Listen now or check us out on one of these platforms!
This week we talk to Brian Ringley, a man who works with robots that can work in the human work environment and don’t need special treatment for their job. SPOT is a quadruped automated robot that can traverse the standard construction site without special accommodations. Personally, we just wanted an opportunity to pack-bond with a robot dog so we can have a friend when the robots take over. Brian brings some really awesome concepts about how SPOT’s vision helps it traverse, how SPOT is being used, and non-construction applications that teach us how to expand our use case.
If you have anything we missed, feel free to text us! - 478-221-7009
Please consider subscribing to our podcast!
Tyler: The Construction Brothers Podcast. I'm your host, Tyler Campbell. And with me like he is every week: My brother, Eddie Campbell.
Eddie: What's up, Tyler?
Tyler: Not much, Eddie. Well, this week we're talking to Boston Dynamics.
Eddie: All right, Brian, thanks for joining us today. Why don't you tell us who you are and what you do, man?
Brian Ringley: Yeah. Thank you so much for having me. My name is Brian Ringley, I'm the construction technology manager at Boston Dynamics. And what that means is that I'm working really closely with all of our construction customers for the SPOT robot.
Tyler: Oh, I've got so many questions for you, sir. So first, I want to kind of clear the air. Is SPOT actually a dog? Is it some other animal? I just want to clarify, is it actually a dog?
Brian Ringley: Our market speak is that it's an agile mobile robot, to differentiate its capabilities from other mobile robots. In robotics locomotion world, it's a quadruped or a four-legged robot, but you know, whatever animal customers want to call it, they can call it. I think it looks more like a grasshopper than a dog, personally. But yeah, I think “robot dog” is probably the most common name it gets.
Tyler: Nice. Well, tell us more about this platform, man. I mean, we've been seeing a lot about it just in the news cycle. I'd love to just kind of hear some more about it from your end.
Brian Ringley: Yeah. So you know, SPOT was released as a platform, meaning that it provided a kind of generic mobility solution for lots of different markets, to be able to carry sensors to places of interest with kind of a specialized capability to move around in human-purposed environment. So up and over obstacles, up and down stairs, things that other types of automation really struggle with. When we were thinking about robotic mobility, there's obviously a mature robotics market in the logistics and manufacturing space. But it often requires that you design the environment around robotics. So when we're looking at the markets, we're trying to target construction environments specifically, we're looking at environments that are really designed around humans. So how do we let people on those sites continue to work the way they're working, but be able to introduce intelligent automation into them?
Tyler: That would be a very, very difficult environment to navigate, I would imagine, because exactly what you just said is that most of the time the buildings are designed with robotics in mind. So we talked about Amazon a lot, the ones that go and pick up the shelves and bring them over to the people, that sort of thing. I think it's just Amazon Robotics now.
Brian Ringley: Yeah. Used to be Kiva Systems and they acquired it. Yeah.
Tyler: Yeah. So, I mean, we talked about that, but the warehouse, it's designed for these robots. I would imagine it would be extra challenging working on a job site though, because I mean, just stacks of rebar will be laying somewhere one day and then not there the next, and then other piles of whatever would be sitting around as well.
Brian Ringley: Yeah, no, what I joke about is, I think people struggle to move through those environments and locate themselves in this environment, let alone robots. And I think one of the trickiest parts is that it changes every day, right? It's an ephemeral environment, it's a work in progress. It's not going to stay the same over time, which is also a very hard perception problem.
Tyler: So how did you guys handle that issue? Like just, I would imagine just a ton of sensors.
Brian Ringley: Yeah. I mean the biggest thing is something we call our enhanced autonomy package. So there are various payloads, which is basically additional tooling or sensors that you can attach to the top of the robot. And the base robot comes with five 3D stereo cameras around its body, and that gives you really good perception and autonomy in what I would call fixed environments like industrial environments. When it comes to construction, we found that having a kind of further field of view was critical to being able to understand what are the more permanent features that can be recognized over time. You know, things like core and shell, superstructure, and what are the things that are kind of more ephemeral dynamic or changing? So the enhanced autonomy payload is a LIDAR unit that basically increases what SPOT can see up to around a hundred meters. And that's been super critical in its ability to perceive those changing and dynamic environments. And it's a problem where we're still working on, of course, which is to extend the life of the robot maps that are used to achieve autonomy in construction and construction sites.
Eddie: What kind of cool things are you seeing SPOT used for?
Brian Ringley: Yes. So, you know, right now it's all about data capture. I think that a lot of GCs, we don't have to convince them about the value of data. They are already looking to enhance their data collection programs, but they're finding limitations with existing methods, whether it's trying to send people out to do the documentation, drone programs work really well for a lot of exterior and civil environments, but are extremely challenging indoors. And then fixed sensors and other systems are subject to visual occlusions and can be costly to set up and swap out. So I think that we're finding that having the mobile sensors through SPOT has been the biggest application. And I would split that into two things, really. One is autonomous laser scanning, which has been a really popular application because anyone who's been on a reality capture surveying team knows that that could be a little dull, just setting up a tripod, and then if you're doing high resolution scanning, you're kind of sitting there for like five to 10 minutes per scan, you're typically hiding behind something so that you don't get captured in the scan, like playing a video game on your phone. And those are highly skilled and invaluable surveyors and VDC professionals, too. So, you know, ideally they would be more engaged with their work and be freed up to do something else. So that's one major application. And the other one is continuous site monitoring. So, things like daily job walks where you're taking a 360 image captures or continuous 360 video of the same points of interest on the site over the course of the construction project, so that you get that kind of four-dimensional timeline and you can go back and have a record of what was done on any given day. You can compare multiple days for progress, and then you can start to have the type of data set that is large enough and well-structured enough that you can also leverage some of the new AI tools coming onto the market.
Tyler: One of the things that you had mentioned with scanning really stuck out to me, because one of the guys that works with us used to work on nuclear sites, and he used to go scan a lot of these sites. They would have to limit exposure in certain areas. So he would have to walk in, set up the scanner, walk out of the room, and then somebody else would have to walk in, turn on the scanner and walk out of the room. Are you seeing a lot of applications where it's hazardous for people?
Brian Ringley: Yeah, absolutely. And I think that's the easiest ROI case for the robot, where one use of the robot pays for multiple robots. So radiated environments for inspections or other operations in nuclear facilities. In non-nuclear energy facilities, like power plants and such, you often have to de-energize certain environments for required routine inspections. And to take that down for 12 to 24 hours can be very costly for that energy company. So the ability to send in a robot instead of a person to do that work not only keeps a person out of a harmful environment, which is invaluable, but also really translates to real dollars for that customer. And I think there are opportunities like that in construction as well, particularly with things like confined spaces where inspection may still be required.
Eddie: What kind of things are you seeing SPOT used for maybe outside of the realms of construction that we might be able to learn from?
Brian Ringley: Yeah, I mean, SPOT is being used in all sorts of really fascinating ways. I mean, we've seen applications in manufacturing and industrial in the entertainment space. That's a fun one. There's Latte, the dog at Universal Studios theme park where they've actually like dressed SPOT up, where there was a spot performance and Cirque de Solei, which is really interesting. There are a lot of YouTube influencers now, purchasing SPOT and making weird videos with SPOT, which is all good. But I think that we're learning a lot from the industrial applications and fixed environments, because those operators are really interested in the kind of mission control paradigm where they're saying, you know, “I need to have one operator responsible for many SPOTs. I need to be able to plan each SPOT’s kind of routine path. I need to be able to schedule those paths. And while I want full autonomy, I of course want to be able to monitor what each robot is doing. And in addition to all of those kinds of repeatable, scheduled operations, I’ll occasionally want to use my human insight to tele-op and perhaps inspect something on the fly.” We're working on a lot of software that's really geared toward those types of users that I think will actually be really valuable for the construction industry as well.
Eddie: I have to go back a little bit to something we've talked about in the past, but I wrote a blog post about tech and tactics. And we're describing SPOT as a platform. And the cool part of that is that platforms are meant for people to get imaginative on, and like use in the way that they can kind of dream something up to use it, right? So, yeah, SPOT is doing that. And are you finding that as you get SPOT out there that people are using it, maybe you guys didn't even know all of the ways that SPOT was going to be used?
Brian Ringley: Yeah, I think that's absolutely right. Last week we made the announcement about our partnership with DroneDeploy, but it was actually a three-way partnership that was very customer driven. So Brasfield & Gorrie were the GCs behind that effort, and working with their engineering team was an amazing experience because they were so skilled and in software and hardware. We offer an onboard computing solution. We have a lot of out of the box things to get customers started, and they didn't need any of that. They built their own computer, they built their own software using our APIs, they put together an entire computing and sensing kit on the back of the robot that really started with an interest in 360 video. But then they were like, “Well, what if we could tag other data points with this video and start to create heat maps? What if we could put environmental sensors on here? What are we concerned about to maintain a safe job site? Maybe we're trying to detect silica and we can start to map that across our site.” And that really all came from the customer. And we try to maintain that humility that we're really good at robots, but we need to learn from the expertise of our customers to really be effective and provide value in that industry.
Tyler: I've kind of got a clarifying question here. Because, is SPOT fully autonomous or do you have to have some level of control over the robot still?
Brian Ringley: I would say that nothing is fully autonomous, and then I'll backtrack and explain. It's one of those horizons that the more you learn, the farther it goes back. So, the primary operational paradigm for SPOT is that you the operator are there on the construction site with SPOT. You pilot SPOT with the controller through the job site. So you are indicating where you want the robot to go and where you want it to perform certain actions, whether that action is taking a 360 photo, performing a laser scan, doing a little dance, whatever you need, all of that gets recorded into a mission file. You can now replay that mission file and that's where the autonomy comes in. So now you're just sending SPOT on an autonomous mission. So it will repeat those exact same motions and actions and work its way through the job site while also exhibiting obstacle avoidance, which is the key. These are unpredictable and dynamic environments. You should tell personnel on site to give the robot the right of way, but if somebody does walk in the path with the robot, the robot will pause and then find a safe route around that person. If somebody left their job box in the route of the robot, it will find a way to safely move around that. If somebody dropped a pile of pipes on the ground, the robot will crawl over that. So, the robot will persist as much as it's safe. And then there's some intelligent user interface features where there are things that's technically capable of doing, but that might not be safe for your environment. So for example, now I've parked a scissor lift long ways, right, in the middle of the robot's path. The robot can work its way around that obstacle, but now you're asking it to move out of what we would consider to be a safe radius. We want everyone to understand and be able to anticipate the behavior of the robot. That's a big part of socializing robotics in human-purposed environments. So you'll get a message to the controller that says, “Hey, there's this thing in front of me, I would prefer that a human operator move me around this thing to ensure that we're doing this safely. And then once you get me back on track, you can resume the autonomy.” And then it goes back. The first thing that we're talking about as we move toward full autonomy is safety, first and foremost. There are a lot of things that can be done with robotics and technology today that I don't think should be done with regard to the concept of operations and safety programs on these sites. And then the next thing is actually starting to think through the technology challenges of what that actually means to have a robot that's on a site alone. So for instance, if we're talking about something akin to lights-out construction, where you have a night shift that's all robots, the robot will fall down. It falls down less and less every day as we improve the software of the robot, but it will fall down. That's just reality. So it has to have a way to get itself back up. So we're constantly improving self-right behavior. The robot has to have a way to power itself for long periods of time. So we have a self-charging docking station that's coming out in early 2021, so very soon. In certain environments, the robot will need to open doors. This is one of the many things that we hope that the arm payload, also coming in early 2021, will help customers achieve. So there are just all sorts of things, all sorts of scenarios you really have to think through, because a disabled robot in an environment where there are no people operating, so now let's jump from like construction sites to like offshore oil rigs. That's a problem. I think a good example of this is the Mars Rover, right? Like all of the kind of precautions and design specs that went into conceiving of this robot, because one significant failure and that robot just becomes space garbage forever. So those are the things you really have to think through when you think toward full autonomy. And I do think construction operators want to see more and more autonomy. That's a big part of the ROI argument for them. And they'll always want to be in a position to supervise that autonomy, whether it's through some kind of control interface or just through their phone, but obviously a big part of the value proposition here is that you are automating these kind of lower level tasks that don't require a lot of kind of complex human effort to free the rest of your workforce up to be more productive elsewhere.
Eddie: When we spoke before, we talked a little bit about BIM and BIM being a driver for some of this, and that came out of our episode with Tessa and some of that conversation. So point counterpoint, where are you at with BIM being kind of a driver of these robotics? Where are you at with that whole thought process?
Brian Ringley: Yeah, I think that— This is a very common question from people in this industry. So I think I'll start by saying that robots aren't going to fix the inherent problems of BIM that we have today. Definitely want to make that clear. There's a lot about the way that we use that software that doesn't fully realize the value proposition of BIM. The way that different stakeholders are siloed, whether it's contractually or because of various liabilities and risk, the way that that data often kind of gets disconnected from the live model when it gets sent into the field. And I think most of all, the fact that I think that we focus a bit too much on a kind of linear delivery process where there's design intent, that's dictated by architects and engineers, and then it's carried out exactly to spec by the professionals onsite. When in reality, a lot of the intelligence that goes into delivering a building and constructing a building is actually coming from the field. So there aren't sufficient mechanisms in BIM platforms as they stand today to ingest data from the field and to have this kind of positive feedback loop. And then even if there were, we’re still back to this problem of kind of disconnected models and siloed disciplines. So what I do hope first and foremost is not that people are going to be path planning their robot from Revit on day one, which I always joke, I barely trust models enough to construct a building, I certainly wouldn't want to use them to drive a robot, but for the robot to start to provide professionals on site with a data capture mechanism that can be easily fed back into the model. Now, again, I don't think BIM platforms as they stand today are really capable of that. I hope that those capabilities come into play. But there are some really interesting startups that are looking at that problem. One that I'm always giving a shout-out to, partially because I've got a couple of really smart ex, we were colleagues, who are working there as well as Avvir, A-V-V-I-R. They're doing some really interesting things where they can ingest point clouds captured on a frequent basis and use that information to drive updates to Revit elements so that you know what's been placed and if it's been placed correctly. All of these things depend on a really rigorous approach to site surveys and field coordination. You've got to have a shared control point and shared project coordinates, of course, and the robot can also support those coordinate systems from the kind of technical standpoint, but you really have to rigorously scan your environments and make sure that your BIM has been corrected to those real world conditions. And that's where a disconnect can happen a lot of the time. And it's hard work because a lot of it ends up being manual again, back to why the autonomous laser scanning application is so popular. I mean, it's a ton of work to go out, scan the site, and not just the data capture side of it, but then you have to take those point clouds. You have to register those point clouds and software. Then you have to superimpose or overlay that registered cloud with the original model or other digital asset. And then you have to kind of correct those two and make those two things shake hands a little bit, which really just means you've got a person spinning a model on a computer and nudging elements back and forth. So there's not enough automation there on the software side just yet. So I think with a combination of automating the data capture standpoint with a robot like SPOT, SPOT can lend a little bit of help to the data processing automation by putting that data into the right coordinate system, pre-positioning that data in a way that makes it faster and easier to register and overlay onto the original model, especially if you're doing it on kind of a frequent basis. Because then you've got like a potential backlog or pipeline problem. But you know, there's still need to be better software solutions to actually giving people insight. And we should really be to the point where no one is actually looking at a point cloud, like no one actually cares about that. That's a means to an end, and the end is some kind of insight, some kind of report, home kind of element update or position update to a model. So we want the result. And right now we're just kind of stuck in a halfway point where we're doing a lot of manual work.
Tyler: Yeah. That seems to be something that we're hearing a lot about, too. So you would say that the software companies in general, in the construction industry, are lagging behind the times and they're struggling to kind of keep up with what we can really do.
Brian Ringley: Yeah. I mean, and I just don't think that, if you look at the origins of like—if we just focus on BIM platforms in particular, I mean, when they were first conceived, I mean, there was no conception of having like this level of data capture onsite. I mean, the whole notion of reality capture, I think, came a bit secondarily. I think the original focus was on digitizing the kind of design process and automating documents from three-dimensional models, you know, having a database of what you're working on. But I don't think at that time anyone could have foreseen four-legged robots performing laser scans on a daily basis. But obviously that could be leveraged to improve these processes. And again, I also want to emphasize, this is not just a technology problem. This is a contract level issue. This is even an issue, a social issue between the stakeholders on the project. So yeah, I think there are things that BIM platforms or kind of these new software startups in the AEC space can help solve, but it's also gonna take some creativity around moving beyond the traditional design bid build contract structures so that the stakeholders are properly incentivized to collaborate in the way that's actually necessary. You’re going to make a kind of building delivery technology paradigm premised on collaboration, you can't expect that to work in an industry that's essentially premised on the opposite. Right?
Eddie: You said it was kind of a social issue. I remember when I first saw SPOT and I saw just kind of a demo of the basics of what SPOT could do. It may have been at Autodesk University, I saw the video from it. My reaction was, “Well, that's cool. What the heck would I do with it?” So, I mean, that's me just kind of being real with my primary reaction. Socially, how is SPOT being adapted by that proverbial crusty old superintendent that everybody kind of throws out there, the guys that are like, “Bah, I don't need that.” Like, construction's famous for that. And you're bringing something new in. How's that going?
Brian Ringley: I mean, it's going really well. I think obviously like, you know, the customers I work with, it's a biased situation where the type of GC who's going to recognize the value of agile mobile robots early on are not the proverbial crusty type, probably. However, that said, I think the industry gets a bad rap. I think there are a ton of incredibly innovative contractors out there that understand that they have to invest in future technologies to really achieve the outcomes they want to achieve as a business, to take on more work and to be able to deliver more work on time. I think it's been an incredible experience. I've met some of the smartest people I've ever met in my life just over the past six months by working with these customers. One of the things we're really trying to focus on right now with our website is telling more of their stories. So that's something, I'm really excited, that's going to be kind of coming out on an ongoing basis. So you're hearing it from them first, because it’s one thing for me to come on these podcasts and say my thing. You know, granted, I did start as a customer of Boston Dynamics as a construction researcher at WeWork. So I understand that perspective, but I think it's important that those customer stories be told, because that's really what's going to make the rest of the industry kind of realize, “Oh, this isn't just really cool—which it is, and that's great—this is a useful tool. This is something I can add to my fleet of equipment to make my workers more productive and to deliver projects more safely and efficiently.”
Tyler: It’ll be interesting to see what people come up with when they hear those stories, too. Because I know a lot of times I'll hear a story about somebody using this robot, or something along those lines, and my brain just starts spinning. I'm like, I could use it for this in my application. And to kind of see how it's going to grow from there, I think that's going to be cool.
Brian Ringley: Yeah. And I think the most interesting conversations actually ended up not being like, “Here are a bunch of ideas I have for how to use these robots,” but, “Here's a business problem I have, and if I logically worked backward through that business problem, it turns out that this robot is an essential part of that solution.” I mean, case in point is we worked with the contractor Swinerton. Their main research agenda was paying their subs faster. The industry average is horrible, I forget what it is, it was like 60 to 75 days, and people want to be beating net 30 payments. And there's a lot of incentive there, too, right? Because potentially you could get your subs to bid more competitively if they knew they were going to get paid super fast, essentially you could get a premium from an owner if they knew you were going to be running a tighter ship. So obviously you're kind of incentivized from both directions, which is rare. And part of automating a payment process is finding a trustworthy way to also automate work in place measurements instead of just kind of putting your finger in the air and feeling the wind and say, “I got about 50% drywall up, that's fine, I guess I'll pay you. But first we're going to argue on the phone for like two weeks and burn up a bunch of cash.” So if you kind of keep working your way backward from that automated work in place detection, again, you need a robust kind of data collection program. Swinerton then started evaluating all the possibilities there and really found that SPOT, plus a couple of other kinds of software technologies, and of course workflow changes, that they could actually start to get closer to that goal. So it was really cool to see this conversation that started with, “Here's a business problem we have, and we're looking for a solution, and we eventually kind of derived SPOT as an essential part of that,” versus a really cool technology in search of a solution.
Tyler: I never put those two together. Having SPOT out there collecting some of this data and saying, yeah, they are 50% done, or they're 43% done. And that ties in really well with our conversation with Pete Dumont on smart contracts, and making sure that people are getting paid on time and quicker and incentivizing. I never put those two together, but it will be really interesting to see how robotics and smart contracts end up kind of working together in the long run.
Brian Ringley: Yeah. I mean, a lot of the technologies and improvement processes that we talk about depend on good, trustworthy data. And you've got to therefore figure out how to establish a meaningful data collection program, and that's really the value that all of our initial construction customers have seen with SPOT. And it leads us to really interesting places, talking to real estate professionals and owners and operators about once that project is delivered and they have this robust record of its construction, how to continue to use this technology for asset management and data capture in a finished building, which frankly is a hell of a lot easier than tracking a construction site. So there's no reason that you couldn't use these robots for years and years over the life of a single property, and just continue to get more value out of it.
Eddie: The what ifs that are spinning up and spooling up in my head. It's happening, you know, all the ideas. We've talked about a feedback loop incessantly, and you think, well, yeah, I mean, SPOT’s great for the contractor for the guys that are out in the field, but what's the designer really going to get out of this? But creating a feedback loop for a designer to know how it went in the field, to return data that says, “This didn't go so well, don’t do this thing again.”
Tyler: Yeah, how did that detail actually work out in the field?
Eddie: Yeah. “Don't use this detail anymore, because when you do, it causes these issues.” Just the ability to capture and bring that back up to the surface and have them know now, rather than just have it buried, just creates a smarter design to begin with.
Tyler: Yeah. You have to have somebody with the wherewithal to be able to point that out, too. Or something.
Eddie: Having these recycle back in, like you're saying, to Revit elements, or whatever BIM software you're using so that things are automatically updating and maybe— I mean, as-builts are, well, that's a topic right there, right? You know, getting a true as-built? Tell the architect, your model is going to match the realities of the field when it's done down to a quarter of an inch in every regard.
Brian Ringley: Yeah. And it's four-dimensional, too. So if you need to rewind it to check on something in the past, you have that information, too. And you can continue to build off of that into the operation of that property. I mean, I started in architecture. I have an M.Arch, I worked on a design technology team at Woods Bagot prior to getting into construction at WeWork. So that's kind of where my heart lies. It's like, how do we extend some of these like hyper intelligent ways of managing design intent to the field, but also incorporate all of the other intelligence that comes from the experience of the trades to really deliver better buildings?
Eddie: I wanna go ahead and hit you with our megaphone question. So if we gave you a megaphone and 60 seconds to speak to the whole construction industry, what would you tell them?
Brian Ringley: (laughs) I believe it was told to prepare for this. And I might've forgotten to do so. But you know, the things that I'm screaming about, there are many things, right? It's to not confuse technology problems with business model problems. There's a lot of griping and groaning about how technology isn’t solving things, whether it's BIM or something similar. And it really comes down to the responsibilities of each individual business to change their models of operation, to write better contracts, to be more creative about how they're generating revenue and to be able to intelligently take on risk where there's adequate reward for that. Coming from the architecture side of things, where there's a lot of moaning and groaning about BIM and that not realizing value, and it's like, well, you've been doing business the same way for like a century, like just automating the production of drawings from a 3D model is not actually going to change the world. Like you really have to rethink how you're doing business. Part of that is obviously embracing new technologies like robotics, but I think an even bigger part is embracing new business models.
Tyler: Man, we're going to need to have you back on for that conversation, too. I'm sure we could go another hour pretty easily. Yeah man, so where can people find you?
Brian Ringley: If you just squish my name together, Brianringley, all one word, you can find me on Instagram and Twitter, or you can search me on LinkedIn. If you want to reach out about SPOT, you can go to Bostondynamics.com and use our contact page, and mention me if you want to have a conversation about construction. So I'm fairly easy to find.
Tyler: Very nice, man. Well, thanks for joining us this week.
Brian Ringley: Yeah, thank you so much, this was great.
Tyler: Hey guys, Tyler here. Just really quick before you go, do me a favor and leave us a rating. If you're listening on Apple Podcasts, it helps us out so much. Make sure that you join our community group. You can text us at (478) 221-7009. And also, go check us out on social media at constructionbrospodcast. We're on Instagram, LinkedIn, Facebook and Twitter, and also on TikTok, believe it or not. So thanks for being with us this week. Hope you have a great one. Go build something awesome. See you next time.