UCI Podcast: Taking it easy

From the University of California, Irvine, this is the UCI Podcast. I’m Brian Bell. Robots are everywhere in this world. Used to bring speed and precision to manufacturing, they roam around our homes, vacuuming our floors, and there’s even some that navigate the UCI campus, delivering food to hungry students.

Thank you. I hope you enjoy your delivery. Zot Zot Zot!

Magnus Egerstedt began his new role as the Stacey Nicholas Dean of the Henry Samueli School of Engineering in the summer of 2021. Upon his arrival, UCI gained not only a new leader on its campus, but also a renowned expert in robot science and engineering. During his time at the Georgia Institute of Technology, professor Egerstedt began looking into a less-examined corner of the field, that of very slow-moving robots placed in remote locations to perform tasks related to agriculture, environmental monitoring, or biodiversity protection. These robots are designed for long deployments and, for the purposes of self-preservation and energy efficiency, they’re deliberately unhurried while doing their work. Dean Egerstedt has written a book about his experiences with these “slow bots.” Recently published by Princeton University Press, it’s called Robot Ecology: Constraint-Based Design for Long-Duration Autonomy. The Dean joined the UCI podcast to talk about his new book. That discussion is up next.

Dean Egerstedt, welcome to the UCI podcast.

Hello, thanks for having me.

So, we’re here today mainly to talk about your new book which is out now. Published by Princeton University Press, it’s called Robot Ecology Constraint-Based Design for Long Duration Autonomy. This is about robots that are out in the field doing work for us. And you’re talking about developing robot systems that have the ability to stay out in the world for a long time, without any human intervention. Tell us a little bit about the history of your work in this field and how you came to write this book.

So the, the start of this entire journey actually was Costa Rica. I was there <laugh> maybe five, six years ago with my family and I got mildly obsessed with sloths, right? I I couldn’t understand how these potentially quite taste the animals could, could just exist. They’re just sitting there not doing anything, waiting for Jaguars or Eagles to swoop down and, and eat them. And I, I just didn’t understand how nature could support that kind of lifestyle. So I you know, I started reading up about <laugh> slots and other slow animals, and it turns out that there are strategic advantages to being slow birds of prey. For instance, they have something called optical flow sensors. Basically what that means is that they detect motions. If you’re slow enough, you become almost invisible to to predator birds. Uh, and there are all sorts of other reasons why slow is energy efficiency being the, the primary one.

So the, the start of this entire journey actually was Costa Rica. I was there <laugh> maybe five, six years ago with my family and I got mildly obsessed with sloths, right? I I couldn’t understand how these potentially quite taste the animals could, could just exist. They’re just sitting there not doing anything, waiting for Jaguars or Eagles to swoop down and, and eat them. And I, I just didn’t understand how nature could support that kind of lifestyle. So I you know, I started reading up about <laugh> slots and other slow animals, and it turns out that there are strategic advantages to being slow birds of prey. For instance, they have something called optical flow sensors. Basically what that means is that they detect motions. If you’re slow enough, you become almost invisible to to predator birds. Uh, and there are all sorts of other reasons why slow is energy efficiency being the, the primary one.

And I’ve spent 20 years building robots and I tried to build faster and more aggressive and agile robots. And I I got a little intrigued by this idea of what if I take a step back and build really slow, hyper energy efficient robots that would kind of change the, the way we think about the design and the way we think about what the robots should be doing. So that’s kind of how it started. I just got fascinated by, by slowness as a design paradigm.

So it’s very different from those robots that are doing parkour and, obstacle courses.

Yeah, this is <laugh> exactly. This is as far from a back-flipping robot that you could possibly get. But once you’ve decided that you’re gonna build slow robots, the next question is why beyond the fact that slow is kind of interesting.

That’s the point. And I had been working on precision agriculture for a while where I deployed robots out on the farm fields of Illinois. I had autonomous tractors dealing with, with corn, for instance. And I started thinking about what if you have robots almost living out on the farm fields for an entire growing cycle. I mean, basically they have to move very slowly, almost at the speed of a plant, if you will. There, you know, being slow is actually kind of, it makes sense. And in a lot of different environmental monitoring applications, you know, you should just be present in the environment for, for a long time. So this is kind of where it started. I, I knew that I was excited about slowness and I liked this idea of its being out in a natural world for a really, really long period of time for the purpose of kind of environmental monitoring and checking what’s going on out there.

So those were the kind of two pieces, and I started talking about this project and I used a lot of terms like robots living in the environment or robots surviving for long periods of time. And I realized that I was using words that smelled more like ecology than robotics. So that was kind of the genesis of the idea of robot ecology with this idea of an organism in its habitat. So a robot it’s in its environment and the way we do robotics normally, and Boston dynamics back-flipping robot is a beautiful example of this is we go to a lab, we build a robot, we have it work in the lab, and then we take it out and hope that it’s gonna work wherever we deploy it. But what if we actually thought of the design of the robot as being intimately linked to the environment in which it’s living, then all of a sudden you think of a system, an ecosystem where the robot is just yet another participant in that ecosystem.

Do you want this thing to blend in and be kind of camouflaged in this ecosystem or is that impossible to do?

What I don’t want is for the robot to become an invasive species. I don’t want it to be harmful to the environment camouflage isn’t that big of a deal. But one thing that we did is we built a version of a slow robot that we call the SlothBot. So this is a super slow cable-driven robot that’s hanging up under the tree canopies and every now and then it goes out and sun bathes and recharges the batteries. And then it goes back in under the tree tops. And we deployed this robot up in a botanical garden on the East Coast, which is where I spent the previous part of my career. And it was really cool to see squirrels actually sitting on top of the robot. In the beginning, there was a Hawk that was nesting nearby that kept checking out the robot to make sure that all was good. And after a while, the hawk didn’t seem to care anymore. And the squirrels kept using the robot as a place to just hang out. To me that was success, that it was accepted as a member of that particular local ecosystem.

And what are these robots out there monitoring? I read in the book, certain climate biodiversity, what are some of the other things that these robots are keeping track of out there?

So ultimately I am a tool maker. I make hammers. And then I give my hammers to people that know how to build things. So in this case, people that are studying conservation biology or field ecologists, and I tend to think of what we do as providing a mobile platform that you can use to monitoring monitor for a lot of different things. What we have focused on are things that are called microclimate data, which is fancy speak for things like pressure, humidity, illumination, carbon dioxide levels. So things that we can measure that have to do with the local climate, and then we have cameras. So we’re interested in, in try certain species. By teaming with a group of biologists, we started looking for particular orchids for instance, and we were gearing up to send the robot down to Ecuador to actually figure out who is pollinating, a particular orchid that we don’t know who is pollinating it. And then COVID happened.

Is that through photography or how does it…

So it’s cameras. Yeah. It’s video.

What are the other considerations? I think one of the things you go through in the book is this sort of pairing of both energy conservation, and they have to go to a certain place to get energy from the sun. Yeah. And that might impact their mission. So how do you square that?

Once you have embraced this idea that the robots should survive and live for long periods of time outside, then that change the design and the book is really about how, how does that change in perspective, influence how we think about robotics? So the normal way we do robotics is we have a task. We encode that task through a cost, and then we minimize the cost. So we’re acting optimally, let’s say you’re a self-driving car. And you want to get from, you want to as quickly as possible stop at a stop sign. So the, the way to do that, if you’re minimizing time is you slam the gas pedal, pedal to the metal. And then at the last second you slam the brakes, right? So that’s, you’re minimizing time. Now, if you get something wrong, then you come to a complete stop in the middle of an intersection and that’s potentially catastrophic, right?

And if you are a robot out somewhere, again for long periods of time, or you’re a robot on Mars, right? Avoiding catastrophic events is much more important than being optimal. So acting optimal is almost completely irrelevant to this idea of surviving. And the big part of surviving is you can’t get stranded somewhere with completely depleted batteries, with no ability to recharge. Let’s say you’re a solar powered robot, and you’re in a shaded part of the world. If you’re, let’s say, under a tree canopy. And if you’re out of battery under this tree canopy, then it’s game over. So you always have to take energy management into account, and that’s almost the first order of business. So rather than making sure that you get to always get to the right spot for the purpose of detecting orchids, say, rather than that, you have to always ask, do you have enough energy to keep going? Cuz if you don’t, then that’s the end of the mission. So it really changes this idea of, of what is important. And again, borrowing from ecology, this idea that survival that’s, that’s what it’s ultimately about rather than some kind of optimality notions

In part three of the book, you talk about the difference between survival and survivability. Can you explain that, that difference?

Well, I mean, you know, I’m a scientist, so I need to have a fancy term for something, but, but survivability is really the ability to, to survive. So always act in such a way that you never end up in a situation where you have no chance of actually surviving. So, let’s say again, a car example, right? So you may not have crashed, but you may reach a point from which it is impossible to avoid a crash, right? So it’s not a matter of, of not crashing. It’s a matter of always staying away from areas where there is no guaranteed recovery. So that’s the idea of survivability that you never end up in a situation from, which there is no path forward to avoid a mission ending, for catastrophe.

This book covers a lot of the work that you did while at Georgia tech. Do you intend to do some work going forward in this area now that you’re here at UCI?

Absolutely. One of the things that I’m really excited about UCI is really that climate research, environmental research is front and center and we’re really world leaders in a lot of these areas. So I am very excited about this idea of having the SlothBot migrate to the West Coast. And in fact I have negotiated with the Crystal Cove Conservancy that we’re gonna put the SlothBot down on the beach. So it’s gonna be a beach-dwelling, sloth robot. And it’s gonna be measuring things like the beach dynamics and see what’s happening to our local beach ecosystems, because those are strong indicators of things happening with the climate.

It should blend right in. There’s a lot of slow-moving organisms on the beach in Orange County.

Exactly right.

So, who is the audience for this book? Who do you intend to be interested in reading this book?

I actually wrote the book almost like two books, so you can read the first part of it without knowing a lot of robotics. And I kind of think of the first part of the book as, as an introduction to this general idea of robots being part of natural ecosystems. And how do you think about organisms in this particular way? So I wrote it almost like a popular science book. So I’m hoping that this is for high school’s students, college students, people that are interested in what’s the state of the art in robotics and how does it connect to ecology. And then the second half of the book is, is more technical and there, you probably need to be a, a graduate student in robotics or an adjacent field to a appreciate what’s what’s going on. But I did it on purpose so that you could read the first half and get something out of it. And if you want to keep going, you could, but you don’t have to.

So you came to UCI in the summer of 2021. Can you tell us a little bit about your experience of coming on board here. How do you like it here in Irvine?

So I’m really digging it. It’s been, it’s been almost nothing but positive. Let me tell you, first of all, why I was excited about UCI in, in the first place, I had reached a, a point in my career where I felt strongly that engineering was such a big part of all the defining questions of our time. You know, how do we feed the growing planet? How do we make sure we leave an environment for our kids that’s not burned to the ground? How do we, how do we structure our future societies with AI and robotics so that we don’t become as in the matrix, right? So these big mega questions, engineering is such a big part of it, but engineering can’t do it alone, gotta connect with other disciplines. So I was looking for a place that was truly multidisciplinary and where there just seemed to be a collaborative vibe built into the, the campus fabric.

I was really impressed with UCI and what I’d heard about it. And my conversations before I accepted the position with other deans and people just seeming to genuinely like working together and also a, a hunger on campus to really be part of making this planet better for everyone. So I came in with that kind of expectation that this was a place where one could do big things in a collaborative way. And even though I’ve only been here since, since July, I don’t think I was wrong in my early assessment. This is a place where genuinely people like each other and where there is a kind of scrappy swagger in terms of just going after projects and problems that matter. And we’re not done when we’ve written the paper and eight of our friends read the paper. No, that’s not it. We also want to see our intellectual products, make it into the communities around and that’s fun.

Well, Dean Egerstedt, I really want, thank you for joining us on the UCI Podcast today, and congratulations on the book. And I hope it becomes a New York Times best-seller.

I hope so, too. And thanks for having me.

Robot Ecology: Constraint-Based Design for Long-Duration Autonomy is available for purchase now. Who knows, it may even be pulled off a warehouse shelf and shipped to you by a robot. You can learn about Dean Egerstedt’s work and other robotics research at UCI by visiting engineering.uci.edu.

The UCI Podcast is a production of Strategic Communications and Public Affairs at the University of California, Irvine. I’m Brian Bell. Thank you for listening.