The robot is the brainchild of Mountain View, Calif.-based Dusty Robotics, but its use on this project is the result of an unusual collaboration between a designer, a roboticist and contractor all willing to try out something new.

“The way Dusty Robotics fits in, is we can take an up-front investment in BIM that is much more accurate and precise [than older methods], and we can bring it out onto the construction site and deck without people,” explains Tessa Lau, founder and CEO of Dusty Robotics. “When people do layout they make mistakes, and those mistakes affect the entire cost of the project.”

But the key element remains the design model. Dusty Robotics’ little robot has been deployed on several jobsites with varying layout demands on concrete or wood floor slabs, but at Project One in Oakland, the tech firm is closely collaborating with designer Planit and developer-general contractor Gurnet Point. The detailed models the robot prints onto the mass-plywood panel floors are the result of Planit’s unusually detailed approach to BIM design, going beyond the usual standards of a digital model into something much more precise. Planit seeks to bridge the gaps between design intent, shop drawings and onsite plans within a single model.

“Right now, subcontractors have the burden and obligation to create shop drawings that make up for any deficiencies in design documents,” says Frank Haase, founder of Planit, and also vice president of operations at Suffolk Construction. “But the way that much construction is done today, none of the shop drawings are entirely correct. We need to also have manual detailing and construction tolerances and design intent—then we will have that single source of truth.”

Project One became a pilot of sorts for the integration of Planit’s design process, Dusty’s robot and Gurnet Point’s foray into mass-timber construction. Mass-plywood panels are cut by a computer numerical control (CNC) machine from designs taken directly from Planit’s model, and Dusty’s robot marks up each floor with layout markings from that same model. Crews need only assemble walls where the robot has indicated.

“Once we panelized with the CNC, I was utilizing the Planit drawings to CNC the wall segments in,” says Scott MacLellan, principal at Gurnet Point. But CNC time is expensive, and the Planit designs called for many precise cuts in the panels. “Dusty came along and hit that sweet spot. We can have some critical dimensions cut into panels, then have Dusty do the detail work. It also helps with field modifications and compensating for any discrepancies we find.”

The result of this BIM-to-prefab-to-robotic layout process enabled the project team to top out the four-story building in only 24 working days.

Gurnet Point is testing several other new products and ideas at once on the demonstration project. For example, cold-formed steel (CFS) moment frames from MiTek have been incorporated with the mass-plywood panels—another first, says the team. But it was the ease with which Dusty’s layout robot was able to transpose the design data from Planit onto the site floor panels that turned a few heads at the project, and may signal that the disconnect between designers and field crews can be overcome. “Looking at quality control issues, one example is we have the robot print where the studs go,” says Lau. “That’s something that doesn’t cost us much time, but it’s cost prohibitive for people to do. It’s because we have these really nice models that we can translate it to the field.”

Related Article Updated Robot Dog Speeds Site Inspections

Changing Industry Attitudes

Project One is one of many examples of how the model and site are increasingly becoming interconnected. But the industry can be slow to adopt change. “Most people still say, ‘Give me my sheet, I want to do it the way I’ve done it forever,’ ” says Rich Humphrey, vice president of construction at Bentley Systems. The construction software giant has invested heavily in promoting greater use of what it refers to as digital twins, where the model is maintained alongside the project, allowing for a feedback loop of data to keep things in sync. This eventually leads to a robust record of the project and its construction, to be handed over to the owner at completion. “It’s not really a digital twin unless it’s connected to the field—the whole idea is you have these live feeds of data coming back from the field,” says Humphrey.

“Now instead we can look at the model, which includes all the work plans for that week, all the things they talk about in that hourlong daily meeting—nothing lost in translation, all the details have been put into place,” he says. Tightening that loop has been critical to getting the most out of digital twins, adds Humphrey. He likens it to the benefits from the way logistical planning revolutionized the manufacturing sector. “I do believe the processes and the mindset in what was achieved in manufacturing is to a large extent possible in construction, but it won’t happen in just a few years.”

Humphrey sees some hope for greater automation of processes in repeatable construction processes, particularly in the buildings sector. “Digital models let you test before you build,” he says. “Right now these tools are available to everyone, so it’s almost foolish to not do 4D modeling and simulation of a project.”

“That convergence of layout and reality capture—when someone is going to document an as-built—really comes in handy with projects where you have a tough site with existing conditions,” says Jon Fingland, general manager of collaborative solutions at Trimble.

“That roundtrip process, where you capture points and use it to coordinate the model in the office, takes the model out of the office and back. It’s all about that closed loop,” he says. But Fingland notes that technology alone isn’t enough. “To really get these gains, we have to change the process and enable it with technology, rather than just enable the old process,” he explains.

Challenging Site Drives New Approaches

At the expansion and renovation of the Phipps Plaza mall in Atlanta’s Buckhead neighborhood, the Beck Group adopted such a BIM-to-field loop. Working to partially demolish and expand the mall while it remained open with no as-builts and limited existing plans, the contractor’s BIM manager, Tim Riefenberg, had to come up with a new approach. While he had experience with keeping the digital model in sync with the site layout, Phipps Plaza forced Riefenberg to get creative.

“From a technology standpoint, we could not do it with a traditional design-to-contractor workflow,” he says. “We have all of the team on site, so the structural engineer is able to facilitate a revised design around a challenging site.”

The design-build effort is the largest such project the Beck Group has tackled so far, and building on an active site with poor extant documentation has required constant updates to the design model as unexpected foundations and utilities become unearthed.

“We would do as-builts daily that would go to the design team across the street,” says Riefenberg. “We kept updating backgrounds to utilities we found, structures we found.” With full demolition not an option, Riefenberg’s team was able to salvage 50% of the existing pile caps, but there were clashes as his field team kept bringing back new suggestions to the design team on changes day after day. “At first when I started on the project, [the design team] was upset that I was in their Revit model. I was giving them new locations for things,” recalls Riefenberg. But initial resistance soon gave way to a tight feedback loop, as the field team would document points in the field with a total station and feed them directly back to the designers and civil engineer Eberly & Associates. Problems that could have taken days or weeks to resolve were addressed and fixed in hours.

One particular design clash almost upended the centerpiece of the project, a 150-room Nobu Hotel planned for one corner of the mall site. Eberly had a rough idea of where an existing storm vault was in the area. But when Riefenberg’s field team put in some preliminary footers to check the subsite conditions, they discovered an active sewer line not reflected in the drawings, and had to act fast. “The whole design was baked in early with the [formwork contractor], but when we were digging, we found this active line. We had to change the whole structural design.”

While there was some initial panic, the team was able to quickly adjust the design thanks to Riefenberg’s maintenance of an updated design model with the latest data from the site. The effort averted major cost overruns. “We had one older super on the job, and he would always joke that was ‘Tim’s Million-Dollar Corner,’ ” recalls Riefenberg. But the superintendent’s teasing reflected a realization about the VDC process they were using. “It’s this whole digital evolution. The quicker we put new data into the model and get it out to the field, the sooner we can adjust.” says Riefenberg. “The models are reactive to what happens. We were creating a model and design as we were finding things out in the field. It’s where the industry has to go.”

Riefenberg also found time to experiment with a drone on site to get stake points set faster. He set up a DJI Matrice 210 with a real-time kinematic module and loaded it with the plans for the next concrete placement. Riefenberg had been disappointed with GPS-based drone layout systems, but found surprising accuracy with this ad-hoc setup. “It was an automated pass: we’d put the drone up and give over the info, and that 3D file already had the stakes [mapped out in a design] plane,” he says. “I was at the computer, overlaying it in AutoCAD, seeing that it all fits.” A laser pointer attached to the bottom of the drone let a field team test the accuracy of the drone’s recommend stake placements. According to Riefenberg, the drone could mark locations within 2 in. of accuracy at an altitude of 350 ft, and within 1⁄4-in. accuracy at 50 ft. It was just a test, but the results were promising. “In the future, we could put plywood down before every pour, have the drone hover and mark where every sleeve goes,” says Riefenberg. “It would help with every job.”

While work on the Phipps Plaza has been slowed or halted in the last month due to the COVID-19 pandemic, Riefenberg says the time saved in the schedule from avoiding rework and protracted fights between the design and field teams have kept the Phipps Plaza project on target, currently at 60% completion.

Prefab’s Time To Shine

Gaining a stronger understanding of the real-time developments on the jobsite and more precision in how components are installed via the BIM-to-field loop may also be the missing piece of the puzzle for wider adoption of prefabricated building systems.

PT Blink, an engineering firm and technology integrator based in Sydney, developed a project-delivery system for buildings using prefabricated elements based on modular, post-tensioned steel panels. “We build a backbone—a chassis for the building that is geometrically precise,” says Murray Ellen, structural engineer and founder of PT Blink. By breaking the floor segments into prefabricated structural trays, the components can be factory produced and assembled to fit a wide range of designs, he says.

 “For us, the BIM model goes to the field—not only just the parts and the backbone and chassis—but in local environment,” he says. “We go to local partners that manufacture the panels, kitchens, staircases and windows, and we all work off the same BIM model. We call it design, manufacturing and integration.” All manufactured components are scanned and loaded back into the BIM model for checking before assembly, with phased assembly planned out in Tekla. Before final assembly on site, sample components are assembled in a dry run at one of PT Blink’s offsite facilities.

PT Blink’s approach of working with local manufacturers to build out the panelized elements offsite has paid dividends on projects in Australia and New Zealand. On one residential tower in Brisbane, crews were able to build six stories in 11 days. “We went for that speed after a developer asked, ‘Why can’t you build fast like China?’ So, we did,” says Ellen. PT Blink is currently working on a U.S. demonstration prefab project in Kansas City, Mo., with Butler Buildings, a major builder of big-box retailers.

 While it has required a great deal of work to coordinate local manufacturers to deliver the prefabricated components for the projects, Ellen says combining the 3D capabilities of Tekla with the laser scans at the factories and the building sites makes the process possible. “Some of these small manufacturers were working in basic parametric design. We got them out of 2D to think in 3D,” he says. “It’s not enough to think in terms of just BIM, you need a model with true 3D vectors  … Then if it will clash, it will clash in the model.”

Small Steps, Faster Schedules

Meanwhile, at Project One in Oakland, the work of the little layout robot continues to pay dividends.

“[Dusty’s robot] printed an entire floor at night, and the next day the plumbers and electricians said, ‘This is amazing, Dusty was in the right spot every single time,’ ” says Planit’s Haas. “Everybody knew it was a robot that printed all the marks on the floor, and every one of marks was dead-on.”

“The important thing is as soon as you use a robot and automate this process, people start thinking and the importance of the drawings­—it becomes more clear to them,” says Haase. “There are so many areas you can leverage this.”