E11: Climate Robotics | Berk Calli, associate professor, Robotics Engineering

Jon Cain: When you think of robots, what comes to mind? Are you picturing a machine on an assembly line putting parts together? How about one moving through a warehouse, pulling objects from a shelf, or maybe a humanoid robot that resembles a person and can do basic tasks or interact with people. My next guest is here to tell us that there may be more to robotics than you think. They have great potential to address climate challenges. Today we're talking about climate robotics. Hi, I'm Jon Cain from the Marketing Communications Division at WPI. It's time for another edition of the WPI Podcast. This podcast brings you news and expertise from our classrooms and labs. I'm here at the WPI Global Lab in the Innovation Studio on campus. I'm happy to be joined by today's guest. Berk Calli is an associate professor in the robotics engineering department at WPI. Berk, thanks for being part of the WPI Podcast. 

Berk Calli: Great being here, Jon. 

Cain: Well, Berk global temperatures are rising. Storms are becoming more frequent and severe. Because of that, people need to better predict and measure the impacts and make changes to protect themselves in their communities. How do robots and the emerging field of climate robotics fit in?

Calli: So in many different ways. For example, they can be used for environmental monitoring. They can be used for prevention efforts like carbon capturing, also intervening to climate disasters mitigating the effect of these, disasters after they happen, or, uh, they can be used to make our current infrastructure more resilient.

Cain: Yeah. It sounds like there's a lot of great potentials there. I'm wondering if you might want to expand a little bit on some of the specific tasks that robots could help with that could be helpful to address the climate challenge.

Calli: Sure. So for example, for environmental monitoring robots are used to inspect the current situation of coral reefs or they are used to take some measurements from, uh, deep under the sea. These can be many different measurements related to the impact of the climate change. Also, uh, sometimes again in environmental monitoring purposes, they are used to get some measurements from lakes and rivers to understand, how the ecosystem of, uh, those areas are impacted by climate change. When it comes to, intervention effects, for example, robots can be used to intervene a disaster. It can be an oil spill, for example, as it is happening. Or it can be a forest fire. So how robots can be actively used over there to basically contain the effects of a particular disaster. And mitigation could then be, for example, after a forest fire or after an oil spill again, it can be about cleaning the environment or bringing the ecosystem back up. And for resilience effects, it can be more of like inspection of our infrastructure or making building stock, more resilient, for example, identifying certain deficiencies in our current building stock and intervening them on the spot. 

Cain: It's really exciting to hear the different types of opportunities. In your lab at WPI, I'm wondering what are the types of climate robots that you're working on?

Calli: So we are working on a bunch of projects. Our lab is manipulation and environmental robotics lab. So, we have an explicit focus on environmental, applications here. One of the areas that we focus on is recycling we are mostly focusing on municipal waste recycling. So this is the waste that is produced in our homes offices. These are usually, transferred to waste sorting facilities. And from there, they need to be sorted to paper, glass, plastic, and metal because we oftentimes, pretty much everywhere in the United States, it's like single stream recycling. So they need to be sorted to these subcategories before they get recycled. And we are investigating the role of robots on achieving that. So that's one of the projects that we are focusing on. Another one is on shipbreaking. This is metal scrap recovery. Vessels usually get decommissioned in around like 20 to 30 years because they are so extensively used. And in order to recover the valuable metals that they are built off they, they need to be broken down into smaller pieces. And this is actually a very laborious process and it needs to be automated in order to scale, those efforts. So we are working on prototyping the first shipbreaking, robot actually. So we did some research on that. We are now in the prototyping stage. And there are some other projects that we are also looking into, like, for example, with our colleagues in soft robotics lab, Cagdas Onal in particular, we are looking into how we can design and control robots that can go into building envelopes and investigate, whether there are certain inefficiencies in these existing building stock so that we can plan how to make them more resilient

Cain: That's fantastic. I was in your lab in the spring and, and your students gave me a demonstration of the first robotic system that you talked about to manage recycling. Wondering if you could sort of just walk our listeners through what happens and how the robot works. You've got a bunch of items that are entering the recycling stream. You’ve put 'em on a conveyor belt. What does the robot do with those items when it jumps into action? 

Calli: So, using robots for recycling requires us to put many interesting robotics components together. One of them is this machine learning model that identifies the type of the waste just by their appearance. So, by just looking at the color image of a bunch of objects, the algorithm is able to identify, whether they're plastic, paper, metal, or glass. That is the recognition side of the problem. And there is the picking side, the manipulation side, and over there, the robot needs to identify how it should place its gripper or its suction cup on the object so that it would be able to securely pick it and then place it to the corresponding bin based on its type.

Cain: And so what's one of the challenges that that can help address is its sort of the, the success rate of that?

Calli: So recycling is a very challenging problem because the recycling facility is required to sort these items in a very high accuracy, this accuracy can go up to 95, 98, 99%. And considering how cluttered, this waste stream is and how the objects are on top of each other, it really is a big challenge. So these sorting facilities have a lot of machinery that makes this happen. But at the same time, there are also manual sorting stations that workers are sorting, these waste items because they cannot do it with the machinery alone. And even then, so, human effort combined with all this machinery, it's still very hard to meet those accuracy thresholds. So, robots become very logical option to be integrated to the process. However, there are main specific challenges here. For example, these waste items, the new waste items come and others go away as new products appear, in the market. So the robots need to be trained by these new items all the time. So the machine learning model that I was talking about, it needs to be updated, uh, constantly and it's also very difficult to obtain these machine learning models that is general, that would work everywhere in United States or in the world. It is quite beneficial to focus these models to particular areas and if possible train them for specific facilities. So what we are working on is this technology that as workers are sorting items it identifies what they are picking automatically and trains the algorithms, trains these machine learning algorithms automatically. And by that way, we are not putting any extra effort to the worker side. But we are collecting very valuable information because if you don't have this kind of technology to retrain the robots, you need to actually manually label each and every object by clicking the screen. You need to mark these objects pixel by pixel and then say, all these pixels belong to this object, and the material of that object is paper, for example. And this is such a laborious process. And this kind of framework actually is a human-robot collaboration framework. One thing that I would like to say here is that robots here will always need to be trained, as I said, right? So humans, the role of humans here will always be there, in order to enable this training process. It is not a process that the role of human will be phased out, but it is a truly human-robot collaboration and enhances both, the capacity of these plants as well as the accuracy of these plants. 

Cain: It's definitely a great challenge as you describe it. I'm not gonna look at my recycling bin quite the same way when I put it at the, at the curbside. There's always some tension and concern when you bring up the topic of robots and how it might impact jobs and humans. With that landscape in mind, I'm wondering why do you think robots are needed to address the climate challenge? Is this a challenge that humans are not able to tackle on their own?

Calli:  This is a great question, because producing a robot by itself has its carbon footprint, right? So we really need to be very careful while providing or developing robotic solutions to environmental problems. But the roles of robots can be really important because considering the urgency of climate change we need to scale our efforts up. Robots are very important tools here in order to achieve this. There's only, that much that we can do with just human efforts. Scaling up is a very important role of robots here. Another one is enabling new capabilities. For example, when we are talking about underwater robots, it is not possible for humans to go as deep as a robot can go and take those necessary measurements for informing climate related research. So these two factors: scaling up and enabling effects are, very important. For scaling up, the example that I can give is the recycling robots, right? So we need to process much more waste with much higher accuracy than we can do with other types of machinery or with humans. So now, by adding robots to the picture, we actually are able to boost the capacity of these plants. However, we need to be very careful about the impact of the robots to human workforce. I think people have very legitimate concerns. They are saying, oh are these robots going to steal my jobs? And sometimes the response from the robotics community is like, well, robots or in general, automation, traditionally always created jobs. So there will always be new jobs. And these are dull, dirty, dangerous, uh, jobs anyways, that needs to be automated. I really think that it is not as simple as that. We need to be very careful. We need to look into specific areas, basically assess this in a micro scale a little bit. What this particular automation for this particular industry will have this kind of effect to human workforce. We cannot really just generalize because indeed some automation technologies could be very detrimental to, the conditions of  human workforce. So designing them effectively is very important. I would again, want to emphasize the role of human robot collaboration here. Again, the role of robots here is scaling up the efforts and enabling new capabilities for humanity to deal with the climate crisis. This has to be a collaborative effort. It is not a competition because this is, this is a really serious problem. It is not about profit. It is not about whether this company is better than the other company. This is really something that we are all in together and oftentimes what I see is the best way of implementing these technologies is to identify how humans and robots can truly collaborate, together. So I think we should be very considerate about focusing our efforts to identifying these kind of entry points for these kind of robotic technologies. 

Cain: So very important to emphasize. It's interesting you mentioned that producing robots themselves has a carbon footprint and a lot of folks that are interested in this may be wondering, well, that's kind of goes against the whole point here. So, are there efforts underway in the climate robotics field to try to, address that by creating robots that may be more eco-friendly?

Calli: Absolutely. So, using materials that are more sustainable and even biodegradable, is a ongoing research effort in climate robotics. So with these new kind of materials, for example we would be less worried about creating electronic waste as well as making another piece of trash in the ocean while we are actually trying to collect some data for the climate scientists. So for example, some of these robotics solutions use less actuators. They use less sensors or in their body, they use soft materials. These become basically means that we have less electronic waste. Or if we are using biodegradable material in robot's body, for example we would be much less, worried if this robot gets lost in the ocean while it is getting used. And this is also important about the cost of these robots, right? If we are talking about scaling up our efforts using robotics, it means that they need to be affordable. And using these kind of perhaps cheaper materials, like utilizing soft robotics technology, for example would allow us to scale our efforts in a much more sustainable way.

Cain: That's really interesting. There's a lot of different advantages to taking that approach. And I think it's another example of just challenging the notion of what a robot is or should look like. I'm glad that you shared that that there's a different way of looking at things here. 

Calli: Absolutely. And oftentimes climate roboticists need to be very innovative, uh, because the concerns of today's robotics is not the same as concerns of, let's say a couple decades ago. And the research field is definitely responding to that and we have very exciting, research projects just focusing on how to make robotics technology more sustainable. 

Cain: I'm wondering if you could talk a little bit more about, um, some of the other research projects at WPI that professors are working on in this climate robotics space. 

Calli: Of course. So we are actually six faculty in the robotics engineering department that, are focusing on, different aspects of, climate robotics. One of the examples was indeed, the work that we did with Cagdas Onal. Um, so Cagdas designs these lizard robots. These robots are soft robots. They are compact robots that can go inside the building envelopes and take some measurements out of that. And this is very important because if you don't have these kind of robotic systems, what you need to do is to basically tear down this whole wall and which is much more costly and invasive, which we wouldn't want to do that and we wouldn't have the capacity and resources to do that, right? So again, the role of robot here makes the process much cheaper. It makes it much more scalable and collect the data that is required for climate resilience. 

Cain: Are they looking for things like leaks, uh, that might help them figure out where they need to put insulation, that type of thing? 

Calli: Absolutely that, yes. So these robots have sensors. So they got some thermal readings as well as they generate the 3D map, of these environments. So if you, for example, think about these hundreds of years old buildings, you really don't have the plans of them anymore. Uh, so it is a very valuable information to understand for people to see how these building envelopes look like, what their condition is, as well as whether they have leaks or thermal inefficiencies.

Cain: And, I know one of the other projects is sort of affectionately described as the, the robo bees. Do you wanna describe what that is? 

Calli: Of course. Yeah, that's my colleague Nitin Sanket's project actually. He is an expert on designing these small UAVs, he actual designs all kinds of UAVs. But, uh, this one is a very small one. It has a particular perception system in it that identifies flowers and it is then used for pollinating. So the idea is to basically design a large set of these bees, uh, so that they can improve the pollination process when needed. 

Cain: Very cool. And UAV. That’s unmanned aerial vehicle. Is that Right?

Calli:  Yes.

Cain: Cool, Cool. So, Berk, I'm wondering if you could tell me a little bit about how WPI students are involved in the field of climate robotics, is there, course material that they're learning on this? 

Calli: Yes, absolutely. So they are involved in all the stages of these research projects, uh, undergraduate students or graduate students, alike. And of course they sometimes do the field work, sometimes do the technical developments, sometimes provide innovative solutions, novel algorithms. So they are participating in every stage. They're actually the people who are doing the work here. We also have various education programs that are actually unique to WPI. For example, we have a environmental robotics concentration with some faculty from environmental engineering department. We put together a set of classes that the students can take if they want to specialize in certain areas of environmental robotics or climate robotics. Also, I teach a course called robots for recycling So this course is one of its kinds. I don't know of any other course that focuses on this project. Again, as I said, the recycling, problem has so many interesting robotics aspects. So it is actually a very intriguing case study for robots and for study in a classroom. So we are taking advantage of this and students love it because, they are very passionate about finding solutions to these environmental problems. And when there's a course like that or an MQP project like that, these classes receive a lot of attention from our students.

Cain: You mentioned the MQP – that’s the major qualifying project. It’s one of the required student projects for graduation. It’s a professional-level research or design experience that’s specific to a student’s major. You know, you talk a lot about the potential for the future, um, but in some sense robots have been involved in environmental monitoring for a while now. So I'm wondering what's new about this next generation of climate robots that researchers are currently working on.

Calli: So the new AI technologies actually, as well as certain improvements in robotic hardware really boost the capabilities of the robotic systems. So just like for example, in, the area of manufacturing, we see more flexible robots, that are able to not just focus on one single task, but can switch between tasks or can be easily programmed to do new tasks. This kind of capabilities are also there for, uh, climate robotics. And they're very important for climate robotics because when we are talking about climate robotics, these robots need to go to the field, take measurements or intervene to whatever is happening, and that means that they can, their role is not constrained into one single environment that is controlled. On the contrary, the environment is oftentimes changing. It's oftentimes very new and robots need to adapt to these changes. And the new robotics and AI technologies allow us to get there. But one important thing here is that we are not just there yet for coming up with these like fully autonomous robotic systems. In order to develop solutions in the near future, we definitely need human-robot collaboration. Let me give you an example about this. Our shipbreaking robot. When we were designing the shipbreaking robot, we noticed that the problem is really complex in many ways, but especially for a robot to understand where to cut a metal piece of a, let's say a ship hull, is quite difficult for humans. It requires long training, uh, sessions and uh, years of expertise and just programming that for a robot, becomes really challenging. And if the robot makes mistakes, it creates dangerous situations. So in order to prevent that, we came up with a human-robot collaboration framework that the humans spray paint the areas that the robots needs to cut. And the robot looks at those distinct color spray paints, identifies those areas and then cuts them automatically. So the time taking part of the job is still done by the robot. But the human expertise plays a key role to enable this technology in the near future. Maybe 20 years from now, 30 years from now, maybe robots might be able to more reliably decide where to cut, these areas. But we need these kind of technologies now due to the urgency, of these problems. And therefore human-robot collaboration plays a very crucial role to enable them. 

Cain: Yeah, I picture, the worker and the robot as sort of a team tackling the challenge, you know, as you've described it, to me today, kind of see that the teamwork and the power of, combining the worker with the robot.

Calli: Absolutely. That team really makes it happen. 

Cain: It's been great to hear about some of the different projects WPI is involved in here. Obviously, uh, WPI isn't alone in working on these types of climate robots. I’m wondering who else is working on these challenges and is there a community that's sort of being built around this, topic?

Calli: Yeah, definitely. There are, many institutes, that have a history in climate robotics. For example, Woods Hole Oceanographic Institution right here in Massachusetts. They have been working on these subjects for decades now. And we have actually good collaborations with them research wise as well as our students go to internship there, for example, every now and then, et cetera. So we have good connections with them. Climate robotics efforts are growing all around the world. But indeed the community building efforts play a very key role here because climate robotics, by definition, is an interdisciplinary field. Roboticists need to work very closely with climate scientists in order to come up with effective solutions. So we are actually, creating a climate robotics network. So the listeners can go, and Google Climate Robotics Network, and it'll come up. So this is basically a group of, researchers and companies who are working on various different climate solutions. We have a climate robotics summit organized every year. And we just organized our second climate robotics summit. This international summit is completely free. And anyone can join. And we bring up experts all around the world to, talk about their research. But it's not only researchers, again, there are companies in different sizes as well as, funding agencies. Because funding climate robotics, as you can imagine, it's not like funding a manufacturing process, right? Which you might find a lot of private funding, whereas the, uh, stakeholder for climate problems is usually public. So the funding landscape of climate robotics is also growing. It is changing and it is important to, uh, get to know about that as well. That's what we are trying to do with many other colleagues around the world to bring this community together so that we can exchange ideas and, knowhow. 

Cain: Yeah, it sounds like a great way to share information, keep the conversation going, and get the people that are really interested in this, all in the, in the same room, you know, whether it be physically or virtually to to keep things moving forward. We'll add a link to the, uh, Climate Robotics Network page in our show notes. You mentioned funding and, and there's often a lot of headwinds and hurdles when it comes to, uh, development of new technology. I'm wondering if you could talk a little bit more about, uh, the challenges that developers of robotics for climate solutions are, are facing. 

Calli: Yeah, absolutely. So it is a very different landscape than any other robotics domain that I know. Because let's say if it is medical robotics or if it is manufacturing, there is oftentimes a clear idea about what needs to be done. There's already a well-established knowhow. There's already a well-established community. And the funding mechanisms can be private, can be public, but either way, they are quite well-defined. Of course, they're looking for innovative ideas, but its framework is well-established. Then we come to climate robotics. It is a very rapidly evolving field. The research is oftentimes funded by public institutions. Um. There is some private initiatives as well, private funding agencies, venture capital firms, that is funding both research and development. Either way, oftentimes they are actually indirectly selling their products to the government, right? So it is not that a car manufacturer is basically requires a certain type of technology. It is like, okay, we are trying to prevent climate change or mitigate its effects. Oftentimes the, stakeholder here is the general public. 

Cain: We've, we've covered a lot of ground today and we've hopefully opened some eyes about the potential of climate robotics. If people are interested in learning more about the field, what do you recommend they do? 

Calli: Oh, I recommend them to go to our Climate Robotics Network page. There, they will see that there's a white paper that researchers from all around the world basically contributed to. So that would give them a good summary of the current and future potential of robots to basically intervene and mitigate climate problems. The other resource we have on that website is a global map, so it's a map of different companies all around the world which I believe they will find it interesting to see all these activities going on. Another one is we have a YouTube page where the Climate Robotics Summit talks are published at. So that would be another resource. And again, I always want to emphasize that this is actually a big community effort, uh, with researchers, uh, all around the world contributed to. It is not specific to our lab in any ways. We are, one of the contributors. But otherwise it is, it is a little bit of a messy field. So it is like, depending on which climate problem that you would like to learn about, you may need to check different resources. And there's no well established, let's say there's no climate robotics textbook, isn't it? So it is uh, it is a developing field and that's one of the challenges, like where to find reliable, well created information. And that's, that's what we are trying to do with the Climate Robotics Network. 

Cain: Is that exciting or is that a little bit scary that you're sort of like on the ground level of something that's just, ready to emerge and evolve into something, unknown? 

Calli: I would say 70, 80% of our time, uh, goes to understanding the problem first and understanding how we can develop these technologies so that it can have the highest impact possible. And then development really comes as the next thing. And again, many other fields in robotics that are already well established looking to develop specific technologies to specific problems do not necessarily have this kind of challenge because they might already have well established network and knowledge base. But it is definitely exciting. But of course it is also like the urgency of the climate problem makes it at times stressful, I would say to understand ways of how we can basically get this technology out in the field as soon as possible. Things are sometimes not as fast as we hope them to be. This understanding the problem, this interdisciplinary network aspect of it takes time and yeah, we don't have that much time.

Cain: Berk, this has been wonderful. Thanks for a great conversation. I've learned a lot and I appreciate you taking the time to tell us about, the field of climate robotics and your work and some of what our students, are doing. And I look forward to checking in with you again as the field continues to emerge. 

Calli: It was a great pleasure. Thank you very much.

Cain: Berk Calli is an associate professor in the robotics engineering department at WPI. You can learn more about our robotics engineering programs, faculty research, and the university's work on sustainability and clean technology innovations by visiting our website, wpi.edu. This has been the WPI podcast. You can hear more episodes of this podcast and more podcasts like this one at wpi.edu/listen. There you can also find audio versions of stories about our students, faculty and staff. Everything from events to academic projects. Please follow this podcast and check out the latest WPI News on Spotify, Apple Podcasts, or YouTube podcasts. You can also ask Alexa to open WPI. This podcast was produced at the WPI Global Lab in the Innovation Studio. I had audio engineering help today from PhD candidate Varun Bhat. Tune in next time for another episode of the WPI Podcast. I'm Jon Cain. Thanks for listening.