But Can Autonomous Robots Handle Manufacturing?

By Drew Vass

For years, robotics companies have worked to mimic the human physique and physical capabilities. “We have robots with the physical abilities to do almost everything that humans can do,” says David Hanson, CEO and owner of Hanson Robotics, perhaps best known for its robot Sophia, a humanoid that engages in lifelike conversation. Now, swift progress in artificial intelligence (AI) is pressing autonomous robots closer than ever to human performance. But how far must they go before they’re capable of replacing human workers? Perhaps not as far as you might think if you haven’t paid close attention to the latest developments.

In February 2023, one of the major players in robotics, Boston Dynamics, announced that one of its humanoid machines “conquered” a construction site. The company’s Atlas robot, once made famous by its humanlike dance moves, was filmed stepping and jumping its way around a mock set-up, where it correctly identified and placed materials. Atlas can be seen retrieving a work bag, which it then carries up a set of scaffolding, tossing it to a human worker. Then, in a magnificent reminder of how robots can be physically superior to most humans, the machine somersaults off the scaffolding in an impressive dismount.

Despite this display of humanlike physical abilities, Atlas and other similar robots aren’t yet equipped with the sort of autonomy required for deployment in manufacturing or on real jobsites, experts admit. But by displaying the physical abilities, one thing is clear: They’re coming.

The Point of No Return

According to analysts at Thomasnet, despite half a century of development, industrial robots have been sparse in manufacturing up to this point. But according to the firm’s recent Thomas Insights, there are signs of real progress. In 2020, installed industrial robot capacity was 255 units per 10,000 employees—double the 176 units per 10,000 employees registered in 2015. According to data from the Association for Advancing Automation, around 9,000 robots were sold in the U.S. during the first quarter of last year, compared to more than 6,400 robots sold during the same period in the year prior. Some of those robots were semi-autonomous, but it isn’t likely that any were humanoids.

While fully autonomous robots haven’t penetrated door and window manufacturing, they’re gaining ground in other areas, such as warehousing and material handling. With optical and laser-based systems for sensing their environments and surroundings, and with the ability to identify objects, some are capable of loading and unloading trucks or picking orders. Add to that the latest AI technologies and it’s only a matter of time before they make the leap to more serious tasks, experts say.

“I’ve been to the robotics seminars and trade shows, and you see some of the stuff they’re doing and you take a step back, and you’re like, ‘Yeah, that’s cool, but how does it fit into the door and window industry?’” says Tom McGlinchy, executive vice president of engineering for GED Integrated Solutions.

To McGlinchy’s point, when it comes to doors and windows, components are highly variable and intricate in design. But already, autonomous robots are replacing human labor in areas that require extreme dexterity and at least some intelligence. In Europe, a company called Fieldwork Robotics developed two robots fitted with 3D-printed, plastic arms that are capable of picking raspberries, which are considered among the toughest fruits to pick. A company representative told The Guardian that the firm is aiming to have a single robot picking 25,000 raspberries a day, compared with 15,000 for a human working an eight-hour shift.

Not Fully Autonomous Yet

Meanwhile, the door and window manufacturing industries already have their fair share of robots. There are no humanoids or fully autonomous robots roaming around (yet), but advanced machines with robotic features have been in the industry for years. Companies such as Convey, Erdman, GED, Integrated Automation Systems (IAS), Stürtz and Urban Machinery all offer (or at least are working on) roboticized machinery for various aspects of manufacturing and material handling. Some systems even include combinations of laser and optical sensors that can be used to view and analyze materials for processing. While none have been known to pick raspberries, several years ago, GED employed one of its robots to prepare and serve coffee at a trade show. Prior to that, the company’s engineers demonstrated how a similar robot could even solve a Rubik’s Cube.

“You could scramble the Rubik’s Cube and give it to the robot, and it would solve this thing,” McGlinchy says. “It would solve it in probably two to three minutes, no matter how you scrambled it. You want to talk about problem solving?”

These days, many of those same machines can also be linked together to automate a wide range of processes. But what you won’t find in fenestration is any fully autonomous (“lights out”) manufacturing facilities that are devoid of humans. No matter how roboticized today’s door and window plants might be, the mobility, dexterity and cognitive abilities of humans are still necessary for directing operations and bridging processes. But experts say—even if today’s humanoids and autonomous robots aren’t ready for manufacturing—the prospect makes perfect sense.

“If you look at the jobs that humans do around those production lines, they’re dealing with uncertainty, dynamics and change,” says Rich Walker, managing director for Shadow Robot Co. “And often they’re the weak point in the whole process.”

A good example of when robots are a better option includes areas that require repetitive operations over and over again, Walker says.

“If you’re going to do something a million times, then you build a machine that does it well,” he says. Those are the instances in which human error tends to be costly, Walker and other experts suggest. “Humans are not good at staying ready for something in the long haul,” he says. “We get distracted or wander off.”

But in cases where there is heavy variability, humans remain the best option, Walker and other experts say.

“What’s interesting is, sometimes, when you see people demonstrating their flexible automation systems, you have a look and the most flexible bit of automation in them is a human who’s taking some things and doing a really complex task with them, then handing them over to the next machine,” he says.

In other words, like it or not, even with robotic machinery and automation, manufacturing is a human’s world. And nowhere is this more true, perhaps, than in doors and windows.

A Tough Case to Crack

“The reason it’s difficult for robots to penetrate the door and window industry is there’s a lot of human decisions going on,” McGlinchy says. “We’re not putting together cars, with three models and a bunch of options. We have this infinite size challenge, with five, six or more types of windows … Most companies have probably over 20,000 SKUs, and that’s where the real challenge is—adapting to all of these types and size changes.”

Even with some existing options for manufacturing automation, “We still have [technologies] that our customers don’t choose to go with, because we can outperform the machine in quality and speed [by using humans instead],” says Mitchell Heckbert, vice president of sales and service for Urban Machinery. In some cases, this dilemma has also stalled progress on robotics, as technology struggles to keep up with the high-speed, high-volume demands of U.S. manufacturers. For this eason, for now, at least, the idea of total human replacement might be the wrong conversation to have.

“I would say replacement is the wrong word,” Hanson says. A better word might be “cobots,” he suggests.

The Debut of the Cobot

While humanoids and autonomous robots haven’t gained the “mental” and physical acuity necessary to fully replace humans, robotics manufacturers have shifted their focus from conveyor-belt bots and other mixed automation tools to collaborative robots that rely on human guidance to perform tasks. In those cases, “The human is actually guiding the robot,” Hanson says. “So, the human and the robot are working together.”

Cobots aren’t replacements for existing workers, but act as “force multipliers,” analysts at Thomas say. As a result, human work partners complete tasks with more accuracy and efficiency than their non-human counterparts. According to the firm’s insights, cobots are less expensive to purchase, operate and maintain, which makes them particularly well-suited to small- and medium-sized operations that can’t afford traditional robotic systems. As a result, the market for cobots is expected to grow from $1.1 billion in 2022 to $9.2 billion by 2028, analysts predict.

While it’s debatable how many will end up in door and window manufacturing, cobots have proven their ability to cope with intricate materials and tasks and therefore shouldn’t be ruled out, experts say. For instance, Walker’s company has humanoid robotic arms and hands that are already equipped with dexterity that parallels humans. His company demonstrates this by performing an extremely familiar task: plugging in a USB cable.

“We’re always striving to build the most humanlike robot hands we can,” Walker says. “And the reason why we think that’s important and so interesting is, well, firstly, because everything in the human world is optimized for a pair of hands like mine.”

Add to those robotic hands humanlike perceptions and mobility, and suddenly doors and windows look less daunting. But, “To really do it, it needs to have all of the mental capabilities and emotional capabilities that humans have,” Hanson says. “And that, I would say, is 20 years away.”

Case in Point

When it comes to the robot produced by Walker’s company, it might be installing a USB plug, but it’s human-guided.

“The person in the background is wearing a pair of motion-tracking gloves,” Walker says. “That’s how the robot handles the task. So, it’s not using any autonomy or intelligence at all … I can put a pair of motion tracking gloves on, and I can control the robot hands with my hands directly to do the task.”

In a tightly controlled environment, those types of tasks might even be replicable without human input, but with door and window systems, “It’s very difficult,” says Ryan McHugh, president of Integrated Automation Systems. “You know, all these companies are doing custom manufacturing and every unit down the line is slightly different. That’s a lot harder to automate than the assembly of an F-150 [Ford truck], where everything coming down the line is essentially the same.”

For now, the gap between robots and humans is a costly space to fill. That’s not to say it can’t be done, but the bigger question might be price.

“If you gave someone a big enough budget, they could build you one,” Walker says. “But the question is, could they build them at a price point where it’s cost effective to use them for a task, rather than to use a simpler machine in a more complex way, or to use humans.”

There are plenty of reasons to embrace the idea, says Adil Sasmaz, managing director for Oz Machine USA. “Robots don’t get sick,” he says. “They aren’t late; they don’t rest; and they don’t need a lunch break. With preventative maintenance, you can keep them running 24 hours a day, seven days per week for years.”

But for the foreseeable future, at least, “Humanoid robots are going to cost more than a lot of other types of equipment,” Hanson says. For that reason, humanoids and autonomous robots might be here, but they haven’t walked (or somersaulted) into door and window manufacturing just yet.

Drew Vass is the executive editor of Door and Window Market [DWM] magazine.
dvass@glass.com

To view the laid-in version of this article in our digital edition, CLICK HERE.