In this first edition of *To Be Automated*, I recall the journey that led me to dive into the world of automation and my first experience as a researcher working to automate the work of masons in Atlanta.

Georgia Tech, Intro to Robotic Fabrication, Fall 2016

Automation is typically discussed in terms of two extremes: great for corporations, disastrous for workers.

By discussing my experience, I hope to shed some light on the nuance of the argument for automation. I want to share some of the process that goes into identifying what should be automated, and I want to highlight the difficult situations that lead even laborers to want their work automated.

A few notes to begin:

Automation takes time. In a capitalistic society, automation is pursued as soon as it becomes cheaper than human labor and no sooner. Over time, small improvements lead a gradual shift toward more automated solutions.

Automation can be simple. Low tech solutions often offer the most reliable performance increases. Processes will be automated in the simplest way possible, high tech solutions are not always the most efficient.

Automation does not have to exclude humans. Partial automation is often more effective than full automation, especially in the short term.

Georgia Tech, *Intro to Robotic Fabrication*, Fall 2016, Grout pouring by hand to reinforce dry-stack masonry.

The Difference Between Human and Machine Labor

Humans excel in adaptability: our cognitive process allows us to problem solve and re-focus our efforts as obstacles arise. As architects and designers we see this in the non-linear nature of the studio process and our ability to jump quickly between abstract ideas. For the foreseeable future, humans have this advantage over machines. However, when industries project production cycles over larger time frames they find that in many cases, machines will outperform humans. Repetitive tasks are the most valuable target for automation. Once a procedure is developed to carry out the single task, that task can be repeated practically non-stop for the operational lifetime of the machine. The machine does not fatigue the same way that humans do, and in many cases a machine does not require the same environmental conditions that humans do to perform their jobs. Robotic arms are often operated in lights-out facilities where the lights in the building are only turned on when human maintenance workers need to service the machines. The machines do not face mental stress from performing the same routine, and they can be designed to withstand extreme temperatures, prolonged vibration, and buildup of waste particulate. For all of these reasons and more, automation is becoming increasingly desirable. To effectively automate a process, enormous up-front research and development costs as well as capital investment into facilities, hardware, and skilled labor are required.

This up-front research is the particular area where I have spent my past few summers, and these are the experiences I want to reflect on. Automation in construction is in the early stages, but the research is well underway.

Yaskawa Denki Future Lab, Summer 2016

**How I as an architecture student found my way into automation.**
As an undergraduate, I wanted to study abroad in Japan, but the architecture school at Georgia Tech did not offer that opportunity. Instead, I took classes in the School of Modern Languages to be able to attend a Language, Business, and Technology program in Japan. I spent the summer of 2016 taking classes at Ritsumeikan Asia Pacific University in Beppu, Japan. For a summer, I was taking classes taught entirely in Japanese, and none of them had anything to do with architecture. However, one of the classes was a current events course, where we learned about contemporary Japanese sustainability initiatives and toured various companies learning about how Japanese businesses operated. During one of these class trips, we toured the headquarters of Yaskawa Denki, an electronics company with a focus on robotics manufacturing. This particular facility was the production facility for the Yaskawa Motoman series of multi-axis arms. The showroom area featured all sorts of robotic displays: teams of small robots programmed to dance in unison, biomedical robots navigating obstacle courses with extreme precision, and numerous arcade style games where the participant could compete against the robot in a puzzle solving task. The robots were programmed to play at a normal speed at first, and then advance to an unmatchable pace, winning every time.

YaskawaDenkiFutureLab2016 — Robots Building Robots

Beyond the showroom, we toured the building where the assembly line of Motoman robotic arms were producing more Motoman robotic arms. The entire process was automated, and the lights were only on so that they could display the process to us visitors. When robots needed to be serviced, they were pulled off of the line, brought to a separate area for human technicians to make necessary repairs, and in the meantime a replacement was already in place continuing production. Somehow, none of the fantastic architecture that I saw on this summer trip compared to this one robotics facility. The idea that the whole building could be functioning without human oversight, yet maintaining a higher level of efficiency than a fully-human operated plant was incredible. I was fascinated, and the night I returned I immediately sent an email to Shani Sharif, a PhD student at Georgia Tech researching human robot interaction. I desperately wanted to learn more, and she granted me a spot in her course for the coming semester: Intro to Robotic Fabrication.

Georgia Tech Digital Fabrication Lab 2017-2018
Introduction to Industrial Robots

One year later, I had been spending much of my academic year at the Digital Fabrication Lab, a shop space for students which had recently acquired a used industrial robotic arm. The arm, a KUKA brand KR90 R2700, was far beyond the capabilities of the arms I saw at the Yaskawa facilities, closer to the type of robot used in automotive manufacturing. To de-code the name, “90” refers to the payload, the amount of weight that the robot can lift: in this case, 90 kilograms. The R2700 refers to the robot’s reach, 2,700 millimeters. This six axis robot was far stronger, had far wider reach, and could move faster and with more precision than me. Essentially, anything that I could do with my own arm, I could program the robot to do much more efficiently. Placing objects, forming sheet metal, and long exposure light drawings were all among the experiments we conducted at this time.
By summertime, I was looking for an internship, and Shani’s PhD advisor, Dr. Russell Gentry invited me to work with him on a project with the Digital Building Lab using the robot to build masonry structures. Though still an academic setting, this opportunity was when I first started interfacing the professionals in the construction industry. The research objective of this summer was to collaborate with Jollay Masonry, an Atlanta-based masonry contractor who were looking for ways to shift their work into off-site facilities.

**Brick-Layers’ Dilemma**

Masonry was falling out of favor in the Atlanta construction market. Pre-cast concrete in particular was cheaper, faster, and required less skilled labor. The elements would be manufactured off-site, delivered, and installed in a fraction of the time that masons could finish the same element, allowing the following phases of construction to proceed far sooner. Masonry could not compete with the speed of off-site manufacturing. Then, there was the difficulty of digitization in the masonry industry. Architectural modeling software are not ideal for masonry applications. Initiatives such as the Masonry Unit Database are seeking to address this issue by providing access to libraries of masonry elements, but fundamentally masonry does not lend itself to precise digital modeling in the same way that pre-cast concrete does. With concrete, parts can be produced with a high degree of precision, allowing for construction with predictable levels of tolerance. Masonry is constructed with small units and with vulnerability to environmental conditions that exacerbate tolerance issues. A digital model trying to predict where 10,000 bricks will be on a façade will inevitably be inaccurate, as there is no way to account for the buildup of tiny differences. This part of the process is where skilled masons come in. With the expertise that these workers build over time they learn to minimize errors and adjust their work as they go. Again, this is a time consuming process. Finally, there is a large risk factor to having the masons on-site for the duration of their work. Even skilled masons will take significant time to finish a façade. During this entire time that they are on site, they are at higher risk of work-related injury due to the high number of uncontrolled variables on an active construction site.

The hot humid Atlanta climate was especially unfavorable for workers on these sites. Not only did the heat increase worker fatigue, but the high moisture made the masonry work even more unpredictable due to the constant need to alter the mortar mixture based on the changing humidity. Essentially, all of the odds are stacked against the masons in Atlanta, and their work prospects reflect that. They were not turning to researchers because they wanted the work to be automated, but because if the work was not improved, then the market would no longer accommodate them.
For that entire summer, I got hands-on experience working with masons on all parts of the process. They taught me how to keep a batch of mortar at the right consistency, how to “butter” a brick, how to modify bricks, and most importantly how to build a wall. For the rest of the summer I spent my time devising different optimizations that would help translate the process into a lab setting to deliver higher precision, speed, and safety. Everyone involved was aware that if we did not improve the process, then the masonry industry in Atlanta was going to continue to decline. This was a case where automation would be saving jobs that were otherwise likely to disappear.

A small portion of the most experienced masons would surely find work doing exquisite custom home additions or some other specialty work, but the majority of the masons, especially the less experienced group, would be unable to continue the profession that they had dedicated some portion of their lives to. Although this seems like the perfect place where automation could actually help people, there’s still something unsavory about the idea that we would be handing over the work to the domain of robots. For that reason, this particular research project heavily focused on maintaining the involvement of humans in the process.

**Re-Imagining the Role of the Mason**

It's important to remember when automating a process that automation research costs time and money. Realistically, work is only automated when barriers to affordability are removed.
Different processes can be automated to different degrees. With a high degree of accuracy we could rely on the robot to place every single brick in the designated spot. We could also theoretically incorporate sub-processes into the brick stacking, such as having the robot add a layer of lateral reinforcing wire every fourth course. Some parts of the procedure weren’t worth figuring out exclusively with the robot, but were suitable opportunities for a robot and human to collaborate. For example, a robot could perform a simple task such as holding a piece of rebar in precisely the right location in space for a human to weld the rebar to a steel base plate.
Then, there were the series of variables which would certainly not be cost efficient to delegate to a robot. Imagine the level of precision and control that a robot would need to apply the correct amount of mortar to each brick before placing. The robot runs into the issue that all masons do where the consistency of the mortar shifts due to inconsistencies in the mix design, differences in admixture ratio, differences in humidity, or some combination of all of these factors.
Alternatively, the procedure is designed to overcompensate, adding so much mortar that the brick will definitely stick in place, but large quantities of mortar are wasted. Beyond wastefulness, this overcompensation ruins the wall, even if the excess is scraped away, the traces of cement discolor the brick, revealing how messy the process was.
To design a facility to control all of these variables is unfeasible for the masonry industry who are already struggling with declining demand. Instead, these tasks are perfect to leave with the already experienced human workers. In this sense, the bricklayer becomes more of a mortar-tender. The robot is much better at placing the bricks, but the mason is well-suited to overseeing the process, tending the variable aspects to ensure quality brick-setting. Moreover, this alternative role for the mason addresses nearly all of the risks related to extended time on the construction site. Workers in a lab setting take turns with the robots, staying away from the moving machinery while it operates, and then doing their part when the robot is safely disengaged.

**The Limits of Automation in the Masonry Industry**

Over the course of the summer, we stress tested numerous concrete and grout mixtures for reinforcing the walls, and eventually developed a working procedure for constructing masonry elements in the lab suitable for transport to site. We were able to lift a segment of wall from its vertical standing, place it on its side (as in a flatbed truck), and lift it back up to be placed vertically. This proof of concept was an effective case study for a scenario where full automation is not worth pursuing, and instead we prioritize automating particularly repetitive and predictable portions of the process. Automation is tedious to achieve, and though we were highly selective about only automating the easiest possible portions of the work, the process still took numerous attempts at every stage.
Speculating beyond the academic setting, off-site masonry manufacturing facilities would eventually automate the monitoring and adjustment of temperature and humidity within the building, perhaps employing custom machinery to mix the mortar and grout. These additions might lead the facility to look more like a food processing plant, where conditions are precisely monitored and quality diagnostics assure that each batch is practically identical. Self-driving vehicles are right on the horizon, and it’s much simpler to automate transport within a confined building than on a road network. Likely, materials would be moving around the facility via Amazon style inventory robots or self-driving forklifts. Then, there’s no reason that these robots shouldn’t be configured in a network such that they signal each other with relevant information. There would be no need for humans to turn the machines on at the right time or initiate the scripts, the various computers within the facility could be controlling all of the robots simultaneously. Numerous optimizations could happen on the construction site as well, such as deploying robots on site or designing robot systems to work from moving platforms on facades.

The Future of Masons

From the worker’s perspective,would this automated future be beneficial? A future masonry facility would likely be less physically taxing and lead to fewer long term health problems. If the facility could lower costs of production significantly then it would also maintain the demand for masonry construction, saving the industry from being phased out entirely. Ideally every plan for automation would have an accompanying transition plan for workers, but that is not always the case. The masons in Atlanta would surely prefer that their work didn’t need to be automated, but they find themselves needing to adapt to keep pace with a changing marketplace. By addressing their profession’s trajectory, the masonry industry has positioned itself to be proactive in the shift toward automation. Some combinations of efforts from all parties will be required to maintain the human role in the masonry process. Masons will need to become familiar with some level of digital interface, but maintain the skills and perception they gain through manual experience. Industries must not take the skill aspect of the labor for granted. The prospect of augmenting the existing masonry process should be prioritized over full automation. Eventually, if properly integrated to take advantages of the adaptability of the human mason and the efficiency and reliability of the robot mason, the augmented mason should have an expanded capability to maintain their profession into the future.