smooth movement in multiple directions. Instead of a rigid robotic arm with a few degrees of freedom, you could build something like a snake or an elephant trunk.

Another exciting area of research (my focus) is that since these actuators are fairly cheap, you could make lots and lots of robots with them. Think thousands. If you had a swarm of 1000 small robots, each of which has minimal power and sensors, what would you do with it? How would they coordinate their behavior? How would they communicate? For that matter, how would you even turn them all on? Swarm algorithms are fun to think about on robots, but are also useful for other problems out in the normal world.

(Don't focus on the "1000x" claim. It's true depending on how you measure, but it's not the exciting part.)


I believe soft actuators have hidden costs. While the actuators themselves may be cheap, the valving and pumps may not be.


Certainly. For what I said about the swarms to make sense, each robot must have a self-contained power source or be externally powered in a wholesale manner (generally by absorbing energy from the environment). For single (or few) robots though, you usually want the neat actuation mechanics and you're happy to pay the cost of the actuators and associated support hardware. The supporting hardware is always getting lighter and smaller though.


Well, since you've worked with those kinds of actuators I hope you can clarify something I wasn't sure about in the article: can the same muscle perform different actions? For example, could you have a muscle that can bend to the left, then to the right of some central line?

I'm asking because the statement [edit: in the article] that "designing how the skeleton folds defines how the whole structure moves" makes me think that perhaps the range of motions each muscle can perform is limited by construction.

Ahem. That's not to downplay the obvious usefuleness of such a device. As far as I'm concerned t's the first time in ages I find a robotics piece of news cool.


In general no, single actuators move in one dimension and then only in one direction. In humans we have muscle pairs, one contracts to open the joint and the other contracts to close the joint. Of course you can do the same thing with artificial actuators, but you're right, the really interesting stuff happens when you have lots of "actuators" (like a sheet that has a hundred individually inflatable cells) or the material that the actuators are embedded in folds in an interesting way. The properties of the skeleton have as much to do with the dynamics of the robot as the actuators do.


Thanks for answering. I looked at the video again and I can see that a couple of setups use multiple muscles- that's what you're talking about I think. I'm guessing that it's going to be a lot easier to wrap a hundred individual muscles of this kind around an artificial skeleton than it is with "hard" ones.

I wonder also what all this means for more, let's say, traditional robots- like the ones we often see from Boston Dynamics. I guess it's still early to say but if I understand this correctly, people can now make cheap, light robots. Where does that leave heavy, expensive ones?

It's not a rhetorical question- Ferrari and McLaren didn't hang up their spanners just because Toyota and Datsun sell lots of cheap cars...


It does seem from the description that only one motion is possible for one of these muscles. Full motion would require combining multiple muscles with complementary motion, similar to the way the body works.

The speed seems much slower than that of an actual muscle. Is this inherent in the technology or just a limitation of the current prototypes?


I can imagine this being modified for spring-like muscles, with a more elastic material. Other than that, you can just suck the air out faster (which has a limit) would work better with shorter muscle "segments".


If you had a swarm of 1000 small robots, each of which has minimal power and sensors, what would you do with it?

Do you need to ask? Look around, although many uses positive and negative would be found the primary two would be: espionage and “warfare” in that order.


Do you know of any way to "charge" the vacuum that drives these over a long period (from differences in air pressure?) so that they could passively build up energy and release in short bursts?


This is tremendous and probably what all robots that interact with humans will look like in the next 50-100. Affordable flexible membranes able to grasp and grapple with our real world, and also gentle enough (when programmed correctly) to not harm humans and other beings, this is the future right here. Very exciting. Focus on the information and less on the headline, guys.


I am extremely doubtful of this future. A big part of the reason we are considering soft robots today is safety, that if the robot hits a human it won't hurt them. If we can make robots smart enough that they never hit humans this is no longer a problem. Another reason is making things compliant so that we can grip objects because we have yet to figure out grasping. If we solve grasping, we no longer need compliant grippers.

In addition, pneumatics which this work focused on, are probably not the future. Pneumatics are not that efficient, are noisy, and are limited by the compressibility of air. The compressibility of air limits how fast these devices can actuate, their stiffness, and even how efficient pneumatic systems can get. Efficiency alone might be enough to encourage future robot makers to use something else.

Stiffness is another compelling argument against both pneumatic robots and soft robots. The max rate at which a robot can do stuff and react to things is dictated by its resonant frequency and mass. Sure we can make our robot very light, but we aren't going to be able to change the mass of things we desire the robot to manipulate. So it is still desirable to have robots with higher stiffness.

Really, a number of different technologies could make this obsolete within 50 years. For example, electric artificial muscles, slightly better rotary electric actuators along with rapid robotic assembly enabling stuff to have huge number of moving parts, or even advanced nanotechnology.


It looks like this is some kind of pneumatic system, and IMHO if you're calculating strength/weight you would need to take into account the weight of the air compressor, to have a fair comparison with biological muscle. Not to mention the fact that air compressors need to have an energy source, and are quite noisy.


The "big new thing" about this is that it doesn't use compressed air the way traditional air muscles do. It doesn't use compressed air at all.

You're probably going to say that vacuum pumps are noisy/heavy next. But this doesn't need traditional high-grade vacuum pumps, very low grade will work. And of course it is entirely moot for industrial machines that stand in place.

From what I can tell the trick comes from the pleating to massively increase surface area, giving atmospheric pressure more to work with.

I've been working with robotics (as a hobbyist) my entire life. This is the most exciting thing that I've seen in a while and I'll be building prototype knockoffs all week.


If you get something working I'd love to see a how-to article.


Heck, you could probably replicate it somewhat with a plastic baggie, some bits of cardboard, and a straw. Add some duct tape and hot glue, and you'd be set.


It could.

If you want to re-extend one of the muscles QUICK ;)


Potentially. But then for biological, you'd also need to include whatever system generated the ATP, oxygen, electric impulse, etc to power the muscle, right?

IMO, the crucial piece of information: how much pneumatic energy is needed per unit force for one of these, compared to that of more naive designs.


Given that the actuators are so light, it means that you don't need the extra power required to lift the actuator itself.

They indicate that they've had a 1000x increase, which means that to lift 1kg you only need force to lift 1.001kg instead of 2kg


I think that's a bit unfair. The importance of the weight of the muscles is how much has to be on a movable part. The compressor does not, depending on your requirements. It certainly doesn't need to move in the same way.


So its the limit, at scale. Build a big compressor, run a bigger machine with artificial muscle, the weight of the compressor becomes a small part of the whole.


Pneumatic to kinetic energy is much less efficient than electric to kinetic, so you're going to waste a lot of energy. And I guess the body is even better at energy efficiency


But apparently the efficiency of these muscle fibres are greater than that of electric ones. So maybe a win?


The 1000x comparison is silly - in a straight up lifting of things, a simple steel wire can probably lift 100000x times it's weight. The tech has it's uses and they should highlight the flexibility rather than perceived strength...


A steel wirte isn't lifting anything, it's just holding.


But a steel wire can be reeled in using a reel. And if we don't include the airpump required for these origami muscles in our "muscle weight" then we don't need to include the winch weight in the case of the reeled wire.


Only counting the steel wire in the winch example is unfair in the other direction. You have to count a small part of the winch (pulley?) for it to be directly comparable.

But this is all incredibly silly, and what we all agree on is that they fail to consider the vacuum pump, valves and reinforced vacuum hoses in order to make their invention seem fancier.


The proper comparison is not with the pulley, the wire or the winch, but rather the structure holding the pulley up.


i don't understand what this means. you understand that g = 9.81 m/s^2 ... every second? the difference between colloquial lifting and just holding is a matter of applying on the order of just 1% more force.


So if we took away 1% of that 1000x it's own weight it would be able to lift it? It can't. It will never be able to. Only thing I got out of my engineering dynamics class - ropes don't lift. Well, that and jokes about couple moments.

A lift is not a hold. A human can hold a ton of weight against gravity, but that's not them lifting it. See the squat. You can put a huge amount of weight on your back compared to the amount you can actually move. If you put them on an escalator, they could probably even move a distance with it. But that isn't them lifting it that distance.


A human can hold several loaded shipping containers stacked upon each other against gravity.

The human will be very flat and leaking all over, but it will be holding the containers.


"I guess he kind of bench pressed that steam roller... kind of."


He lifted it several millimeters off the ground!


as someone who stalls out at the midpoint of a squat coming up very often i can tell you that locking your knees with weight on your back is not holding anything - it's putting your posterior chain under compression. my point was that something like a barbell hold (like this http://www.myfitnessstudio.co.uk/wp-content/uploads/barbell-...) is just as hard as just curling.


There is a difference between a material’s chemical bonds resisting the acceleration of gravity and actually being able to lift an object from one height to a second, higher height.


I'm not talking about force, but about energy. To hold something suspended in air, you don't need any energy. But a steel wire can't lift anything, you need to have a motor (or muscle) that actually expends energy to move the object higher.


A steel wire can lift things all by itself. Just cool the wire. If you disagree with this, consider the motor or hydraulic cylinder that can't lift anything either. (Without an external source of energy like a battery or compressor.)

I'm half serious about this. The other main "artificial muscle" technology is nichrome wire after all.


You mean infinity more force, since at rest (holding) no force is being applied at all.


I think the gripping part is the important bit. You can't get a steel wire to open and then close on an object and lift it.


Maybe the fact that flexible, water-soluble materials can even show up in a comparison to steel is enough of a breakthrough?


I recall when I was building animatronics in the 90s there was a pneumatic air bladder "muscle" at the time wrapped in stainless steel weave basket. So instead of using a rigid cylinder with rod, when you inflated the bladder its length contracted. You could make an arm bend and the appearance of a bulging bicep! Somewhat similar principle, but animatronics doesn't use high-volume parts, so it never took off. These origami-inspired actuators seem really cool!


Neat!. I am working on artificial muscle as a side project[0], especially the kind you can (cheaply) 3d-print. Earlier work used these sorts of soft actuators similar to regular muscle-- you make a bunch of actuators that move in a particular direction, strap them to a skeleton, and then activate them in various combinations to move the skeleton.

3D printing, on the other hand, allows you to build more complex actuators (that don't necessarily apply force in a line). Origami-inspired designs (particularly rigid origami[1]) are related, in that you can design a particular folding pattern and have it fold and unfold to exert force in a particular way.

I was originally inspired by the work on artificial muscles actuated by a phase change (liquid to gas, with attendant increase in pressure) from Columbia[2]. Some combination of the two techniques might be better than either alone, allowing for fast-twitch soft actuators to fill the roles that servos stepper motors have previously occupied. Plus, they're likely to be cheaper in general, customizable to specific tasks, and probably safer in situations where humans might get in the way of the robot's motion.

----------

0. Most of the time I am working on reinforcement learning theory, and so building an actuator with difficult-to-model dynamics seems strange. However there's a lot RL could offer here, either learning how to control those dynamics from scratch or refining an existing model.

1. Wikipedia and its related/external links have a good overview: https://en.wikipedia.org/wiki/Rigid_origami If you just want a cool example of What Rigid Origami Can Do For You, check out: https://en.wikipedia.org/wiki/Miura_fold

2. See the press release: http://engineering.columbia.edu/news/hod-lipson-lifelike-rob... and the associated paper: https://www.nature.com/articles/s41467-017-00685-3


I cannot stop but recalling the last prodigy of Boston Dynamics.

Paraphrasing Elon Musk's words: This is nothing. In a few years, ...

https://twitter.com/elonmusk/status/934888089058549760


They are basically inflating and deflating balloons filled with some flexible structure.

So 1000x might not be that impressive, because they are very lightweight.

And I guess it requires a big pumping device in addition to the muscle (which weight is not included in the calculation, but certainly should be).


The weight of the compressor is important, but one compressor could be driving lots of 'muscles' via opening and closing valves, so it isn't completely disingenuous. It's analogous to, say, a robot with six motors, all powered from a single battery.


A compressor driving more of these hydraulic actuators need to be larger and heavier than one driving one hydraulic actuator.

Rather than battery and many motors, its more like motor with many transmissions and clutches. Vacuum is also a bad choice of medium... It requires stronger pumps, stronger hoses, better seals, etc.


So how much is that compared to a human. A 3kg biceps can lift 30kgs of weight, so artifical muscles are 100x stronger than human's?



I wonder if this would make a better penile prosthesis for erectile dysfunction...


"Artifical muscles can also lift, grip and twist objects."

I want to see it crush as well.


Reminds me of cell flagella.


So when do I get my Doctor Octopus style robo arms? :-)




Consider applying for YC's W25 batch! Applications are open till Nov 12.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: