A few years ago, I spent a week helping my friend Paul with the daily operations of his ski lodge in British Columbia. One of my first tasks was to change the duvets in the 10-room hotel. It’s simple: you turn the duvet inside out, reach into the far ends of the two-sided sheet, grasp the corners, and then pinch the comforter before turning the cover down around it. I then set about folding heaps of flannel sheets. And there I struggled. Unable to fold a fitted sheet so that it lay flat on the shelf, I sheepishly turned to YouTube. Online, I watched none other than domestic goddess Martha Stewart similarly struggle with the task. Since then, I’ve often wondered why, well into the 21st century, we don’t yet have in-home robots to fold our laundry.
As an early adopter of Roomba, the much-loved mobile vacuuming robot, I am someone who would surely welcome a clothes-folding robot if there were one to be had. Having a Roomba in the house is like having a friend stop by with a broom to sweep whenever you are out. Roomba keeps the domestic peace and sucks up dog hair that would otherwise have never left the house. Unfortunately, I will be waiting for some time before Roomba has much company. This is because many of the cleaning tasks that we love to hate are extremely specialized and unlikely to be done by robots in the near future. This became clear to me after diving into the research done by a team of computer scientists at University of California, Berkeley, who studied the efforts of robots to fold towels. It turns out that vacuuming the floor is literally one dimensional, while folding laundry is another beast entirely. “Cloth,” as the Berkeley researchers write, “is highly nonrigid, flexible, and deformable with an infinite dimensional configuration space.” Laundry folding is something easy for people but challenging for robots. And here’s why: to fold a towel is to know that a towel is a towel.
“Laundry folding is something easy for people but challenging for robots. And here’s why: to fold a towel is to know that a towel is a towel.”
To see why robots won’t take over our homes anytime soon, you need to consider the act of folding a towel through the machine mind of a robot: Once it has visually detected a towel, a robot must also detect its thickness, find a place to pinch it, and then know that it can pinch another edge within the span of its appendages. A robot must use gravity to create a new shape and then understand what to do with this new shape. A robot, as the Berkeley researchers found out, faces “challenges in grasp point detection and selection.” You? You grab the towel, pull it taut between your hands, snap the first fold, and move on. And, for heaven’s sake, you can fold about 50 times faster than today’s state-of-the-art robots. And this is just a towel we’re talking about; robots fail when it comes to T-shirts. And those pretty little underthings? Not a chance.
It’s fair to say that while researching the uncertain future of the automated home, I unwittingly stumbled on the answer to why folding fitted sheets is so hard. As the Berkeley team writes, for “many articles of clothing, the ability to grasp certain corners is a key enabler towards spreading out and then folding.” And fitted sheets, of course, don’t really have corners. In the end, struggling, as both Martha Stewart and I have with the fitted sheet, allowed me to realize how much my human existence can’t be easily replaced, how magnificently alive I truly am. And so next time I face the rounded elastic of a fitted sheet, I will rejoice in being human, being some sort of laundry-folding machine, even if a flawed one.