Embedded - 189: The Squishiness Factor

Episode Date: March 2, 2017

Kari Love (@ikyotochan) spoke with us about creating soft robotics. You can see her edible soft robots talk from 33c3. Kari works at Super-Releaser. Her personal site (and blog) is Kari Makes. Kari mentioned that the Super-Release intern Aidan had some picks for soft robotics on Instructables. Super-Releaser created the Glaucus soft robot and Adafruit has an in-depth tutorial for how to make it. Some videos of soft actuators and soft robots: Super-Releaser Playing with Heat-Sealed Actuators (including the spiral) Silicone gripper from a cardboard mold Voxel Soft Robotic Simulation Evolution Super Long Mylar Robot MIT Tangible Media Group AeroMorph Soft Exoskeletons http://blogs.wsj.com/japanrealtime/2014/11/12/wearable-power-assist-device-goes-on-sale-in-japan/ http://biodesign.seas.harvard.edu/soft-exosuits http://www.roamrobotics.com/ Rat heart cell robot from Popular Mechanics First Autonomous Entirely Soft Robot (Harvard Octobot) VoxCad Tutorial for simulating soft robotics Also, if you haven’t seen Big Hero 6, you should. Consider it computer science homework. If you just want to see Baymax, here is a short video. Octopus: The Ocean's Intelligent Invertebrate (Elecia’s latest octopus related reading, the previous one was called Kraken)

Welcome to Embedded. I'm Elysia White, alongside Christopher White. Today we are going to talk more about robots, soft ones this time. Soft-bodied robots mean squishy robots. You know what else is squishy? Gummy bears. And no, actually, I haven't lost my mind any more than usual, but our guest is Carrie Love, and she's making soft-bodied candy robots and spacesuits. I'm sure it all makes sense. I'm confused already. Hi, Carrie. Welcome. Hi, don't worry. I'm confused too. I still haven't decided what I'm going to be when I grow up, and that's how you end up with these combinations. I understand that. And yet we didn't even mention costume designer and puppet maker.

Those all are all coming later. So there'll be more. Before we get there, could you introduce yourself like you would if you were, I don't know, attending a super con panel for Hackaday? Sure. So I'm Carrie Love, and I am one of the two collaborators at Superreleaser, a soft robotics consultancy. When we say soft robotics, people often get really focused on soft actuation, and it is one of our primary areas of interest. But really, we ask people a broader question, which is, have you considered a soft part?

Because sometimes you don't need to make an entirely soft system in order to have compliance, compliance in a physical sense, make a lot of improvements to your hardware. So we want to consult with people and find out whether or not we can add items or workflows that will create more robust, more simple, more novel ways to approach your prototype. The other things that I do are that I teach soft robotics at NYU's ITP program as an adjunct professor and I work wardrobe on Broadway. Right now I'm working

at Amelie, which is a new show that's opening. And I also have worked in spacesuits in the past as a NASA contractor. I am a partial owner of Final Frontier Design design i own a very small percentage of that and then i also have worked in puppet building and um as a puppet builder i've had the pleasure of working at a lot of different places including the jim henson company well we could just talk about that for six hours i know oh my goodness so uh i know you have an agenda and I'm going to ruin it. No, it's fine. Go ahead. So, soft robotics is an interesting term and it's one that I don't know that many people are aware of. So, can you give like a 30 second? I mean, doesn't everybody just skip to the robot in Big Hero 6, the super artificial intelligence and he's giant and he's very squishy i think the thing that's

interesting for me about baymax and being part of the soft robotics um conversation is that one not as many people saw that movie as should have no it was great movie really really wonderful but the other thing is i think that people don't realize that Baymax was inspired by a real robot. That the team that was working on the creative development for that movie wanted to take inspiration from real world robots. So they went to different labs all over the country. I don't know if they extended the labs all over the world even. And Baymax was inspired by the work of Chris Atkinson at CMU, that he was working on the idea of soft-bodied robots for healthcare, and that really captured their imagination. But there are also other robotics ideas that are in Big Hero 6 as well,

that they're working on at that institute, that are inspired by other sorts of real world robots. So I think that when introducing the idea of soft robots, they may have seen that, but they didn't really take it out of the fantasy space. Okay, so if we do take it out of the fantasy space, what do we get? That is a very fun question. And the reason why it's a little daunting at times to even define what this is, is that there's a really broad spectrum. So things like Baymax are inflatables. There are lots of pneumatic soft robots, there are lots of hydraulic robots, they are maybe made out of silicone, or they're made out of heat sealed fabric or there's ones that are made out of paper that are pneumatic and self-folding and origami.

But there's this whole other side of what's called the field of soft robotics that's in the biospace. So there are things that are made out of rat heart cells that are actuated inside their own power source, which is some kind of glucose. There is a very, very broad field of what soft robotics is. There's theoretical computational robots that are only predictive of soft behavior pixels that don't exist in the real world. And those are also soft robotics problems. So it really is just, if you are trying to engineer softness into a complex integrated system

that will utilize some computer control, that's usually what soft robotics is. But there's all of these kind of fuzzy spaces in between. And a lot of it is very interdisciplinary. So you'll have chemists working with biologists, working with people who are electrical engineers, and you're coming up with these bridging sort of systems and bridging problems. Did that make more questions than answers? Oh my God, so many questions. Soft robotics pixels. Pixels.

Pixels. Soft robotics pixels. What is that? I mean, is this a display? Yeah, what is it? So that particular work is being done by an integrated team. I would actually have to look up i know part of the team is at columbia university but i don't know where the other where if there are collaborators at other

spaces they it's a voxel uh based pixelated robot system where they have uh essentially their their blocks like little lego blocks digital le blocks, and they have been assigned different physical behaviors that it moves at some times, it has a certain amount of compliance or elasticity, and then that different ones actuate in different times and different patterns. But a lot of the work in that is about evolution. So in an example, maybe you start with a cube

and then you have the computer auto generate different forms, like different areas of softness and different areas of different patterns when each part actuates. And then they're given parameters. So the goal is for it to get from one end of the screen to the other screen, end of the screen as quickly as possible. And then it becomes an AI problem where they just run as many variations as possible. And they evolve over time to become more and more complex and being able to solve the problem more efficiently. They're really interesting. Yes. Yes, they are. I'm not sure I got all of that, but I want to actually go on

anyway. Whenever you say compliance, you mean squishiness, softness, the not, I mean, we talked some about regulatory compliance, and that's totally different. Yes. And actually, it's kind of something that I struggle with a lot talking about soft robotics, because a lot of the words that you would want to use to describe soft things are already taken by other fields. So you say soft and people jump to soft where you say flexible and they're like, oh, flexible systems or you say compliance and they think regulatory compliance. So I really do struggle with finding words that aren't loaded with other meaning in different domains. When I talk about softness, there's kind of several different factors that I point out to people when you're talking about material softness. So there's the flexibility of the material. So if you think of some thin plastics or even thin metals, fabrics, those things are flexible. You also then have the durometer, which is like what I would call like the squishiness factor, like how hard is this

thing versus how much does it compress? And then there's also elasticity. So does it elongate? Does it stretch? And that these components that we would all call soft embody different physical characteristics and possibilities. And that you can have things that are one of these three and not the others. So you can have something that is, yeah, so you can have something that is, has a squishy durometer, is flexible and elongates and is elastic. So, or you can just have one. And then, so that changes the complexity of your problem and your solutions. Well, you have, I mean, you definitely have terminology problems because the word soft has like a whole bunch of different antonyms. It can be hard hard it can be rough

and so you you have to distinguish i mean you didn't even mention is it plushy is it physically soft like fur or is it sandpapery and that probably doesn't work so much in soft robotics but if you were building something medical, you might want it to be non-slip for prosthetics, or maybe you want it to be soft like a plushy animal that you can snuggle with. And so you do have even more terminology. Yeah. But let's go on to soft actuation.

I had this idea that I would explain octopus physiology, given my very limited understanding of it. And you would tell me if it's the same model you would use for soft robotics. Does that work for you? Or do you want to describe it from like first principles and not involve cephalopods whenever possible. But I did think about whether or not that would be illustrative for the particular problem space. And you end up against this, like, very unsatisfactory answer, which is it depends. And I think the it depends in this space is kind of interesting to me too, because there's this idea where a lot of roboticists were really interested into biomimetics, so actually mimicking nature. And so there was a lot of effort, and there's still a lot of effort being put into directly copying systems from nature. But there has also been the growth of the idea of the bio-inspired design. So you're not trying to like slavishly reproduce what exists in nature. Instead, you're taking principles from nature and trying to extrapolate that into something that's simple and usable and

applicable to a particular problem. So if you tell me how a cephalopod moves, I can say, yes, you could create a soft system that mimicked all of those things. But probably what is going to be more functional is to take some bounding around that and say, what would get me a more simple solution than the exact copy oh yeah because they're complicated they've had millions billions of years to get it right for their situation um i just i don't have a good mental model of even how you would start soft robotics. I mean, all of mine go back to putting air around, encapsulated air, putting an air-filled balloon around a hard skeleton, which is not what soft robotics is really about. So where do I start? Where do I even get the concept of what we mean? Let's see. So if we, so we're going to narrow down a bit. So like I said, it's,

we're starting out a field that's so big that it could be made out of like literal biological cells, or it could be just a chemical system and, and not involve any kind of traditional engineering solutions. Let's say we're going to just talk about pneumatics and have an inflatable robot. In this system, it ends up being that you, it depends how complex what you want it to do is. And then the way most people go about solving these problems are through direct prototyping. You have some concept by observing something in nature or having built something before where you know that it has a property. One of my favorite simple soft actuators is just two pieces of airtight film and then you cut them into a spiral that has volume and then you weld the

edges together and then you as you blow it up what happens is when it's not filled with air and you hold the top the whole thing just hangs down. And you have like kind of a wavy sort of piece tubing, essentially. But as you blow it up, it becomes more stiff. And as it becomes more stiff, it goes back to its original flat shape before gravity worked on it. And it pulls itself up. The spiral tightens and pulls up through the air and by going through that motion it lifts itself the distance back to where the air supply is coming from um the way in which i predicted this would move

had a lot to do with understanding how spirals work in fabric and then extrapolating like how inflatables work. There is more of a movement now to be able to computer model these things. But we're in the part of the understanding loop where the heaviest lifting is done is in the iterating process, the physical iteration. So you build and test and build and test and build and test. And then once you have something that does something interesting, then maybe you want to model that. And maybe you turn that into a design principle that then you turn to an application. Does that make sense as a process or a way to understand the problem space? It's actually really hard without visuals.

I'm so happy this is a podcast. Common problem. I guess that makes the process makes sense. But I got lost in imagining spirals. And, and I think maybe I want to go back to that because it was important. So I cut out something that looks like a spiral, and I end up with a balloon that when it's flat, it's in a spiral shape. Yes. When it's unfilled, I guess. When it's sitting on a table, it looks like a spiral shape.

Okay. And then I put air into it, and the air makes it go flat. So right now, let's start where a place where you're visualizing, when it's sitting on the table, your balloon is a spiral. Okay. But when then you take one end and you lift it up off the table, you just have a long dangly tendril. It is no longer a spiral. It's just there.

A pigtail. You can't see that it's a spiral. Oh, all right. It's just like a wavy looking tail. Okay. Okay. Now, as you inflate it, the air pressure causes it to stiffen and as it does the stiffness makes it want to return to that spiral shape

and as it's going against gravity it's pulling itself up and up and up closer to the air supply until it's a flat spiral suspended in air it's it's now a flat sheet again so the air supplies at the top okay and so we start out with something that looks sort of like a line-ish like thing and as it gets more air, it becomes more spring-like until at the top it is now a flat coil spring flat spiral. Okay. Yes. And because it is this shape, we can use it like we might use a spring. Yes. Now you can exactly, now that you know what it does, you can apply it to a problem.

And you can, I mean, you can change the spring constant depending on how much air you have. And that actually now gives you a whole realm of possibilities. Exactly. Okay. I'm in for that. I got it. So now kind of to take another kind of, a more classical shape. I started with one of my favorite shapes.

And also I like working with heat sealed materials. So like it's a, it's a realm I enjoy, but, um, kind of a very classical soft robotics problem is the idea of like a gripper. A silicone gripper is a very, very popular sort of soft robot. Because there's so many applications. You think about fruit and being able to move fruit from one spot to another or grip. I mean, even holding hands or yes. Okay. I'm on board with grippers. Might have called it a claw, but a gripper is a better word. Yes. So a soft robotic gripper is like a very, very popular problem space. It's something that people are trying to commercialize and you'll see lots and lots of versions of them across soft robotics.

And there's various ways in which you can solve that even within that problem space. It's kind of wonderful thing about soft robotics is like there's a million ways to solve the exact same problem. I think it's probably true of every element and every domain of engineering that that's true. But I think because soft robotics is so new and everything feels novel, like any variation feels like totally different. But in those particular systems, a lot of them are relying on the elasticity of something like silicone. So in those systems, you fill them up with air oh you know we should start with a classical model let me start again so a very classic soft part is a McKibbin muscle um I assume you're familiar with McKibbin muscles would you like me to cover

what they are okay McKibbin muscles are a kind of me to cover what they are? Okay. McKibbin muscles are a kind of pneumatic system that was invented in the 1950s and by a guy named McKibbin, surprisingly enough, who wanted to apply them to prostheses. And they are a air bladder that is then covered with a kind of woven fabric or some kind of other structure on the outside that as the tube inside inflates as it grows in cross section the weave causes the piece to shorten in length. Inchworm. Yes. And so that's how you get it to do work. And so like this is a very, again, a very classic soft system.

They're applied all over the place. So the ways that you are dealing with that system is that you're causing something to constrain the elastic material, and that is changing the performance. You can do the same thing by constraining just one side more than the other side, or you can constrain in strips around it, which creates ribs. And by doing that, you can then make bending actuators. Because you are putting pressure on one side differently than another. Yes. And then the uneven inflation allows it to bend. Okay. I can get that.

When I was preparing with the octopus, I had this idea of talking about partially filled water balloons, the long ones. And then you put it in a little bit of water, but you don't put any air in. And you can push it. And when you push the long, so you have the long part and you push it on the shorter. You have a hot dog shape and you push it in the middle of the hot dog because it's a water balloon. It just squishes out. The ends don't really move much.

And what you're saying is if you squish it a lot in the middle, you hold the hot dog in the middle. Chris's head is in his hands because he hates when I do this. But you're squishing a lot on the water balloon and the ends have to get further apart. When you squish in the middle, they get further apart. When you release, then they get closer together. Am I building the right mental models or should I be really thinking about air? Water is a little easier for me. I think we should move on from the specifics because it's very difficult. The specifics I think are probably our problem here because the mental models you're

building are correct. And actually what you've just switched to is that you like to think about hydraulics more than you like to think about pneumatics. And that's perfectly reasonable. Like you can touch water. It's a little easier to get your head around. I think the thing that is the truth about this conversation we're having is that the way you understand how these soft systems work is actually that you observe them. You observe them in nature, or you observe them in the laboratory or in your garage. That's how people are coming to understanding about soft systems. Are there YouTube videos or maybe you can recommend a book or something? Oh, yes. Watching YouTube videos of soft robots is a wonderful rabbit hole and people should totally go down that. Okay. I'll get a list of some of the ones you like best after the

show. We'll put it in the show notes. Excellent. Yes. I think that when you're building these mental models about soft robots, yes, the answer is if you have seen a soft system and a soft system can be something as simple as a water balloon, then yes, that is how you're coming to understanding about how different materials work under different amounts of pressure. And the thing that's interesting to me about the water balloon example is that, so in that one, you're dealing with, again, something that's flexible and something that is elastic. And as you deal with elastic systems, a lot of that starts to depend on the thickness of the walls. That if you have a thicker wall of something that's elastic, it's a lot more stable. And if it's thinner wall, it will distort more. So like the complexity mounts very quickly.

I mean, I see why more thickness would add less elasticity but why does that make it more complex that just seems like a constant in the system you have to figure out how much force to use or how much resilience to breakage you have so um it's not that it's thicker is more complex it's that if you have a gradient of thick and thin. Oh, inconsistency in manufacturing is always complex. Yes. Okay.

Oh, not consistency. Intentional. Intentional. It's a design parameter. Yes. Because you want to be able to be flexible over here, but not over there.

Yes. Like, yes. Like my neck wants to be flexible, but my but not over there. Yes. Like, yes. Like, my neck wants to be flexible, but my cranium does not. Okay. Yes. Christopher's head is no longer in his hands, but he still looks doubtful. Christopher, what are you doubtful about?

Nothing, nothing. I'm just enjoying listening. Should we talk more about actuators and different types, or should we go on and talk about candy? I feel like if we spend a lot of time on actuators, it's just going to be confusing. Well, maybe we could talk a bit about building them up into more complex systems. Because maybe describing some of the systems you can build out of actuators might be... Well, now we have an inchworm and a spiral spring. Right. What do I do with those? Well, the inchworm can walk. The spiral spring can spring.

We can build Tigger from here. Sorry. Go ahead, Carrie. So let's talk about why you would want a soft robot. Yes, thank you. That's what I'm trying to get to. Okay, great. So there are lots of reasons you'd want a soft robot. And then there are some downsides why people don't want soft robots. So we can talk about both. So people want soft robots because we would like to be able to collaborate with robots. So we would like them to, for one, we would like to be able to wear them on our bodies. We'd like to be able to do that because there's medical utility maybe from, you know, things like gait training robots. You may want to wear a robot

because you want to be stronger or faster. And that once you put a human in a hard robot system, there are many reasons why that might be suboptimal. So one is the human factors, like humans are soft and hard together. We are not hard systems ourselves. And when you try to put us in hard systems, we are uncomfortable. We can be injured by it or injure ourselves with it. And also, you end up with this mass versus power problem. Like you have to have a power supply that

is so large in order to move the massive hard robot that then you're losing the gains that you're trying to get so another one uh advantage of soft robots is the one that you talked about which was like when we talked about grippers we're talking about fruit so fruit is a wonderful problem space to think about because one fruit is soft you you it needs to be handled delicately and the other is that fruit is irregular. There are no two fruits that are exactly the same. And so traditional hard robots try to overcome these kinds of problems with very robust sensing, but that sometimes you can get around some of that complexity by actually making this physical system soft.

You can have more flexible applications and handle a wider variety of materials. One of the popular ideas of why you would want a soft robotic gripper would be kind of these new large warehouse systems where you're packing all kinds of dissimilar items. So how do you get a robot that can handle all of those differences? Another reason would be environmental constraints. So the Navy is very interested in soft robots because there are all of these very, very successful biological systems in sea life. The majority of sea life is either all soft or soft and semi-soft, like fish tend towards cartilage rather than hard bone. And so there aren't a lot of hard moving systems.

So when you start looking at things like submarines, there's some feeling that they have already reached the limit of how successful they can be in these hard-shelled vessels. So they may be able to have a lot of advancements in terms of speed or stealth or ability to use sonar in some ranges where they can't use sonar now if they were able to switch to a soft system. What else would you want? I cannot wait to drive my personal jellyfish. Me too. That sounds really amazing. And then also space. So I'm interested in space problems and have worked on space problems before. And so when you're dealing with space or undersea or like in a volcano, you know, wherever these like extreme environments are, you suddenly have other kinds

of problems. So for example, in a spaceship, you're not just limited by how much weight you can put up because the fuel is very expensive. But you're also limited by the stowage space. So you often want to build something that's a lot bigger than something that you can launch. So by adding kind of soft or flexible components, maybe that you can pack these things much more efficiently and then be able to deploy them. Okay, that all makes sense. One question that comes to mind to me is, why didn't we start this way? It seems like mimicking nature is our natural way to kind of figure out how to make devices and other things that act the way we want them to. It seems like we would have started with soft robotics instead of gears and metal. Was it just a materials issue historically, probably? I think it's a predictability issue. Okay. So the complexity of these soft systems is the thing that's one of the cons.

So people, they're nonlinear systems. So that's what people always say. It's a nonlinear system. So that means that we can't model it under our current understanding of physical modeling or computer modeling. So it is much more dependent from things like manufacturing error. So variations and consistency become a big hurdle in terms of these kinds of behaviors. And then I think that also there's the durability issue. So, a lot of the soft materials that we have had are really averse to friction, like wear from friction or puncture or kind of all of these normal day-to-day hazards that things will be running up against.

I mean, that makes sense. I mean, if I go back to my water balloon example, there are many things that go wrong pretty easily with that. And of course, pneumatics have the same problem that you don't even know that there's a problem sometimes, depending on what state you're already in. But you mentioned mass in wearing hard robotics. How does a pneumatic system make that lighter? I always think of air generating pneumatics as relatively heavy and expensive power-wise yes and i think that that is one of the arguments again in the against the current so i think that um there are some systems right now where it is a pneumatic system where they're putting on people's bodies. But then in that case, even though the pneumatic system to generate the air is heavy,

you can switch to things like compressed air so that you're not generating the air as you go. And so you could switch to that, and then you'd have some way where you'd switch out the cartridges. But the thing is that even though it is heavy and power inefficient to create the air, you do still need less power than you do for an equivalent hard system that's heavy. Some of the systems that are wearables are actually like a more cable driven. So they're like spooling up cables to pull them along the body. And those ones I think are pretty interesting to me because you're using the hardness of the existing bone structure of the human to provide the predictability on like the hard components to anchor against. Yeah, okay. Cable things, I always immediately want to lose my cables because frankly, I lose my cables because frankly I lose my cables.

With pneumatics, if you're cabled to a spaceship, a space station, then that's one thing, but if I have an exoskeleton that is helping me with fixing my gait so that I don't walk so that I damage my knees or something on that order. That's something I want to be able to walk on different surfaces and try out in different places. So I think we need to figure out how to put the pneumatics on the person. But what you said was important. Compressed air is a thing and you don't necessarily lose it in these systems. You don't need to generate compressed air. You just need to pull it from one place to another.

Is that right? Yeah. Well, it depends how leaky your system is, intentionally or unintentionally. Like you can build leak into the system as part of the operating parameters, or you can try to make a less leaky system so you can keep the air and move the air over different amounts of time. But I think that when we're thinking about soft robotics right now, I think it's often helpful to think about the fact that it is such a young field. Kind of as people define it right now, it's only like 10 years old. As you look back on the history of soft robotics, there's things from the 80s, there's things from the 50s that we would now fold into this field. But it hasn't been a field that we would call it such until quite recently.

So I think a lot of the finding the application space for it right now is driven by finding an application where we know that the constraints that we have are still functional. So when we talk about like gait training, the easier to market thing may be that you're building a gait training rig that gets used in a physical therapist's office rather than one that is like totally out in the field. So you have these like parallel discussions between like what might be able to be commercialized in the near future and like where we can see gains right away. And then you also have like the blue sky side

where people need to make like really big assumptions about what will be possible in the future and then do the incremental work to get there um because there are tons and tons of possibilities but you have to do kind of all that nitty-gritty short-term maybe we're still tethered um steps and i think that you know there aren't that many robotic systems that are like human walking systems that are untethered now either. The tethering question is a question that we still have for some realms of traditional robotics too. Yeah.

I worked on an exoskeleton. It was hard. And we did get it untethered, but it was non-trivial. Let's go with that. And it had to carry its own weight so yes yeah that is certainly a big consideration when you're doing that sort of thing yeah so I think in soft robotics it's that you're tying all of these like new problems to an already already non-trivial problem yes everything everything gets more complex as time goes by because that's what we do. But so making candy robots, this seems like a hard problem.

And yet now you're just going to use some gelatin. Okay. So the story about the soft robotics, candy soft robotics so, uh, I have not traditionally worked in silicone a lot, but my, my collaborator, Matthew Borgatti, uh, at Super Releaser, he has a lot of deep expertise in silicone casting and, um, some of his popular work, including the Glockus, which is a, a single cast piece of silicone with two lines in that can walk as a quadruped. So he and he and I have worked on kind of more simple silicone robots together. And he's been teaching me about his area of expertise. And during one of our conversations, because we're writing a book right

now, we're co-writing a book, he's going to write eight chapters and I'm going to write four and we're working on the projects together. It came up like, how are we going to structure this? Like, what are we going to demonstrate? So we were coming up with different projects that demonstrate like some foundational piece of soft robotics. And during the discussion, it came to mind to me that I was like, oh, you

know, silicone is a lot like gummy. Like it feels a lot like gelatin based things. So wouldn't it be really great if we made a candy robot? And then I was like, oh, I don't know. At the time, were you eating gummy bears? I mean, I may have been. I really like candy. But yeah. And so I was like, well, maybe we should make a candy robot. And then I was like, but it doesn't really meet our agenda. We already have these silicone robots that we're putting in the book. And he was like, no, part of the agenda is to be awesome. And so that's how I got put on the path of actually following up on the crazy idea of making robots out of candy. Making robots out of candy.

I feel like that should be the show title, even though that's way on the nose for us. Yeah. Okay. So, yeah. Candy. It's like jello shot robots. Tell me more about jello shot robots. That is an excellent idea.

So I will totally consider that. I can see how there are some similarities between silicon, the molding materials and gummy bears um there's a they have the elasticity and they have uh i want to say the lack of denseness the squishability but how i mean you're not making the gummy bears dance are you have you considered that sorry so um so the way that i like to work on problems is that because it is so much of the process is materials driven for me. I think that has a lot to do with the fact that I've worked in lots and lots of different materials over my careers. And you really hit this place where if you choose an inappropriate material to build something, it just it never works right. And you're you're fighting it the whole process. And so for me, when I wanted to tackle the problem of making candy robot for me, it was very process driven. So so the process that I like to think about when working on emerging technology, it starts with why. And so in the case of Candy Soft Robots, the why was as an art project. I really, really love edible art. I think that people talk a lot about interactive art and I can't think of anything that's more interactive than food because it engages all five senses and then it becomes a part of you. Like I literally can't think of something more interactive than that.

And then the other reason was like the practical why, which was we could put this in a book. And then the other why is that like, it's delightful, like it's delightful and or disgusting. And these are our strong emotions. And like, if you're going to make something, making something that elicits a very human response is appealing to me. As I move then through the next stage of what I want to do when working in an emerging field is that I want to talk to experts. And for candy robots, I was like, there aren't any experts there. You know, it's it's it's new as a as a field and putting air quotes around things that you can't really see. And so the thing about this was like, well, who are the experts?

I was like a food scientist, a soft roboticist. I ended up speaking to a chemist for a good long time. Um, a person I know who's an extreme candy enthusiast. She's eaten every mass mass produced candy in North America, as far as I can tell. Um, so I talked to all of these people and was like, okay, great. Like, let's talk about what this means. Like, let's build out my understanding of candy, um, generally and like soft robotics, um, with kind of a more focus when we're narrowing in on making a candy robot. And then, that, I define the specs. And so in this case, some of my specs were like, it can't be only technically edible. So people were like,

well, you can eat string. And I was like, no, we're not going to make a robot where we're going to eat string, even though yes, technically it's edible. Another was that I wanted it to, again, like either bring joy or disgust. That was another one. One was that I wanted you, I had seen some various edible art projects that were like edible electronics where you essentially had to eat around the wires. And I was like, I don't really think that that achieves my like aesthetic goal if you have to eat around something so um for that then I was like you have to be able to eat up to where it joins to the computer control so I wanted to for example if I do a pneumatic system wanted to use airtight tubing that was also edible um and so then once I have the specs in place, I start a materials exploration. I look at

all of the materials that seem like they might fit the problem space and I start playing with them. So in this case, I started both with making homemade candy and I started with store-bought candy. I did both paths simultaneously and I did a bunch of materials exploration where I just tried to find out what are the characteristics of these materials and then could I apply them to this problem or not. So I looked at licorices, I looked at, I didn't limit myself away from the idea of a hybrid robot, a robot that has soft and hard components together. So I looked at like drilling out holes in sweet tarts. I looked at fruit leather, like is fruit leather airtight? I made many different recipes of gummies. And I also, yeah, so I went on this kind of big, broad exploration of all different

candies that I could apply to the problem. And then once I get through doing the materials exploration, I see what's promising and what's not promising. Like a lot of the candies I bought, ultimately, I was like, okay, now I'm going to give these away. And I'm going to be the person who gives lots of candy to people. That's also a nice side effect. It is Halloween every day, especially when I got in a time crunch and I was like, I'm just going to order a case of Nerds Rope. And so I gave a lot of Nerds Rope. So anyway, that's kind of the stage I am at right now like i'm built out one simple actuator that was like multiple kinds of licorice uh which actually made me laugh because people who hate licorice was like you have gone into the technically edible zone oh they were like black licorice is

not food and i was like oh i really like black licorice. Okay, so let's, no, I want to, this licorice thing, I saw it on your video. You gave a talk and you had it. And it was a red vine on the outside, so a hollow licorice, which could make excellent tubing as well. And then a solid black licorice vine that was not hollow solid. And then you threaded the black licorice through the red licorice and it bent. I don't remember how it bent. It had little voids cut out of the red vine. And so because it's flexible, it's wherever you don't have. Oh, you cut out vents. Yes. Okay. And then when you pull, so did the black licorice, it was connected or it couldn't retract through the hollow licorice?

So as you pull on the black licorice, it it was a spiral piece of black licorice so i left the spiral end to be the portion where that's the anchor point for the pole and so as you pull the black licorice cable the red licorice um where the where the vents are that's where it bends and then it slides back and that was funny too because um then i started exploring edible lubricants for candy because i wanted them to slide nicely and then almond oil and olive oil would be pretty good but and that's hydrophobic so it shouldn't get in with the candy with the sugar you know i had that theory too but essentially oil also works as a solvent on sugar. So that was something that happened.

And then I talked to a food scientist about it a little bit and he was like, you know, pretty much most liquids that sugar dissolves so easily, it just, it's a property of sugar itself that almost anything liquid would would cause the sugar to become gummy or melt and so i had to switch to dry lubricants only such as crushed up sweet tarts such as uh yeah i was using um corn starches powder sugars um someone suggested that i could should use um the sour um powder that that might be really delicious yeah citric or acidic or yeah one of the acids citric yeah there there's like 10 different interchangeable edible sour acids that you can buy if you go to the like

molecular gastronomy stores okay and then the other thing I saw in the video was you were talking about fruit leathers, which is, I mean, fruit rollups, really common pounded fruit into jam and then dehydrated so that they're flexible. They're not particularly durable,

but you can definitely make a bladder with it and connect the edges. Either using a sugar syrup or sometimes just licking them. You can create little pockets. And I know that because I have made the pockets and then filled the pockets with other things. But since we were talking about soft robotics, you could make the inchworm sort of thing if you made a pocket and could pump the air in and out have you tried that i made essentially what looks like kind of a heart so since i only got through the materials exploration what i what i wanted to test was whether or not i could seal it and it would work like an airtight film and the answer was yes i made like this thing that kind of looks like a heart because when you pump the air in, it expands and then it contracts again.

Yeah, I think that the thing about going down this materials exploration path is that it's really, it's really provides the answers to what I might want to make. I didn't want to start where like I already pre-designed what the robot was going to look like or what the robot's function would be or what it would move because I can come up with a better idea once I know how the material behaves. And so the question that I have for myself about like now that I'm on this path and I'm making all of these things is um okay now which of these is is like the most fruitful and so that's kind of again talking about the process of fruitful intended um and then as so now that I I have all of these things that I've tried now say like what do the trade trade-offs? So what are the trade-offs of

going with a fruit leather versus a gummy versus, um, versus like even the trade-off for like the most delicious gummy versus the most functional robot gummy is like an interesting question for me too, because I'm, I'm writing a project based book with someone. So I need to make sure that the project is easily reproducible. So that's another one of my specs. You talked about durability. And that was something that in my specs, I don't want the robot to be durable. I think that actually when working with food, making something that lasts a long time is actually not a desirable trait. Well, no, but they should be able

to actuate it five times it shouldn't be one thing and then it disintegrates that's not as much fun yes so so my spec was that for that was like you need to be able to make it the day before you want to present it so it needs to be durable enough from like a static standpoint to last 24 hours because i don't want to put anyone in a situation where like you have an expectant room of people who want to eat your robot and you're like frantically trying to make it and then it fails and then you've had a fail. We've all been there. Yeah. The demo never works on the day of the demo. Exactly.

So I wanted it to be that you could make it the day before. And then for you to be able to actuate it, maybe half an hour would be ideal. We'll see what the trade-off is. Maybe I can come up with something that's delightful, but you can only cycle it for 15 minutes. It probably would be fine for this particular application. And then, but I didn't want the spec to be that it lasted a long time because like nobody wants old and dirty food that's been actuated everywhere.

And like you kind of lose your opportunity too if no one gets to eat it at the end. So if it's so durable that it becomes so precious that no one will ever eat it, then I think that's a missed opportunity. And yet I want to make little gummy worms that crawl across the table and I'll just sit on the floor and look up with my mouth open and they'll crawl into my mouth and it will be horrible. And Christopher's face once again, I wish you all could see the look on his face about how horrible

this sounds to him, but I am jazzed oh so um yes definitely crawling into your so things that have have come up when that's the other thing that's really wonderful about this project is that people are really eager to talk about it um so definitely in the disgusting side some of the things were like eyeballs that pop out of something hard or like um uh there's a cling-on dish in star trek where it's like a plate of squirmy things and they're like oh can you make that and i was like actually that's probably not a very hard problem um like of things that we could do with it making a plate that kind of wiggles in an indistinct way would might not be a hard problem compared to some other things. Another one was like a snot-based thing, like snot coming out of a nose, but it was actuated

kind of on the disgust side. And then kind of on the delight side, you had said like dancing gummy bears, like that is a delightful kind of interaction. And like both are interesting. How are you, assuming pneumatics, how are you going to, how do pneumatics work at all? I mean, I know how a pneumatic drill works, but we're not talking about that level of, or a pneumatic hammer hammer that's the word right jackhammer sure uh i know i know i know i can go to home depot and get this giant loud thing that will make compressed air but i don't want that for my party of dancing gummy bears how do how do you make something small that will do what you want it to do or is this a bike pump i so a lot of times when prototyping will use like a sphygmomanometer it's the like little bulb uh you know that's the part of the air like a turkey baster yeah like a

turkey baster so the one i was thinking of and the one that I use is for blood pressure, yes. And so it's a little bulb pump and you have cable. You have tubing that you connect to that. But also you have used compressed air a lot. So the compressed air that's for beer making. Oh, okay. They make little cartridges so you can hook that up to some little

valve systems. So it just depends. A lot of my prototyping I will do with a hand pump though, just because you want to be able to iterate quickly. And since I don't have a specialty in electronics or code, I often will collaborate

with someone who that is more their area of expertise and I will handle more of the like soft goods and embodied complexity side. How much does puppetry and the ability to make interesting joins with fabric that do non-obvious things apply to soft robotics? At least the scale of robots that I work on, which is kind of this small size onto like human wearable stuff. Yes. I think that the fact that I have worked on costumes and on puppets is the entire way I have for visualizing these systems. So I know a lot about human factors and I have also worked with lots and lots of different materials and kind of learned tricks of the trade for motion and now can apply this

just to a new space. I have a small hand puppet and I really like it. Christopher still doesn't know why I got it, which is fine because I don't know why I got it, but it's sort of a not well-built frog. And I had to play with it for a long time to realize what expressions I could get because of the way it's built. There's some fixed things about it. And I suspect those are design choices from the puppets, puppet masters, puppet makers, puppeteers. Puppetologists. God, that's a terrible word.

I usually say puppet builders or puppet designers. It was the puppet. Puppetographers. I can't get this. Sorry. Yes. So, I mean, this is also an origami problem in some ways that you have a fixed surface and you want to be able to move it in certain ways. And you mentioned vents as one way that allows things to move in one direction. How do you learn this stuff? I mean,

I play with origami, but even as I thought about robotics, I was like, I don't know how to get from here to there with the origami. And I was looking at hard robotics, little tiny actuators. How do you learn this stuff? I think that at least in these kinds of problems, it's that everyone is bringing all of the things they've already done. So you'll have a toolkit of all of your knowledge base and then you are finding a collaborator or you're collaborating with different parts of yourself even to say like okay i know this from this and this from this like what happens if i blend these two pieces of knowledge and then that's how you make something novel so for me um i because i have worked on the costume problems and the puppet problems and the spacesuit problems, like I have a very deep toolkit for physical prototyping and materials understanding.

And so I'm just like, once I have a new problem, I'm sifting through all of the things I've experienced before and say like, this looks like this, like maybe no one else would make this connection. But like, let me figure out, of all of the things I've tried before, which ones speak to this problem. And then if I look through my toolkit, and I don't have anything, then then the next step is that question of then then who is the person I talked to? Who does have the tools that I need to get this done? That's totally fair. But I was more going to a, if I wanted to start, where do I even start? Is this learn to sew and learn to make puppets? And then that will get me further into soft robotics is this understand silicone

better and start making more jello in my kitchen it depends again it depends on what kind of robot you're interested in if you were interested in silicone so i um i just taught a bunch of students about casting silicone. And they had these questions. They said, okay, so you taught us how to cast silicone. We made one actuator together. Now, how do I do more? And I said, okay, well, probably what you're going to want to do is, one, cast more silicone things because you will need stronger skills in order to reproduce your work and to get the outcomes you want. So that's one phase of it.

The other phase of it is that there's an existing body of work. So look at all of, if you want to work in silicone now, look at all of the silicone soft robots that exist and say like, okay, what design patterns are people using that are working well? So some of them are shapes that you learn the shapes from this library that people are unintentionally creating by making all of these soft robots. And then you like, you look at thickness. So some areas are silicone, uh, thick for the stability. And some of them are thin. Some people put fabric inside for constraining

things. Some people are wrapping things around the outside to change the way they inflate like some people are making them like McKibben muscles so you're like okay so this is now a library of ideas by looking at what other people have already done and then also then maybe you're wanting to look at this bio-inspired space you're like I want to make something totally new if I look at the natural world and the majority of things are soft or soft and hard together is there something that is really interesting to me that I can simplify to the point where I can actually make something that approximates part of its function and then then you try and then you make one and it fails and then you make one and it fails and then you make one and it fails.

And then you iterate until it works. And then you've discovered something. I guess. You're telling me there are no shortcuts and I'm going to have to actually learn something and try it out? Yeah, I guess the thing is, too, I guess maybe the answer is there are no shortcuts. But I guess the thing for me is that it's also not a mystery. Like, it's a process. But you can learn the process.

And also, I guess one of the steps that I forgot is that when looking at the silicone problem, I think people get really narrow in where they want to look for ideas. So they'll say like i looked at all of the silicone soft robots coming out of universities and i said well did you look at all of the special effects artists because they have done all kinds of things that are totally different than the things that are being done in universities and so if you bridge those two domains are you going to come up with something that's like even more interesting and more exciting? Yes, I do enjoy breadth of knowledge. I mean, that's why I read about octopus intelligence and musculature, because who knows when I'll need that. So yes, I do agree that the cross-pollination can lead us to really interesting places.

One of the things that if someone is listening and they want to try out, you mentioned a book, which I want to ask you a little bit more about. But before that, there's Glaucus. And you also mentioned this, it's from Super Releaser. And is it your collaborator who's on the video? Yeah. Um, Matthew Borgotti, he's actually, he started super releaser, um, by himself and then, um, was working on these problems. And at a certain point had asked me like, you know, you know, what next step should I take? And I said, Oh, I think that maybe you need a partner. And I didn't mean myself. And my husband said to me, so, and then you said, and you should pick me. And I said, no,

I didn't say that at all. That hadn't occurred to me. But it was funny. Then I went back to Matt and he said, oh, no, that's what I thought you meant. So I guess I unintentionally put myself on this path. But nothew has been working on soft robotics actually a lot longer than me he's worked on it for going on six years now and he developed glaucus before we started working together and it's a very unique um quadruped super like i said it's a single cast piece of silicone so it has it's a one part uh actuator which is kind of amazing for a quadruped walking robot and people can download the mold for this from thingiverse and then use silicon to build it to do the the mold, and then hook up the little,

what do you call them? Spire? Phenomenon. The little bulbs, the little air bulbs. That one always gets stuck in my craw. Spigmomometers, I think is the right way. The turkey baster thingies. Yes.

And then they can have a little robot that will walk across the table. I mean, assuming this all goes perfectly the first time or the seventh time, however long they spend on it. But it's all, theoretically, someone could go out and get all the parts. Yes, theoretically, someone can make this. In reality, we have had one person recreate the work. She was a university student in Korea, and she actually came to our lab and brought her glaucus with her, which was fantastic. That's so exciting. The thing about the glaucus, though, is that because Matthew is an expert in silicone casting and does that at extremely high level, it is actually very difficult for other people to reproduce.

So what we are working on right now is a book that is for Make Media. We're not sure when it will come out. We're still in the process of writing it and writing projects. But the goal for us in this one was to simplify, to make projects that were much more achievable and repeatable. And so that is very exciting. So for example, I had put out that I wanted to come up with a soft robot that we could cast in a not a custom mold entirely. So you'd only have to 3d print one small part. And then the rest you could use off the shelf parts to make the molds. So you can cast it now in a PVC pipe. And

so that robot I had 16 people make last week. And I think we had 15 successes, and we took two hours to make them. So that's kind of what we're working on for the book to like make this a lot more accessible. in terms of like the silicone cast robots, there's quite a few like small projects that you can find on instructables or kind of Aiden Leach is our intern and he likes to, he goes by XYZ Aiden online. And so he has done a lot of simple work that people reproduce and it's pretty fantastic to see kind of how the community of DIY soft robotics is growing in addition to the academic and the military pieces. Sure. I mean, being able to do it at home and to think about it and then go out and apply it to our careers.

I mean, that's one of the reasons a lot of people are into the maker movement is just that thing you wouldn't even get to try at work until you're good enough. You can say, oh, no, this will work. I'm sure of it. So you mentioned that you don't do a lot of software. You don't do the control system. Does your partner do the control system? He does some control systems, but I think long-term we have found that more of the people we work with have expertise in that area in terms of clients and potential collaborators,

whereas fewer people have kind of this like fabrication plus design base to do the soft component side. So we thus far have been putting more of our eggs in that basket. But we started working with someone who is a postdoc at Yale doing, she's doing cosmology actually, but she also is interested in doing science on physical things. And so like she is one of the people we've been talking about, about like starting to do more robust analysis of kind of the underpinnings of the physical work that we're doing.

So we're definitely interested in collaborating with people who kind of the, either the electronics and coding side is their specialty or people who are interested in the simulation side of the problem. So simulation is something that is in the process of being built up around these nonlinear systems. And so I think that it's really interesting to us to hopefully long term meet people who are working on that side of the problem as well and kind of add that to the set of tools that we're applying to the space oh i bet and it is so non-linear that simulations will improve a great many things and since it is non-trivial to construct the simulation is very good i'm just going to go with that. Stop babbling on. And yet I'm sitting here thinking about the control loops you'd use and what feedback you might get, what sensors you could put in to make sure you weren't damaging. All right, I'm going to go off and not bother you with this, Carrie? Oh, no, I want you to bother me with this as much as possible, but, but we also don't have to,

to, to make the audience privy to that part of the conversation, unless you're interested in that. No, I think we have kept you for long enough for a beautiful weekend day. Christopher, do you have any questions you, before we go?

So many questions. So many questions. No, I don't, I don't think at the risk of starting a whole other discussion i think we should probably i mean yeah i guess the thing is that we we barely scraped the surface of the potential things to talk about but um as long as you guys feel like you got what you needed i, it is fascinating to me because it does seem like a really very early days of something that almost shouldn't be. But it's kind of amazing that there's this whole field that's, it's very untapped, right? I mean, people are just kind of learning the principles and discovering and making the

principles, I guess. I am in the process of reading all of the soft robotics journal. So it's soft robotics is the name of the journal. And it's the academic journal that is entirely devoted to soft robots. And it is to the point where there are only 120 articles. So I can literally read the entire primary source for this work. Obviously, there's some very high-profile work that's going into science and nature and other big-name journals as well, so it won't be the whole field. But even to say, like, okay, I can read 120 articles and have read the bulk of the papers is kind of amazing.

And then you'll write a book so that other people can get involved. I think that's pretty cool. Excellent. Gary, do you have any thoughts you'd like to leave us with? I think the reason why I'm really drawn to soft robotics as a field is that for me, the most exciting space to work and the best problems to solve are the ones that are entirely novel and the ones that can't be solved by a single person. So I just have this passion for thinking across domains and being like, what if we mix this and this? And what if I get to have a conversation with someone who knows everything about something I've never even heard of before?

And how exciting is that space? And so I feel like this kind of play plus rigor is a space that there's not enough time for in some domains and that I feel like the best problems are the problems that require this kind of combination of ideas plus rigor and I just really want to encourage other people to get excited about these in-between spaces. You have certainly made it sound very appealing. I'm pretty excited about soft robotics now. So yes, and I totally agree. This is fun and it's going to be great. I look forward to your book. So as soon as you have it out or have a release date, I will be asking you for information on that so people can get it. Thank you so much for being with us, Keri. Thank you so much for being with us, Keri. Thank you so much for having me. Our guest has been Keri Love, soft robotist collaborator at Super Releaser. There were some other titles in there, but we'll stick with that one. Links to everything will be in the show

notes. You know, the notes that live in your podcast app, but also on the embedded.fm, where there is also a blog, a contact link, assorted other goodies, and a new search bar. Thank you to Christopher for producing and co-hosting. And of course, thank you for listening. I have a thought to leave you with this week from WB Yates. I have spread my dreams under your feet. Tread softly because you tread on my dreams. Embedded is an independently produced radio show that focuses on the many aspects of engineering. It is a production of Logical Elegance, an embedded software consulting company in California. If there are advertisements in the show, we did not put them there and do not receive money from

them. At this time, our sponsors are Logical Elegance and listeners like you.