Autonomous construction drones map progress on billion-dollar projects

Who knew, but maintaining state and level of progress on massive construction jobs is actually a really hard problem.

Exyn Technologies has adapted their mine mapping level 4 autonomous drones to work on construction sites and feed completion data into construction software packages. The global construction industry wastes about $3 trillion a year on mistakes, scheduling errors, re-work, and fixing errors. The promise of always knowing current state of your massive construction project is that you’ll be more efficient, scheduling will be better, staggering of work and trades and steps will be more efficient, and you’ll have to do less demolition and re-construction due to errors.

In this TechFirst, we chat with CEO Nader Elm and COO Ben Williams. Check out the story on Forbes, or keep scrolling for full video, audio, and transcript.

(Subscribe to my YouTube channel)

It’s super-hard to deploy drones in construction sites. They’re always changing, messy, busy, noisy, and full of people and machines. Exyn, however, says they’ve been able to do it safely.

Subscribe to TechFirst: Autonomous construction drones

Transcript: Exyn Technologies deploying flying construction drones for Obayashi

(This transcript has been lightly edited for length and clarity.)

John Koetsier: What enters your mind when you hear the phrase “flying construction drones?” Impossible? Science fiction? What are you smoking? Construction sites are hard, they’re busy, they’re dangerous. It’s always changing, and flying drones have battery issues, strict cargo capacity limits, stability issues in wind, and they can’t just accidentally, sorry, chop someone’s hand off.

One company, however, says they’ve cracked the code. It’s Exyn Technologies. We’ve had them on TechFirst before, talking about their Level 4 autonomous drones. Back to chat about actively using flying drones in construction scenarios is CEO, Nader Elm, and COO Ben Williams. Welcome back, Nader. Welcome, Ben.

Ben Williams: Thanks for having us.

Nader Elm: Thanks, John. Good to see you again.

John Koetsier: Good to see you both again. Let’s dive right into it. What have you built?

Nader Elm: So, we’ve, as you know, had Level 4 autonomy available on drones for over a year now, and its primary use has been in underground mines. And we’ve identified an opportunity beyond that. And the near adjacent opportunity we saw was in the construction vertical, where similar kinds of requirements exist. They’re trying to map a project on a frequent basis, to inform their kind of construction decisions on a frequent basis. So, that’s what we do in underground mines, and we saw the opportunity to move into the construction space doing the same.

John Koetsier: Underground. I mean, vision is limited. You might have to use some sort of radar or echolocation, even, right, to find your way around all that stuff. Above ground, you’ve got vision, but you’ve gotta deal with weather, you’ve gotta deal with wind, all those other things. Ben, talk about what this drone will do.

Ben Williams: So, the real key for us here is that in most of these industrial environments, you can’t count on having an immobile or static environment. Things change, it’s dynamic.

Even if you have nothing planned to move, someone’s leaving a forklift, or a truck, or a bag, or whatever in the way of things, and so, because of that, you need a system that is able to respond in real time to changes in its environment.

And in many of these cases, also, especially if you’re within the skin of a construction project, within the building, or in large facilities where you can’t count on being outdoors with good GPS, in those contexts, you really need a system that is able to be resilient to loss of GPS, loss of line of sight, the ability to self-direct its own safe navigation path and corridors.

And so, that’s what we bring here, is it gives you the ability to do operations within very complex, difficult environments above ground, but still indoors, or in above-ground areas where you are still in constrained environments and GPS isn’t enough, or GPS is not available or consistent, and you still need to conduct all of these kind of surveys.

John Koetsier: So, give us a quick sense then. What is the drone for? What is it doing on the construction site?

Ben Williams: So, from a high level, it is basically a way to capture a representation of the physical reality digitally.

You hear this most commonly in terms of reality capture, or VR environment generation, or digital twins. These are all different terms for roughly the same process of digitizing a representation of a physical environment.

And where that becomes really important is, previous to this, if you wanted to go see the status of a construction site, you would have to go walk the site, and, typically, people would bring, you know, before cell phone days, it was you bring a camera with you, DSLR or point-and-shoot or whatever. And nowadays, it’s probably a cell phone, but you’re still just taking pictures, and then you have left a bunch of pictures that you then have to try and contextualize, and that is all stuff that happens in someone’s head, unless you try and put it into photogrammetry software, in which case you have to make sure that your lighting was good, you’ve captured enough angles for it to calculate where it is, or you have enough information captured generally.

And that’s a ton of work, and it takes a long time. Or, you can put a tripod system down and do a capture every 15 or 20 feet, and then you still have to take a huge number of measurements. So, the basic idea here is that you are massively streamlining that process.

The same thing that would take you all day to capture from traditional methods might take you an hour, or even less, with an autonomous system, and so you are both increasing the accuracy, the speed with which you can operate, and you’re able to actually run these sort of data analyses digitally that you weren’t able to even do before.

John Koetsier: So, I’m guessing the people doing this work and spending the day, pre-drone days, to take those photos isn’t … they’re not just putting those photos in “my construction scrapbook,” or posting them on Facebook, you know? “This is what I do, mom.” Why are they taking these photos? What is the purpose here? Why is this important?

Ben Williams: So, there’s a bunch of different things that we know about already, and the mining use case is instructive for us, because we give these systems to them, and they’ve found a dozen more use cases as they’re able to gather more data.

But the things that we know are really important, and that people are already buying for, there’s the ability to track progress of an install or progress of a build, especially for a large facility, where you have layers and layers of different systems going in. Those all have to go in in a very specific order.

If someone installs something in the wrong order, then what that means, if you don’t catch it early enough, you have to pull down walls, or HVAC systems, or piping, or whatever, in order to get to what was supposed to go behind it all, and you get this massive amount of rework. You get schedule creep and scope creep, and it costs a lot of money.

And so, the whole point here is you’re able to get an accurate representation of what’s happening in the project in near real time, for someone that may not have the capacity to go and walk every inch of every site weekly, right? That’s a huge amount of time.

John Koetsier: Interesting. So, what I’m getting the picture here is that there’s a digital reality in some software of what should be built, which people get plans from, and they go and build it. There’s a physical reality of what is actually being built, and here’s a step to check, is what’s actually being built what we planned to build? Is digital and actual on track, on pace, synched in lockstep? That is really interesting.

So, you’re saying that this drone will fly inside, potentially outside as well, of a large building under construction, check every room, check every space, map it all, understand it, be able to do that autonomously, and report back. You’ll get … what will you get out of that? You’ll get some mapping which will enable you to tell it’s on track here, or it’s not?

Nader Elm: I was just gonna add to that. So, that’s exactly the opportunity that we saw, which, just to put it in numbers…

So, McKinsey actually did a study where they quantify the global construction industry as an $8 trillion industry. And they also quantified that $3 trillion of that is actually down to inefficiency and waste. So…

John Koetsier: Every homeowner doing a construction project or renovation project knows that [laughing].

Nader Elm: They’ve absolutely lived this. Anyone who’s done large-scale projects has absolutely lived this. And the interesting thing is that it’s not surprising, because the industry itself is actually quite complex.

If you’ve got a general contractor reporting to an owner, and managing all the different trades, the friction is down to the information that gets passed from trade to trade, from trade to the general contractor.

And, as Ben said, the information capture, up until recently, has actually been very, very clunky. So, now, with the much better capture devices, and combining that with robotics, that’s basically where the opportunity to unleash greater efficiencies and less friction. And in doing that, that’s basically where we see a huge opportunity, and, to your point, it’s twofold. Number one, it’s comparing the as-built to planned, so you can identify variances and understand whether that’s okay or not, and what you can do to remediate.

But also, comparing yesterday to today. So, there’s a temporal element as well, because in doing that, now you can understand how well you are tracking, or even do measurements, like, for example, how much concrete has been poured, and should we be paying as much as we are for that particular material? So, there’s a variety of different things that are being calculated from the data that is being captured, and we can really make that more efficient.

John Koetsier: Nader, the software implications for this are immense, because if you can fly this drone on a daily basis and just capture daily information here, input that into some construction software platform that then sort of automatically checks what’s happening, what’s done, volume, features, build-out, all that stuff, and just gives you a report card, that would be huge.

I’m assuming somebody still has to actually look at all this footage.

Ben Williams: Well, not all of it. A big piece of what is really interesting about this is that because you’re not just gathering a bunch of pictures, where someone has to interpret what’s going on there, you are capturing structural, geometric data.

And so you can actually run automated comparisons, where it will highlight for you every major change in physical structure, so that you, as a project manager, doesn’t need to go through and look at every element of the data.

Instead, the software that you run can flag two or three things where it says, “Hey, I’m not sure what’s going on here,” or “Here’s a big change. Should this have happened?” And then the project manager doesn’t have to spend as much time kind of wandering through a thousand photos. You just bring up the two flagged areas, and can I identify that, “Yep, that was supposed to happen.” Or, “No, that shouldn’t have happened,” or “it shouldn’t have happened in this way,” or whatever.

John Koetsier: That sounds super helpful, because you’re not just working with Joe’s Construction Limited right here. You’ve actually got a beta client, and the beta client is the largest construction firm in Japan.

Ben Williams: Mm-hmm.

Nader Elm: No, that’s right. So, Obayashi is a very, very interesting firm. They get involved in very, very large-scale industrial and commercial projects. And what’s unique about them is that they are vertically integrated. So, for us, it’s great, because now we get involved with a firm that gets involved right from the beginning, in terms of the design, all the way through to handing over the keys, so, of the project. And that gives us the entire lifecycle view of a project, and we can understand exactly where we’re having impacts, and where we can actually deliver most value, so we can focus and refine our products from there.

John Koetsier: Mm-hmm. Excellent. So, let’s talk about features, capability, specs.

When I got the pitch, I was thinking “flying construction robots.” My mind instantly went to, you know, is it welding? Is it hammering? I’m joking on that one. You know, is it carrying stuff around? What is it doing? Obviously, in your scenario, what it’s doing is it’s mapping, it’s seeing, it’s sensing, it’s detecting, and it’s reporting on that stuff. Ben, what are the capabilities here? What are the specs?

Ben Williams: The core of it is actually very similar to what we talked about last time, which is to say the core software, the autonomy software, is effectively hardware-agnostic. So, you can use that on a number of different aerial systems, ground systems. You can hand-carry it, or mount it to a pickup truck or a construction vehicle, and continuously map that way as well.

What we are specifically piloting with Obayashi is a modular aerial system that has a light, sort of mesh cage around the props. Because it’s such a difficult and debris-strewn environment sometimes, we’ve sort of done some customizations for survivability of the system, and to make sure that we can operate in the areas where they most need it, which is often where there are construction materials lying around.

And so, in this particular case, we are delivering on a Ascent AeroSystems drone. It’s a dual coaxial system that is actually modular, and so, easy to fit inside of packing cases and that sort of thing. And then, the cage around it as well. This system is a little bit custom in the sense that we are trying to get it through smaller spaces. Now, one of the biggest challenges for these types of systems is that you need a certain amount of compute, and a set of sensors that weigh a certain amount, and are, you know, nontrivial in size. And so, you need a certain size of drone in order to carry those any meaningful distance. And so in a lot of cases, you’re managing tradeoffs between flight time and size of the sensor, and that sort of thing.

So, we opted for a system that is a little bit narrower, and a little bit taller, which allows it to get through slightly narrower spaces, which is a good trade-off in the case of these construction clients.

John Koetsier: So, somebody comes to you tomorrow. Maybe the biggest construction firm in the U.S. or in Europe. Can you deliver the system? Is it operational? Is it still beta? When do you launch?

Ben Williams: So, I would still call this iteration beta, but it is available for purchase. You know, we’re delivering to … we already delivered to Obayashi. They’re off doing stuff. And the core of the system is the same software.

It’s almost complete reuse from mining [inaudible] clients, and we also do infrastructure inspection and critical asset monitoring and that sort of thing. And so it’s, at that point, it’s almost entirely reuse from those other use cases. But, the things that have been customized a little bit here is that we have focused some of the set points, some of the settings files specifically for these environments, and that’s part of the purpose of this pilot with Obayashi as well, is that we’re going to be testing in their real-world build environments, for real projects, and learning about it.

There’s, I’m sure, things we don’t know yet about how the system will need to be tweaked and modified. And so that’s why this is so important of a first step, for both us and for Obayashi, because they can learn how to integrate it in their workflows as well. And, you know, that’s a great partnership to move forward with.

John Koetsier: Are there software platforms that you’re working with as well to ingest the data? You mentioned already that in some cases, this is already automated. Are there multiple… I have no idea what software somebody uses to build a $1.5 billion skyscraper, but I assume there’s some package out there. Are you working with people like that?

Ben Williams: Yeah, we have some early partnerships with a handful of companies that are doing what’s called digital BIM [ Building Information Modeling ] platforms. And so, those platforms are able to ingest the data that we create, and then they can do some, like, comparisons.

You can also use a more manual process. There’s a lot of sort of point cloud, or geometric mesh software solutions, where you can do a little bit of this manually.

And so, sometimes folks will take that first step with an off-the-shelf platform that is just for manipulating 3D models, and then get a sense for what’s possible so that they can select the right construction-specific platform to manipulate their, or to manage their digital BIM, or compare against it. You know, fill in the blank. There’s a lot of ways to go with it. As Nader pointed out, it’s a pretty big industry.

John Koetsier: Nader, maybe contextualize this for a little bit. We’ve seen attempts to involve drones and robots in construction. Not a lot of huge success stories yet, because it’s such a challenging environment, and because it’s always changing. This is actually a little bit interesting, because the data that this provides could be input into sort of a generalized knowledge space for drones that are actually helping in the future, or robots that are helping in the future, with construction, as, you know, this, “Expect this. Know this. Here’s your path. This is the next step.” Those sorts of things. Contextualize this for us. What does this mean?

Nader Elm: Well, I think the industry as a whole has been very excited by the new technologies coming to help digitize, and I think, if we focus on the digitalization, the building the digital twin, reality captures Ben spoke about, they’re all invested in that.

I think the key challenge, to your point, has been the environment, the complexity of it. It’s unstructured, it’s dynamic. And I think a lot of firms have been struggling with that. And that’s where the autonomy becomes, you know, comes to the fore, because you do need a very, very high, higher-order level of intelligence on board the robots for it to be able to navigate safely, and completely, through a project, to capture the data which the customers ultimately will find useful.

Up until now, I think what we will have found is that because robotics more generally has been kind of going up the maturity curve, they’ve required either GPS, or they’ve required preparations and infrastructure in the environment to help the robot. But now we’re getting to the point where the robot is intelligent enough not to require any of that. And the project manager, the surveyor, can very simply drop the robot in place anywhere, and the robot can execute the mission without any oversight or intervention by the operator. And I think that’s the exciting point we find ourselves in. We’re at the beginnings of the transformation of an industry.

Ben Williams: Yeah.

John Koetsier: Super interesting. And given that stat that you mentioned earlier, almost 40% of the $8 trillion spent annually on construction globally is waste. There’s a lot of room for improvement on that. Ben, thank you so much. Nader, thank you so much. Do appreciate the time.

Nader Elm: Thanks so much, John. Appreciate the opportunity.

TechFirst is about smart matter … drones, AI, robots, and other cutting-edge tech

Made it all the way down here? Wow!

The TechFirst with John Koetsier podcast is about tech that is changing the world, including wearable tech, and innovators who are shaping the future. Guests include former Apple CEO John Scully. The head of Facebook gaming. Amazon’s head of robotics. GitHub’s CTO. Twitter’s chief information security officer, and much more. Scientists inventing smart contact lenses. Startup entrepreneurs. Google executives. Former Microsoft CTO Nathan Myhrvold. And much, much more.

Consider supporting TechFirst by becoming a $SMRT stakeholder, connect to my YouTube channel, and subscribe on your podcast platform of choice: