ROBOTS ARE LEARNING TO WORK ON CONSTRUCTION SITES
BY STEVEN H. MILLER
The long-predicted robotic future of construction has begun. The industry has lagged behind others like automotive and aerospace in adopting robotic labor because, until now, most building robots could only operate efficiently in a controlled, consistent environment like a factory. The construction jobsite is neither controlled nor consistent, because the nature of the job is to transform the site continuously.
Four years ago, a brick-laying robot that operated on rails alongside the wall began to address that challenge (see Rise of the Robots, Constructor July/August 2016). Now, advances in machine learning and artificial intelligence (AI) that were developed for technologies such as self-driving cars — especially visual recognition — are enabling a new generation of robots. They can find their way around the site and perform a job with very limited human assistance or supervision.
At World of Concrete earlier this year, there were at least three robotic construction technologies on exhibit, two that tie rebar and one that performs laser scans. For those specific construction tasks, they are potential game changers, and they point the way toward a lot more non-human labor. As contractors seek strategies to cope with the shortage of qualified labor, these machines offer ways to perform boring and/or dangerous tasks without tying up increasingly scarce human workers.
TYING INTO THE FUTURE
“We are in the midst of a new industrial revolution,” says Jeremy Searock, co-founder and president of Advanced Construction Robotics (ACR). “If you zoom back to the industrial revolution of the 1900s, it completely transformed how construction operated. It went from buck-ets and shovels to steam shovels and trucks and power drills. The model was to take the way the industry did the job before and add unlim-ited power through electricity, motors and combustion engines. The model for this revolution is you take the way the industry does the work and add unlimited brain power — robotics and artificial intelligence — and you have a massive increase in productivity, in safety, in schedule stability.”
ACR’s rebar-tying Tybot is a robot that does one specific task. It is a gantry-style assembly that is positioned over a bridge deck or other large concrete slab in-the-making. The rebar is placed by humans, and then Tybot is enabled. It visually identifies right-angle intersections of rebar, positions itself over them, and ties them using a purpose-built bulk tying unit much like a large tie-gun. It makes a map of the location on the fly and figures out which intersections to tie for 100 percent coverage, 50 percent, etc. It does not use input from BIM, there is no software programming needed, and it does not depend on GPS to locate itself.
The gantry can be extended as wide as 100 feet. It can do up to 1,000 ties per hour. If it misses a tie, it is immediately aware of the failure, repositions and tries again, up to three tries. As a broad generalization, Tybot is about as fast as five to eight human workers over the course of a shift.
Tybot, in its first version, has been doing real-world bridge deck projects on a lease basis. Feedback from those jobs was used to develop the second version, which was scheduled to be available this past summer for lease or purchase. For purchasers, Searock says, it requires very little training to operate, mostly for site assembly of the machine (it’s designed to be transported by an F-250 pickup and a trailer), spool-swapping and similar operational assistance and long-term maintenance.
The company will soon offer another robotic option in the rebar process, Ironbot, which will carry bundles of rebar and place them ac-cording to specification. Ironbot is expected to perform pilot jobs this year, with commercial availability in 2021. Searock points out that to-gether, Ironbot and Tybot will fully automate transportation of bundles, placement and tying, “essentially transforming the way in which rebar installation occurs for horizontal surfaces.”
IT’S A BIRD, IT’S A PLANE
Another emerging approach to automated rebar tying is much smaller and lighter. SkyTy, from Skymul, is a flying robot, a quadcopter drone with on-board sensors and a computer, and a tie-gun mounted underneath it. Developed by Georgia Institute of Technology alumni Eohan George, CEO, and Stefan Posey, CTO, SkyTy is designed to fly over the deck area, visually spot the rebar intersections, land on them like a giant insect, and shoot the tie-gun.
A tiny onboard computer, just 200-300 grams, makes all the identifications and decisions the drone needs to perform the job efficiently.
“If you had tried to do this five years ago,” says George, “you would literally have had to mount a PC on board.” The tie-gun can rotate, and the drone is smart enough to position itself in a location where it can perform several ties from one landing. It calculates which intersections to tie for 30 percent, 50 percent or 100 percent coverage and can do up to nine ties per landing.
The human crew only places the rebar and ties the ends to keep it in position. Then, the SkyTy technician defines the work perimeter and activates the drones — he or she can supervise four or five at a time — and the drones start flying and tying. SkyTy can do four to six ties per minute, and the battery can last 300-400 ties. Humans can certainly tie faster if they try, but they slow down over the course of a shift, and SkyTy does not. The company claims that overall, their drone is 1.9 times faster than a human, 13 percent cheaper (approximately, depending on local labor costs) and most importantly, requires 84 percent less human labor.
A CONTRACTOR’S BEST FRIEND?
These single-purpose robots are both designed for one specific environment: a deck covered with rebar. There is a more versatile, mul-ti-purpose robot that has started working all over a construction site. Spot, made by Boston Dynamics, is a dog-like walking robot that han-dles rough terrain, walks up and down stairs and avoids collisions. It can not only be operated remotely, it can also learn a path or “mission” and go there autonomously.
FARO Technologies, the laser scanner manufacturer, is using Spot to mount a scanner. Spot can walk a construction site at night autono-mously, find its way to pre-defined locations and trigger the scanner. It can do as-built scans without getting in anyone’s way, while all the other trades are home asleep.
Scott Diaz, FARO’s director of field application engineers, says that the system was developed not only because of the tediousness of scan-ning, but also for hazardous locations. “We have a public safety division where one of the customer demands is ‘I don’t want to send a human being into a dangerous situation.’ I can send a bot.”
One of the early adopters of this technology is Swinerton, a member of multiple AGC chapters. Scanning has been part of Swinerton’s op-erations since 2014. They have used it extensively in healthcare renovation projects to capture as-built conditions. According to Vincent Marquez, virtual design and construction manager, Swinerton “is always looking for ways to be more automated,” and they believe the Spot/Faro combination has potential for “huge time savings and productivity.”
This level of optimism is particularly impressive because, as Marquez acknowledges, some significant limitations of the automated scan-ning system are still being worked out. Currently, its battery life is short, so it needs a human on hand to replace its power every couple of hours. It also has a limit on how long a path it can learn.
“Our floorplate is big,” explains Marquez. “Automating the path is fairly easy, but Spot is timing out; it’s getting too long of a path.” They are addressing these issues with Boston Dynamics, and Marquez hopes to have the system running smoothly by the end of the year.
The Spot/FARO combination has a broader significance. Spot can be used for a variety of tasks, such as remote observation, including walk-throughs of hazardous areas. If a resourceful contractor gets such a device on the jobsite for one purpose, he may invent other applica-tions for it. The ability of robotics to operate in the same environments as humans will inevitably lead to more automation options. The trend is in its early stages, but it has definitely begun. Searock believes that the time for getting in on the ground floor is not in the near future. “It’s here and it’s accelerating.”