Research Matters

Robert Shepherd on the future of robotics - Research Matters S2E8

• Cornell University • Season 2 • Episode 8

Share episode

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 26:28

In this episode of Research Matters, Cornell University mechanical engineering professor Robert Shepherd explores a radically reimagined future of robotics – one where sustainable, self-repairing machines partner with biology rather than imitate it, opening new possibilities in medicine, environmental monitoring, agriculture and beyond. Shepherd explains how biohybrid robots use living cells as actuators, how mycelium can function as both building material and sensing network, and how volumetric 3D printing allows entire soft robots, complete with internal skeletons, to be created in a single step. Watch here.

Robert Shepherd:

We actually ordered a king oyster mushroom kit from Amazon and started to...

Laura Reiley:

And there are all kinds of problems with the king oyster mushrooms. They've escaped and are now kind of taking over in the wild.

Robert Shepherd:

Oh, I did not know that...

Laura Reiley:

and out competing, yes.

Robert Shepherd:

Oh, really. Oh, OK.

Laura Reiley:

So you may be part of this problem.

Robert Shepherd:

I'm sure that happened before we did that. We follow all appropriate protocols.

Laura Reiley:

Welcome back to Research Matters, where we explore how Cornell researchers are solving real-world problems in surprising and inspiring ways. I'm your host, Laura Reiley. Today we're diving into robotics, not the kind made by made of bolts and steel. My guest is Professor Robert Shepherd, director of graduate studies for mechanical engineering and the John F. Carr Professor of Mechanical Engineering. He designs robots made of living tissues, fungal networks, squishy materials and even structures printed all at once in 3D space. He's showing us what happens when machines stop trying to imitate life and actually start partnering with it. Rob, thanks so much for being here.

Robert Shepherd:

Thanks for having me.

Laura Reiley:

All right, so I want to start with the class that you teach. We just saw what was in his book bag, and it was an awful lot of let's just say it was a catholic, small c, array of reading material. But what is this, this class that you teach, that you just finished?

Robert Shepherd:

It's a mechanical engineering class, co-instructed with systems engineering, and it's the framework for how to invent new products that people want. And the books I showed you were more about the influences I had and draw from and invention in my research, but also that I try to apply to my course.

Laura Reiley:

So is it an entrepreneurship class, or is it more kind of an esoteric theoretical the origin of invention?

Robert Shepherd:

It's more theoretical with a with a framework and tools you can actually use. But I wouldn't say it's an entrepreneurship class. We do delve deeply into patents and how and look at how invention progresses over time. Before this started, you were mentioning aluminum cans, and one of the lectures is about how aluminum cans proliferated and why they didn't exist in mass 100 years ago, and why they're used a lot today, as an example.

Laura Reiley:

Oh, that's fascinating. Yeah. Oh, that's great. So do you, for the students in this class, is there an invention component to it? Do they have to invent something, or at least, kind of contemplate what they would invent?

Robert Shepherd:

Just a small task of having to invent something

Laura Reiley:

Easy.

Robert Shepherd:

So my co instructor, Sirietta Simoncini, from systems engineering, she takes the students through an empathy field work process where they try to identify sometimes outlandish behavior and figure out why it exists, how, if it's useful, and then how we can adopt it and make it something that's serviceable to more people.

Laura Reiley:

Fascinating. All right, so I have read a bunch of stories in the Cornell Chronicle about your work, and to me, they always seem kind of phantasmagorical. You know, I read the recent one on kind of fungal networks and robotics. I think there was one. I'm not sure if it was in the Cornell Chronicle, but it was about robots that have little explosions that advance their their movement. So where should we start with the work that, maybe we'll start it more recent work, and then work our way backwards?

Robert Shepherd:

OK, those sound like very different things, but they were arrived at using the same process. Our group, we try to think of future needs for robotics and then find the fundamental principles required to enable it, whether it sounds strange or not. If it sounds strange, it's even better, because there'll be less people working on it. And so the explosions came first. That came from wanting to have something high power, but small. Batteries...

Laura Reiley:

How small are we talking? What's the...

Robert Shepherd:

Three-millimeter diameter disks that can actuate to similar three-millimeter height heights, but using 10 Newtons, like a like 10 apples, kind of forces.

Laura Reiley:

OK...

Robert Shepherd:

So a lot of power, small area batteries don't work well for that. The smaller you make a battery it's mostly packaging and not much energy or power delivery. So at that point, we thought, what other power sources can we use? We're not going to use a nuclear power source or little black holes. So chemical fuel.

Laura Reiley:

Solar seems bad. Wind also very bad in this case.

Robert Shepherd:

Yeah, it doesn't scale well very small. But, of course, at large scale, those are great solutions. But chemical fuels have a lot of energy in a small volume. So that's, that's where we went through there. And then we made little insect-scale robots that could do very high-power motions. And then aggregates of these actuators to make Braille displays that are full scale Braille displays that don't require being plugged in and solves a big problem in the in the blind community.

Laura Reiley:

OK, wait, go back to that. So the Braille, so Braille. So you're saying these are tactile, these are things that a visually impaired person can feel. How does it change the Braille experience?

Robert Shepherd:

Well, it's very difficult for a - and I should say the first author on this paper is Ronald Heisser. Fantastic student of mine is now at MIT. He wanted to, he wanted to demonstrate this capability in Braille. And so he did an i-core experience with that was enabled by Cornell as well. So we got to talk... He went to a conference to the blind. He met Stevie Wonder all these kinds of things.

Laura Reiley:

Wow, that right there is worth the price of admission, for sure.

Robert Shepherd:

He loved that part. I would have too. And the thing is, the blind have a have a difficult time with devices that allow them to read music or do math, things that aren't line by line kind of things. And so a full...

Laura Reiley:

Because it's six dots, right? Six dots in different...

Robert Shepherd:

Yes, yes, that's right. Well, you know a lot, and

Laura Reiley:

That exhausts my Braille knowledge...

Robert Shepherd:

You used phantasmagoric earlier. I'm still gonna look that up. But so the the but this braille display issue is one that's a problem when you need, when we want something portable. Getting the the number of dots in a the way Braille displays work now would create a very huge, heavy thing and where most of it would just be, you know, a little bit of area you can read. So this solution allowed us to make forceful, tactile inputs in small areas without creating a huge, bulky, expensive, heavy display.

Laura Reiley:

Yeah, I can totally see the utility of that. So something I've read in your work is the word biohybrid robot appears a lot like, what is that? What does that mean?

Robert Shepherd:

That's where you use living tissue to augment your robot to provide new capabilities. And so biology is really good at self assembly, like it starts with amino acids and creates proteins, and proteins assemble into other functions that we see and express at macro scale into if you're a human or mammal or something like that. And we just can't do that artificially. We can't start...

Laura Reiley:

And self repair too, right? I mean, isn't that like a thing that living tissue is good at?

Robert Shepherd:

Yes, that's right, yeah. And we, you know, there's a lot of work. We've got work on self-healing chemistry, which is synthetic too, but yeah, living tissues will heal themselves. Mammal tissue is really good at high-force density. It can, like, create a lot of force in a small volume, and it can, you can grow it in, like, weird architectures and things like that. But mammal tissue is sensitive and it can... It's a very arduous process to culture it and and get it to work appropriately. Actually, funnily enough, Ronald is also working on mammalian tissue biohybrids in his postdoc. But what I wanted to do is work on biohybrid systems without the difficulty of mammalian culture. So we started... We we actually ordered a king oyster mushroom kit from Amazon, and started to...

Laura Reiley:

And there are all kinds of problems with the king oyster mushrooms. They've escaped and are now kind of taking over in the wild.

Robert Shepherd:

Oh, I did not know that,

Laura Reiley:

And out competing, yes.

Robert Shepherd:

Oh, really. Oh, OK.

Laura Reiley:

So you may be part of this problem.

Robert Shepherd:

I'm sure that happened before we did that follow all appropriate protocols.

Laura Reiley:

So anyway, all right, so the oyster mushroom kits, which are common now on Amazon.

Robert Shepherd:

So we ordered it. And the problem with culturing mushrooms is not them dying, it's making sure other mushrooms other fungus don't grow in the same petri dish. So, but that's the only issue we had. And once they're grown, they're stable for months, unlike mammal tissue. If you brought it outside and just put it in contact with the air, it would die almost immediately worse. Anyway, so that means that allows us to put the mushroom mycelium, which is the sort of you can think of it as a root network...

Laura Reiley:

Underground, so mushrooms up here, and mycelium is under.

Robert Shepherd:

Right, yes.

Laura Reiley:

So you're working with the mycelial part of the oyster mushroom. Why is that a good because of the product for your your work?

Robert Shepherd:

Because the fruiting body, which is what most, most people think of as the mushroom, is temporary, like it grows and spores and dies and goes away. The mycelium is persistent. So we wanted to keep that in contact with the... We use electrodes to interpret the signal and control a robot, and so you can make sheets of it. Right now, I'm working with Larissa Shepherd in Human Ecology, who's spinning fibrous mats, and we're culturing these spores into like sheets of mycelia that you can wrap around robots for wide area perception. And I think thats...

Laura Reiley:

So I know these sheets are being used for packaging and a whole bunch of other things right now, and even for, you know, fake bacon and, you know, all that kind of thing. But so these, these mycelial sheets, basically, do they live in perpetuity or or like, what's the what's the lifespan of one of these kind of things?

Robert Shepherd:

Yeah, a big difference between - which the work Ecovative does, which you were mentioning, is really great. The difference is theirs is not alive, and ours is alive in how we're using it. And so it lives for months. Then it actually does start to die as it doesn't can't grow anymore in the volume it's in, it starts to die. And then we notice the signals decaying in our robots. And it's a very interesting problem to solve, to how to deal with controlling a robot with degrading signals, and then even thinking about how to rejuvenate it afterwards. It's a whole different problem space. And by working on mycelia instead of mammalian tissue, we can start to address it now in real systems, rather than wait another 20-30, years for these mammalian based, bio hybrid robots to be stable in the wild.

Laura Reiley:

OK, so how do you convey what you want the mycelium to signal to the robot? Llike, how do you how do you program the mushroom?

Robert Shepherd:

That's a really great question. We aren't programming the mushroom. We're listening to what it's saying and telling the robot to do what we want based on the information the mycelium perceives. And I intentionally use the word perceive instead of sense, because the mycelium is taking environmental inputs, multiple inputs, and encoding it into voltages that we're reading. It's making sense of the environment, and we are taking that signal and trying to do something with it. An example is- so mycelium mushroom, the fungal kingdom is - and if a mycologist is hearing this, I'm sure they're gritting their teeth. But as an engineer, the fungal species kingdom is similar to mammalian, mammal.

Laura Reiley:

It's the largest organism, right?

Robert Shepherd:

It's a large organism. They, we use antibiotics from them, because they don't like bacteria either, and, you know, obviously the bad kind. And so, if a to be simple, a mushroom doesn't like the environment it's in, we may not like it either. And so if we can understand the action potential, the voltages and electrical signals coming from these mycelia as good or bad, then our robots can make a decision about searching the space for something that's maybe harmful to us so it would be going against what the mycelium favors.

Laura Reiley:

Favors as in the way a plant leans towards the sun, or the way I don't know, muscle will kind of avoid, I mean, it doesn't have a brain, per se, right, but it'll avoid adverse perception, or condition.

Robert Shepherd:

Yeah, so heliotropism, like plants trying to follow the sun, and maybe a more complex reflex action, where, you know, we blink when something goes in our eye. It's we're more thinking of the mycelium as helping us understand if the environment is bad or or good. And in the case of, let's say, a nuclear reactor meltdown, we could send a fleet of these biohybrid robots into it, and then they could quickly assess the largest areas, the concentration of radiation, for example.

Laura Reiley:

Sure, yeah. Well, I was kind of thinking about, kind of, what's the all of this is so cool, but what are the practical applications for some of this? I mean, I can imagine kind of tissue self-healing tissues being advantageous in a robot, if it were in space or deep under water or in some place where fixing it for humans would be difficult, like, what are the other the kinds of questions that you're asking right now? What problems are you trying to solve?

Robert Shepherd:

Well, you kind of nailed it there. The robots we make are, we don't shy away from complexity, which is one of the things that differentiates us. Our group, I think that in order to have robots that are autonomous, agile and capable of operating untethered for long periods of time have to be complicated, an integrated network of multiple systems, some amount of redundancy, and things like that. And so these are things for unmanned missions to Mars, things like that. And if something breaks, it does need to heal. And so we approach that from artificially but also with living tissues like you're describing now, too.

Laura Reiley:

Wow, that is wonderful. I know that you've done a lot of work in kind of, you know, volumetric, 3D printing as well, and how does that fit in with your robotics work?

Robert Shepherd:

Well, as I said, it's a good, that's a good connection, because we want... The robots we make are complicated. There's a lot of multifunctional, integrated components, kind of like organ systems. To assemble that is hard to do. The reason we're using living tissue is because they build from the nano scale up to the macro scale. We can't do that, but we can kind of get there from printing, where you start from the bottom and grow out and so the and get complexity at the same time, and it's the same amount of time to print a brick as it is to print a replica of La Sagrada Familia, or something like that. So...

Laura Reiley:

Although, I'd like to see the latter, that'd be great.

Robert Shepherd:

It's wonderful, the best thing I've ever seen in my life. So then we can get complex things and co-integrated systems at timescales of a graduate student.

Laura Reiley:

Yeah, actually, Gaudi is like a really good 3D printing kind of environment.

Robert Shepherd:

Yeah, absolutely. I think about it all the time.

Laura Reiley:

Yeah, that could be wonderful. You could redo all of the wonderful places in Barcelona. All right, so in terms of the 3D printing, what specific things are you working on? What are the specific kind of thorny questions that you're trying to overcome?

Robert Shepherd:

The most recent challenge in what's, what's called volumetric additive manufacturing. It's not a layer by layer process. Everything solidifies all at once, like a Star Trek replicator, and in that case, the the issue is it's small, like we're talking cubic centimeter volumes. The problem that the community had primarily was thinking about absorption of light. It's a photo photo polymerization process where light has to penetrate through the volume. But actually we discovered that a bigger challenge initially is that when you do the chemistry, it creates heat, and then that heat moves the liquid around that you're trying to solidify all at once and makes it impossible to make the thing. So we invented an active cooling technique that allows you to print large things in the same amount of time.

Laura Reiley:

So the cooling is this about kind of structural integrity, or what are the what does the cooling allow you to do?

Robert Shepherd:

Well, when you... we actually, in this case, we don't want anything to move when it's being printed. We want the liquid or solid to allow light to pass through it, but everything stays in the same place. When you create heat, if it's a liquid, you cause motion, convective flow that then makes it so you can't solidify the object where you want it to. So by cooling it, we remove the heat, so we don't get these flows, and then we can make the object solid there.

Laura Reiley:

So this process, as opposed to the layering, what bottlenecks or what problems does that potentially solve for industry, or for, you know, scaling robotics?

Robert Shepherd:

We think a lot. In fact, we started a company a year ago trying to commercialize this. One is because the liquid, the material doesn't have to move you have a lot more options of materials. So you can use more viscous materials.

Laura Reiley:

Like what would be? What would be one of these viscous materials?

Robert Shepherd:

Well, polyurethane that has an oligomer, like a precursor that's longer. And so it would when it's longer, these chains are harder to move around. They're thicker, like oil in your car, like but even thicker than that, even more viscous than that, so and then you have to somehow transport it in most other material system, but in ours, you just it just sits there. It doesn't matter how viscous it is. It could even be a solid. And you could do that. And with that, you can have tougher rubbers, for example, or tougher plastics, these kinds of things. So if you were to try to print a shoe this way, you could print it out of materials that are even better than what shoes are made out of now, whereas other kinds of printing technology, you could print a shoe, but it's going to be worse materials than now and also take a longer time. Or the feet of robots.

Laura Reiley:

The feet of robots, yeah, the shoes of robots?

Robert Shepherd:

They need shoes, that's great. We got it.

Laura Reiley:

All right, so I know that some of the work that you've done has had direct applications for agriculture, and I've done in my life, I've done a lot of stories on the adoption of robots, like the slow and maybe overly optimistic adoption of robotics in the field, because obviously labor for agriculture is the big headache, you know, for all kinds of reasons. So what kinds of things would your array of robotics allow to happen in the field?

Robert Shepherd:

Great question, I think, because it's the our main so we want to make robots that are agile. They can perform dexterous movements, manipulate objects that are in an intricate way, but also do that for a long period of time. A lot of people in the robotics community are doing great work solving the dexterity agility problem. Tons of money going into robot hands, humanoid hands, things like that.

Laura Reiley:

That don't like ruin the fruit, but pick it at maximum ripeness.

Robert Shepherd:

Yeah. But if you try to do that for long periods, if you try to deploy a humanoid robot in the field to pick fruit, it's going to be able to do it for like, 45 minutes.

Laura Reiley:

And then what happens?

Robert Shepherd:

Then it has to go recharge.

Laura Reiley:

So that's, that's worse than a lot of you know, F2A visa workers, right? Yeah, that's, yeah.

Robert Shepherd:

Yeah. So we need to, so we're, we're imagining this challenge is going to be solved. Robots are going to be agile, and all this other stuff. But after that, then what? Efficiency matters. Energy storage matters. How long can they do this for? And so that's those are things we're trying to solve by making multifunctional use of energy, where parts of the machine not only are the frame or the actuators, but they're also the energy storage. So 90% of the weight is also a battery. It's a lot like how humans work, especially me with...

Laura Reiley:

Yeah, I hadn't thought about that as as is even harder than those kind of, you know, the the grabber making the grabbers better?

Robert Shepherd:

Yes, yeah. I don't know if it's harder. It's just, it's something that isn't being addressed right now, and we want it to have that solution ready for when the rest of the problems are solved.

Laura Reiley:

So are there agriculturally, are there crops, or are there situations where this would be especially useful? I mean, I imagine corn and soy is already pretty automated, so are we talking about specialty crops or tree fruit or what like? What are the things that this that in the next five years will pivot to robotic? You know, planting, harvesting...

Robert Shepherd:

You're right. The current row crops are designed for the equipment we have now. There's a lot of nutrient-rich produce that isn't appropriate for that, that isn't being farmed at scale. There's also the opportunity to mix crops together, and not in rows and ways that have some that allow reduced pesticide use and better use of nutrients.

Laura Reiley:

Kind of more biodynamic planting strategies or...

Robert Shepherd:

Some that are resistant to certain insects that shield the other ones and... But that's a very complicated process, but if you have robotics that can that are adaptive and can work in these in these systems, then you can maybe do that, and then we'd have more nutrient-dense food with less agricultural waste.

Laura Reiley:

So is that predicated on kind of use of GPS coupled with AI like, I mean, I imagine you have to have a robot that can see, can identify, oh, this thing will be ripe tomorrow that I've just passed over. I have to remember precisely where it is, right?

Robert Shepherd:

That's certainly a lot of the story. But there's also, like, my favorite fruit is the pawpaw.

Laura Reiley:

You're the only one.

Robert Shepherd:

You've never had it?

Laura Reiley:

I've had had a pawpaw.

Robert Shepherd:

Oh, they're delicious.

Laura Reiley:

I think Florida... they grow in Florida sometimes.

Robert Shepherd:

They grow here.

Laura Reiley:

Do they really?

Robert Shepherd:

They grow here, in Michigan.

Laura Reiley:

Is there a pawpaw festival or something?

Robert Shepherd:

I don't know, but I know that actually, I'm not gonna give away my hidden grove.

Laura Reiley:

Okay, you don't have to do anything...

Robert Shepherd:

Actually, you know what I will at the macadam nut farm or nut Grove, there's a bunch of pawpaw trees, and anyway, you can go there and pick them. They're only ripe for a short period of time. They taste like mango, banana custard, but it's...

Laura Reiley:

But you never see them in stores, and that's because they're so ephemeral, right?

Robert Shepherd:

Exactly. Yes, that's right. They don't transport well, they don't have a long ripening period, and you can't...

Laura Reiley:

Where you didn't think this conversation was going today, the pawpaw.

Robert Shepherd:

I'm very passionate about these things, and you can't see that they're ripe. You have to feel that they're ripe. So vision and GPS is only part of the story. Being able to feel it and get through occlusions like leaves that are blocking things and stuff like that. It's also a big part of the story.

Laura Reiley:

Wow, we're almost out of time. But one more, one more kind of agricultural question, I imagine robots, kind of a phalanx of guardian robots could ward off pests, or large scale, you know, blights. Or is that another arena that this could be brought to bear on?

Robert Shepherd:

I've seen some efforts in weeding robots and probably insect hunting robots, but I think there are... I prefer the mixed breed plant, where we work on we integrate plants that are resistant to some of these things, or fight them in their own way. Like plants, they practice chemical warfare like nobody's business. So I think there's probably a way to use plants to do it, but maybe robots could maybe do that. It'd be a very energy intense way to do it.

Laura Reiley:

All right, so if we're I, all of this is so fascinating to me that we could go on and on, but I do want to ask, like, is there any kind of what's the next big question on your horizon, or the next big problem you're you're aiming to solve?

Robert Shepherd:

Well, it is the use of the biohybrid mycelium based robots for threat detection, radiological, chemical, that thing, and how we can coordinate swarms of robots to localize the origins of these these threats.

Laura Reiley:

Well, Rob, thank you so much for coming in today and for showing us how robotics is reshaping what machines can do and how they can help us in fields from farming to medicine to space exploration. If you want to learn more about Rob Shepherd's lab, look up the Organic Robotics Lab at Cornell Engineering, and if you enjoyed this conversation, remember that Research Matters brings you stories just like this, research that doesn't stay in the lab but reaches into the real world to solve real-world problems. To watch more episodes or to learn about the breakthroughs happening across Cornell, visit our website or subscribe wherever you get your podcasts. I'm Laura Reiley, thanks for joining us.