Balancing with Drones: Using UAVs for Site Balancing and Haul-Road Design

ANDY THOMA: So my name is Andy Thoma. I work for Grunley Construction and I'm going to be talking today about a project that we started to use as a general contractor on a project. And the attempt was to utilize a UAV or a drone in a different sort of way. What just happened right there? Not what was supposed to happen. Let's try that again.

Anyway, the class summary kind of started like this. What we were looking to do was get along with how can we use a drone on a construction site and then, more importantly, what kind of data can we get from it? How useful is it? How accurate can it be? And then really the goal was to figure out a way can we control? Can we validate? Can we actually do something that will give us some value from this tool? And that's really what ended up creating this experiment and then, more importantly, came about creating this topic that I wanted to show to you today. So we were doing a case study out of it.

The goal in the learning objectives is I really want to talk about how I did this. So in other words, how I'm going to present this is really talk about the project, the challenges on that project, why we chose to use a UAV on the project, some of the objectives we were trying to do with the technology. And then really, I want to get into what we did, how we did it, and then how did the results look at the end of the day.

So the first thing that we got in to-- is this not wanting to cooperate with me-- there we go. So the first thing I want to get into is the project and why we use the drones, and really what our challenges and objectives are. First, kind of need to pay the bills. A little bit about Grunley, but also, more importantly, for you understanding why we chose to use a UAV in the first place. We're a general contractor in the DC area. We do a lot of work for institutions, for campuses, for the federal government, for the GSA. We were the ones that restored the Washington Monument after the earthquake. Really what we do is our core experience or what we were probably the most specialized in is this idea of phased renovation. So in other words, people are occupied in this space while we're doing the renovation.

What's interesting about that is it creates two problems-- it's very schedule dependent, because I've got to worry about moving things around, but it also deals really well with what we were trying to do with the drone in that we were trying to figure out a way to be able to phase earthwork around people and a campus all at the same time. And having that element of what we do a lot of, which is this occupied renovation, really kind of drove one of the reasons why we wanted to figure out what we were doing.

So anyway, the project that we were working on, it's 182 acres. Now, I'll preface this really quickly. When I started this presentation, I actually worked for a different company, a general contractor in the DC area, and I have this same concept being utilized on that project. And it's kind of right now is a bin buyer's market, I guess you want to say, in the DC area and I got recruited. And these guys had some really cool projects so I moved over to Grunley and I tried to do the same thing there.

With that, what we have is we have a 182 acre campus. Now what's interesting about this campus is it's actually a historic landmark. The original buildings are from about 1850. It's a phased renovation. The current renovations that we're working on is about a 4.5 million square feet. Budget wise, it's about $300 million of a $4 billion phasing plan of the entire thing. The client themselves, they have a lot of interest in using technology and a lot of interest in using these solutions. So when I proposed doing this with them, they're all on board with wanting to do this.

The other nice thing about this campus is it's got about in total-- if you count all the little small buildings because they have large buildings and they have like 2,000 to 3,000 square foot buildings-- there's about 22 buildings on this entire campus. And, obviously, from 1855, they also have a lot of infrastructure in place, have a lot of underground utilities. I had a lot of things in the ground. So what's the project? I can't actually tell you what the project is, but I can say it's a GSA project and it created a bit of a problem. Not only can't I tell you what the project is specifically, I'm not allowed to actually show you any images. So a lot of the images of what I'm showing you is me doing the same exact thing I did on this project, but I actually did them in a local park to try to give you some ideas of what we were doing in that process.

What's intriguing is probably one of the reasons why they really liked me doing this on this project is the most I can tell you about the client is they own the largest number of drones in the US government. I don't know if that helps. Any rate, so some of the things that motivated me to use this as a solution was some of the experience I already had with the UAV. I created a presentation last year with a good friend of mine named Peter Marchese from Microdesk there, an exhibitor downstairs, but anyways, he and I went through and we were like, OK, what could we do with the drone?

So what we did was we went through and tried to figure out what does it take to fly a drone on a construction site? What can you do to make it reliable and continue to capture the data? And then as we started capturing the data, one of the things that kept popping up in the heads of the people that were showing it, my superintendents mostly, is that I would show them the images and they're like, great, this is a wonderful way for me control the site, but you're telling me you can do volumes, how accurate is that volume?

So that project was really trying to validate that data. And what was interesting about that is we ended up getting about a 5% to 10% error and it was on the low side. So when we were doing fill-- so most of our superintendents were fine with that number. I'm going to kind of go back in further this experiment that we did with this one on this project, also. So it's kind of this is almost a continuation of what I did last year. So definitely I have this one out on AU's website so if you're interested in looking at that one, check it out.

So what we needed to do when we were using a UAV is we needed to be able to design haul roads. So this is an existing campus that has a lot of buildings on it. And more importantly, we're doing a lot of demolition on the campus to do restorations. So we're doing one of those supported demos where you support the exterior shell-- there was actually a picture of it earlier-- and then you hollow out the inside and then you rebuild it from there. Plus, I had some other buildings that had to go in there so I was removing a lot of debris.

The other problem with this was there was a lot of infrastructure that had to go in. So I had to dig in to put in a lot of new sewers and a lot of underground utilities. In doing that, that wouldn't have been normally a big problem except this was actually an occupied site. So in other words, I had actually people working on this site that were not part of our project and we had to provide access to other buildings on this project. So what we had to do is design these roads so they couldn't impede access, egress, and anything on this particular project to allow us to continue to work without impacting their work, too, at the same time.

To add more excitement to it was there was originally some ornamental gardens. They were going to be restored, but that meant I had a lot of trees to protect and a lot of landscaping to protect. And with that, I also had a lot of landscaping that was going to be put in. So all of this being phased, of course, means I have to worry about what time of year I can actually plant and what time of year I can actually put these in. So me figuring out how to put this road in here means that the road is continuously changing and I'm going to simultaneously move things around.

So what I have to make sure is that as I'm doing all of this, how can we control it? In other words, how can we make sure that it is being put in the right spot, that the progress is being done correctly. And then also, it's constantly shifting so that means I need a way of surveying the data, being able to create a digital terrain model to be able to understand how the materials are going to move. To make it all even more complicated, the client really had no interest in buying fill or bringing fill into the site namely for cost. And then also, I had limited access, in other words, where this was at. It just kind of goes through a lot of city, it comes through a lot of area where bringing in fill is just not very practical. So a lot of the material I needed to find had to be on site in some way or another.

So what we came up with was this idea of surveying the area with the UAV to be able to understand what I had there and what could be moved and how we can move it. So with that, we kind of came and did some challenges. First challenge was, well, we had some rule changes. When I first actually proposed this topic, we were still not under the rules to allow us to fly and that's just because when you look at the timing of how Autodesk University goes. But what was also interesting was that I knew how I wanted to do this and I actually had some data and I wanted to actually push it through, I just didn't know when I was going to get the time to do it. So one of the things that we had to look at was how the rules for FAA were going to change.

Some of those things that changed was the 107 ruling for UAVs or unmanned aircraft. That really changed a lot and allowed us commercially to be able to fly again, but some of the things that we also had to worry about as a challenge was location sensitivity. In other words, I had specific areas where we're allowed to fly or not. In DC region, there is a gigantic no fly zone. I was right on the border of that thing. Literally, if you flew too far on the property itself, I'd hit the no fly zone and I was using DGI so the thing would actually land. It's kind of an exciting time when it happens because you don't realize it until you hit it and then things come crashing into the ground. But it's also a really good point of location sensitivity of making an understanding of where you're going to fly that was actually a big part of what we were trying to do, also.

Client requirements-- like I've said earlier, I'm not allowed to actually show you any imagery from this even, the DTM which is strange because you can look at all this up on Google Maps and it's not really anything different, but it's a good example of a challenge in that the client really needs to understand what you're doing and, more importantly, the client really needs to have some buy in on all of this. And then, the final one that's a really big issue is-- oh, that was my picture for that-- is an active construction site, and being occupied, I have to be really cognisant of doing that the entire time.

So when we were working on this, the other thing that I wanted to do was how can we validate what we're doing, the data, how can we create a system to prove what we're doing is correct? Because working as a general contractor, working with superintendents, the first thing everyone will ask me as a superintendent or PM is great, this is a great tool, but does it work? Does it give me anything I can use or trust? And that's one of the things we had to do is we had to figure out a way to appropriately validate the data so that we can get buy in that the tool that we're actually using made sense and it had a lot of value to them.

The other thing that we had as a challenge was we needed to figure a way to acquire the data appropriately, time wise. In other words, I had a lot of work going on, I had a lot of scheduling going on, I had to make sure that no matter what I was doing I was meeting the needs of the project and getting the data as soon as possible. So we came up with an idea. An idea led us to these objectives.

So what I wanted to do was I wanted to have a way to be able to repeat it at a moment's notice. I wanted to be able to go out and use it again and again. I wanted to gather as much information as possible, as quickly as possible, and I wanted to have that for a real time control. In other words, I wanted to be able to use this in real time to be able to create the earthwork data that I would need to be able to feed to an earthwork sub, but also use it as a general contractor's tool to be able to figure out is the earthwork sub delivering on what they're telling me they are doing. So I wanted to do not only real time control of grading, but real time quality control of that grading. So this is basically what we did on this project. And I'm going to explain this two different ways. I'm going to go through and actually talk about what we did, and then I'm going to get into how we did this.

The first thing we did was we flew a mission, we took pictures. This was done multiple times, different times of day, and it was done at different angles with different images. The goal was to build as a robust a DTM as possible with as much information as I could capture as possible. Once we did that, we created a DTM from images and then we geolocated that image. I was actually using a program called Pix4D. There's a couple of reasons I've used Pix4D more than reCap. I'm trying to be a little bit software agnostic because of Autodesk University, but Pix4D is a very easy tool to use. The nice thing, also, we're using a DGI Phantom 3 Professional and you can control everything through the app through Pix4D. So being able to do all of this is actually very easy and pretty straightforward. So that's actually why I chose that tool.

That said, you could do all of this same thing with reCap. One of the reasons I didn't choose to do reCap on this was by the time I got the 107 certification, I had to go. So I didn't want to learn a new system. I already knew Pix4D so I went and did it. Hopefully, now, that we have all of that and we have a little more time, I can experiment with reCap to see if I can get any better results, but at this moment, I'm going to be talking mostly about Pix4D on this one.

Once you use the tool to be able-- the UAV, you capture the data and then you convert that data into DTM and you locate it you take it in to Civil 3D as a DTM. Once you have it in a DTM, one of the big things that you can do with it at this point is take this and put all the other pieces of the puzzle together. In other words, you can reference in a BIM Model. In this case, what we were using is the utility models, and then also the Civil 3D models of all of our underground materials, and we're putting them all together under the same survey control in the same model so that we can understand how they would all correlate to each other.

Once we went through and did all of that, we could start the design process for what we wanted to do with this road. So we would just straightforward use Civil 3D's Corridor surface tools and then we would create a surface volume from that so we could compare the two. Now to be able to balance it or to get it into a state where I'm not hauling in or removing material, I would then manipulate the vertical profile inside of Civil 3D and then come down to a zero by moving things up and down. In doing that, I also had to calculate in materials that I knew I already had on site. I had some stockpiles on site that I can move around and put into it, so I didn't have to be necessarily 100% at zero in my profile or my volume surfaces as long as I was within that material that I had access on the site. That said, sometimes I didn't have enough material and I actually had to cut in a few areas and bring it out and move it around from there.

Once we got that data from Civil 3D and once we were able to create those volume surfaces, we were able to take that Civil 3D surface and create survey points, export it out to our earthwork contractor. The earthwork contractor would then take it in and use their GPS grading equipment to be able to actually go through and grade the solution for us that we had worked with. To get all of this to actually happen, we had to work together a lot. In other words, I had to understand how their system was working. I also had to understand how they were doing GPS control, what they were using for their control points so that when I'd send them data, it would make sense with what they were doing. So we had to do a lot of coordinating our survey control and all of our survey points throughout on.

So how did we actually do all of this in the process? First, let me get into one of the things that's probably the most important out of this and probably going to be one of the more asked questions is if you're going to do any of this mapping, you're going to need to actually start flying a drone. To fly a drone as a commercial entity, you have to meet these rules that have been put out by the 107 certification or the part 107 ruling that the FAA did. This actually changed on us. They implemented a no fly, what was it, last Christmas? And then this came out for review, I want to say March, April. And then, actually, didn't get instated until about July.

What you're supposed to do is you're supposed to register your drone. Register your drone, you actually get a tail number just like a regular aircraft. Once you have that tail number, you need to then, as a commercial flyer-- and this is different from a commercial flyer versus a hobby flyer-- you need to get this 107 pilot certification. Really what it is, is it's an exam that you take that is meant to say that you understand the rules of the road. Do you understand how to operate the drone? And more importantly, that you know how to operate it within the parameters they ask you to do. Some of those parameters are things like you've got to keep it in your line of sight, you have 400 feet of elevation. It's really not a hard test, especially if you've flown a bunch of times, but it is something that you need to study for and you actually have to take.

Probably the worst part is you have to schedule it, you have to go to an FAA approved testing center. I'm fortunate that I have a few near me where I was at, but it's one of those things you actually have to organize and do. If I would give any recommendations on taking this class or taking the certification, fly a bunch first, get experience, understand how it works, and then try to figure out how the whole thing works together. I'm going to get the handout because of my flight getting in literally this afternoon. I didn't get it loaded up in there, but I put a whole bunch of links on my handout that go through about a lot of the information that's for this class. I've got some links to some practice exams. They have a really great website for FAA explaining everything about this and trying to help you go through understanding what you need to do to take the certification.

Site mapping itself-- so when you start flying on a project then you need to understand a few different things. First off, area of interest-- you need to understand what you're trying to take pictures of. If you're going to go through and actually go through and map the surface or create these images from for surface, it's actually pretty straightforward to take them, but the problem with it is you can get some situations where you don't get all the information you're looking for. In particular, if you notice in this hole, there's a bunch of trees there. This is actually a local park near me, but it actually simulated a problem I hand which was I had a large area of trees on my property that I'm flying and those trees block because the ground cover any information that can be gathered inside of it.

This is kind of an example of you need to know the area of interest where you're trying to gather data. In my case, it was the same basic example. I needed an area up here and an area down here. I was OK with the influence of those trees because it was kind of outside of what my surface needed to provide me. That also led to another thing that you have to think about when you're actually trying to understand the site is hazards. You don't want to run into one of those trees if you're trying to map around it.

When you're taking these images, one of the things that I've discovered in doing these maps with imagery is being able to do as many images possible. Flying more missions is very useful, but taking images at differing angles is actually better. Pix4D actually came up with this really interesting software change recently that does this. And what it does is it flies a mission and it flies it in a pattern where it flies at 90 degree angles of each other. So what it's doing is it's literally taking an image from one direction and it takes an image from the next direction.

What I find is even if you're doing a manual flight or preprogrammed flight, taking pictures at a fixed angle and flying that pattern and then rotating the whole thing again, like 90 degrees, 45 degrees, and flying it again, it's just going to build a more robust surface because what it's going to do is it's going to help you avoid that situation like the big hole in the trees, because it's going to be able to see around more things, get you more information, and give you more data. If I can liken it to anything, I'd say think of a laser scanner and how the laser scanner works. Whatever is in the way, it's not going to be able to see behind it, so the more images you can take, the more images you can get, the better the experience you're going to have with those pictures.

One thing that's really nice about the way the Pix4D or reCap works is you can feed in as many images as you want. As long as you figure out a way to geolocated them, they'll all kind of mesh together. So once again, if you ever played with a laser scanner or a point cloud, once you start registering and once you start matching, you're just going to build a more robust image and you can actually get everything together.

The other thing I'd recommend when you are flying these things in taking these images, that you might have to do some experimentation. In other words, you have to fly at different elevations, higher or lower, obviously, you want to miss trees on your site. You might want to try different times of day, all of it to ensure that you can try to get as many clean shots. You might find one time on day, one elevation, and one angle working better than another, but what I've noticed is it doesn't necessarily flow from job to job because a lot of what you get from what you're trying to take pictures, whether there's buildings in the way, whether there's trees in the way, that might change the height, the elevation, the angles. Plus, I've noticed time of day, how much sun will wash out shadows and then reflection of the ground if you got a lot of bright concrete reflecting up onto you, it will affect your quality. So really, the advice I have is to fly a bunch of times and try to see what else you can see, try to recreate the experiment and see where else you can get a better image from.

The other thing that I've noticed is when you start trying to do the coordination of all these images, the easiest thing to do is to have a target on the ground. Last year, I was espousing to use natural objects on the ground, whether it's a building or edged curving or something that you can identify from a picture as an object. It will give you something that you can use as a geolocation target.

Kind of changed that a little bit in that I'm finding that creating a real target is your best bet. And reason why is you can put that real target into Civil 3D and you can put that into a survey instrument very easily. And then more importantly, you can do a target something like that. What I did was I actually just printed something like that on an E sized sheet, put it on my target, laid it on the ground, weighed it down. And as I would fly, I'd see that. I'd put multiple down. I'm not really doing anything radically different than what normal aerial survey companies actually do, the only thing that I'm doing a little bit different with this is I'm doing it through a drone instead of an actual aircraft. But this technique I've noticed was the best way for me to line up the surfaces because I'd have these targets.

Other thing I found is using three so I could do a triangulate it and to be able to put all of those. As long as I had one on a known survey marker, I was able to do that as a baseline. But that should also probably help you understand one other thing which is if you're going to fly a mission, make sure your mission takes a picture and actually captures that target, because if you don't capture that target as one of your images, it's not going to help you. OK.

So now that you've flown the mission, you've gathered the data, you've taken a bunch of pictures, and you're going to go through this concept of photogrammetry. Photogrammetry is the idea that it's going to compile all those images together. One of the best things is all these tools, whether it's reCap or Pix4D, they pretty much do it all automatically for you. You really don't have to think about it, you just have to click a couple of buttons and it starts processing the data. It's sort of amazing when you think about it, but it's also going with the problems of it only uses-- the better information you get, the better you're going to get result wise. Crap information in, then you're going to get crap information out.

One of the things that I found was noise was actually a huge problem. Noise is anything that you don't want to capture, but the images caught it anyway. That could be sides of buildings, walls, things that you didn't want part of your surface, but it was in the way so it wasn't able to resolve a close image around it. Last year in my experiment, I was trying to do volumes of materials and I could do stockpiles, but my problem was with going up to where stockpile was up near a building. What I wanted to do was OK, I've got to backfill. How much material could I backfill in there? It would be great, but I would start getting the building in the way and the building would start trying to triangulate to it and I would get way too much information. So what I had to go through was actually sort some of that information out.

What's interesting is all the programs have gotten better since last year in that they deal with noise a lot better. So if you're looking at this image, the little dots on the left image are actually the trees that are shown on the right. What's interesting about it is you can see how few images are taken because what these are, are the individual images that actually produced that in the model. So it's relatively, actually, straightforward to eliminate, you just say hey, cut a plane and remove all that information out of there.

But the other thing that's very interesting is if you capture enough angles, it's kind of hard to see in this image, but I have a flat image of this soccer field all the way to that edge, which is pretty much what it does and then it drops down from there. So if you capture enough images from different angles from flying that different pattern, you'll actually get a good enough image map to be able to create a pretty accurate surface from that. The only way you're going to figure that out, though, is to experiment. That will mean doing things like different angles on your images, taking it that those different-- like 90 degrees, the different flight patterns. But you also might want to play with your path. In other words, maybe you want to instead of having the camera pointing towards the inside of this field, you want it to point towards the outside towards the field itself. That may not be necessarily something that you can't do without manual control. In other words, you're going to have to flip the drone and you'll have to turn around and fly in a different direction.

The flight elevation will also definitely affect it because what you're doing is you're changing the angle of how the camera's perspective is working with that. My only real advice on that is you have to experiment, you have to keep trying it, and see what resolves the best for you for the time in there. What I found, generally, is when I went into an area to fly, I'd go through four batteries. I'd go through all four flying two missions per and what I would do is I would just literally just re-fly the same flight, I would just do it from a different way. I'd change the angle of the camera, I would change the elevation of the flight, and then I would track it all.

Now if you actually take the 107, one of the things about the 107 from FAA is you have to have a log book. In that log book, you're supposed to talk about the flight and what you're doing. I added in there all the specifics of my missions. In other words, I would add in what the conditions were, where part of what you're supposed to list, but I'd say conditions of what was the light, was full bright? Was it sunny? Was overcast? And then, I'd add in other information like, what was the angle of the camera I flew? When I did the pattern of the flying, what was the area that I actually covered? So if you list all that information, you can go through, you run through your experiment, dump all that data, and then try it again, you get to see what works and what doesn't work pretty quick.

So with operations, getting all of this to work required a lot of schedule coordination. Some of it is pretty much logistics wise. In other words, some people don't want you to fly when there's active construction going on. The client may not want you to fly certain times of day, too. So you have to actually work through that with everyone to understand when to fly. So what that might mean is you're flying early morning or after work activities are done.

The other big key is if you're going to use this tool as a tracking tool, as a QA QC tool, you also need to be able to do it in a timely manner so that my earthwork sub is working every day, they're going to charge me how much material is every day. If I want to capture that data of what they're doing, I probably should capture the data after they're done with their activities for the day. So being very intimately knowledgeable about the schedule will help you figure out when you're going to put all those together.

The other big one was survey control. It took a little bit of work to get everyone to get everything together, namely because the architect doesn't work and sort of they control, they work on a project coordinate. And I've got all the subs, meaning my building mechanical subs, are all telling me, oh, we're just going to work on projects specific for each building. When I have big campus, that means I'm literally going to have like 22 project controls for every building, it gets a little bit out of control.

So as the BIM manager for the project, what you have to do is you actually have to create a system where everyone is going to actually control the data with the survey that is done on site. The key is you want to make sure everyone's using the same coordinate system, because I actually did run into this that the earthworks sub was using real world and we were all using an assumed project control and we didn't realize that until we tried to put them together, and one's way out here and one's way over here. So this is one of those things that you want to definitely coordinate with everyone to make sure you have the survey controls set up. In doing the drone, the nice thing about all the software is whether Pix4D or reCap, they actually will work in real world coordinate systems, so it is nice because you can actually put the data into a real world point, but that might mean you have to do some translation to some of those underground utilities from some of the other people that are not working with it.

As a QC tool, being able to continuing to capture the data becomes huge. The more data you have, the better of an idea you're going to have of the quality of your image data. So in other words, you get to continuously capture those. When you're actually using this as a comparison tool, all you're really doing is you're taking a surface and you're comparing it to another surface. Inside of Civil 3D, you have the Compare Surface feature of the volume services and you can actually go through and actually compare the existing to proposed. And what I would do was I'd fly, create a surface, then I'd compare it to existing, and then I would consider it the proposed, and I'd know how far we went and then how far we had to go to.

So with that, I went back to my idea of my last year's experiment and that was, how can we validate this? So I'm going to step back for a second and talk about my last year's presentation and also, the concept of what we did with this. So when we did the last year's presentation, one of my questions was by my supers is, how accurate is what we are doing? So we were like, how can we prove this thing actually works? So what I wanted to do is I recreated the method that we were using outside on a known object that I had a volume of, a box, because I could calculate that box volume and I could figure out what my size was.

What I ended up discovering as I was getting about a 5% to 10% error on the low side. In other words, I was low-- my volumes are smaller than what the box was in reality, and really all I was doing was I took a drone and I mapped that box. When I did some research into this, what I was finding was the validation by published studies and some of those-- some of the things I found on those published studies was that they were coming up with a 0.1% to 3% error. So they were getting a much more accurate number than I was. And I didn't understand why until I started doing a little bit better research.

One of the things they found out was their image quality was much better. They were using a much higher end camera. They were using a higher end drone. And they were getting much better imageries than I was. The other big thing that they had was line of sight. They were getting a much better line of sight so they were getting a clear area. In other words, they were doing stockpiles with nothing around them and they had a nice clean image path so they could get a great shot.

So those two things were some of the things that I took in account to play with this year. First thing I did was we got a better drone and actually I bought the Phantom 3 at the time, the Professional, which had the much better camera than what I was flying on my Phantom 2 per the reCap guys' suggestion when I was at AU last year. And then I focused on my line of sight during my photos. So what I did this time was I took what the earthworks subs were charging me. In other words, I wanted to know how much material they moved and they would tell me, and I used that volume to try to figure out what my difference was.

When they would tell me how much they moved that day, I would take my surface and compare it to the existing, and then I would do the comparison between the two. What I was finding was a 3% to 5% error on that. And generally speaking, I was on the low side. In other words, they had a little bit higher number than I was. I was always taking the average on the low amount of volume in there. But would it kind of lead me to believe is I'm in the same range as I was before. And it also helped me believe that some of the things that you can't control, you always have to deal with this little bit of error in here on this.

Now, I worried about this number a lot and then when I talked to my subs, they're like, oh, yeah-- or my superintendents, they were all like, oh, yeah, 3% to 5% is not a big deal because at the end of the day, that could be settling, it could be a volume difference, they're not counting trucks right, they're giving me a bad volume, I'm not flying at the right time, I'm not getting enough air. They're like, that's not that big of a difference. But it kind of led me to a little bit of a conclusion that you have to take everything as an estimate. You have to have a realistic expectation of what your results are going to be.

I'm thinking that my next bet would be probably doing a fixed wing drone instead of a quadcopter drone and flying more patterns and having it do more during the day. But in the end of the day, what I was noticing that I'm not getting as much air as I was expecting, and I was actually feeling pretty good about the results that I was only able to get a 3% to 5% difference between what these guys were calculating and giving to me as a volume through what they were actually moving.

So with that, I'm kind of running out of time, but I wanted to leave it open to any questions you might have. I definitely apologize for my voice. It's like dying on me.

AUDIENCE: How many-- when you're using your targets, how do you decide how many you need to use in the size of an area that you have got to fly over?

ANDY THOMA: Generally, I try to do a minimum of three targets. And how I'd want to establish it was I'd try to be on a control and then I'd stake the other two. I would go with the theory is the same idea is setting up a total station. If I have the one known and the two knowns, I could calculate the angle between them. I could probably have done more for the larger areas, but what I tended to have a problem was with battery time. I could only capture a certain amount of area so if I could get the three in the area I wanted to capture, I felt pretty good about it. I could probably set a fourth target and I could have gone maybe a little bit larger, but the nice thing about my site is it's kind of square so it's actually easy for me to cut up.

The last project, last year's project, that one wasn't really a square, it was like maybe a strange trapezoid the way it's shaped. So if I would probably have a more obtuse shape that I couldn't necessarily do, I would set up definitely more. Maybe it's because of the GC in me and I think in laser scanning terms, sometimes, is when I'm doing a laser scan, the more targets you set up, the more overlap you can create, and that's kind of the same idea. The more overlap is really what I'm looking to do. So a lot of times, too, when I would do the target-- and this is probably something I should do a slide on-- but I would set a target here, a target here, a target here. And then when I move over, I'd keep one of those targets and then move over to there. But I always try to keep one on a known, a hub that was in the ground.