Human hands are extremely good at making clothes. While many manufacturing processes have been automated, stitching together garments remains a job for millions of people around the world. As with most labour-intensive tasks, much of the work has migrated to low-wage countries, especially in Asia. Factory conditions can be gruelling. As nations develop and wages rise, the trade moves on to the next cheapest location: from China, to Bangladesh and, now that it is opening up, Myanmar. Could that migration be about to end with the development of a robotic sewing machine?
There have been many attempts to automate sewing. Some processes can now be carried out autonomously: the cutting of fabric, for instance, and sometimes sewing buttons or pockets. But it is devilishly difficult to make a machine in which fabric goes in one end and finished garments, such as jeans and T-shirts, come out the other. The particularly tricky bit is stitching two pieces of material together. This involves aligning the material correctly to the sewing head, feeding it through and constantly adjusting the fabric to prevent it slipping and buckling, while all the time keeping the stitches neat and the thread at the right tension. Nimble fingers invariably prove better at this than cogs, wheels and servo motors.
“The distortion of the fabric is no longer an issue. That’s what prevented automatic sewing in the past,” says Steve Dickerson, the founder of SoftWear Automation, a textile-equipment manufacturer based in Atlanta, where Dr Dickerson was a professor at the Georgia Institute of Technology.
The company is developing machines which tackle the problems of automated sewing in a number of ways. They use cameras linked to a computer to track the stitching. Researchers have tried using machine vision before, for instance by having cameras detect the edge of a piece of fabric to work out where to stitch.
The Atlanta team, however, have greatly increased accuracy by using high-speed photography to capture up to 1,000 frames per second. These images are then manipulated by software to produce a higher level of contrast. This more vivid image allows the computer to pick out individual threads in the fabric. Instead of measuring the fabric the robotic sewing machine counts the number of threads to determine the stitching position. As a consequence, any distortion to the fabric made by each punch of the needle can be measured extremely accurately.
These measurements also allow the “feed dog”, which gently pulls fabric through the machine, to make constant tiny adjustments to keep things smooth and even.
Dr Dickerson patented the idea in 2012 and won a $1.3m research contract from DARPA, a US Department of Defence research-and-development agency. The military interest in sewing arises from a 1941 requirement that the department gives preference to American suppliers when buying uniforms. Bill Lockhart, a SoftWear Automation executive, says last year the project began successfully stitching together pieces of fabric robotically and that one machine being developed is now able to stitch a perfect circle—something that only a highly skilled human operator would dare to attempt.
But neat stitching is only part of the sewing process. Most garments are assembled from various different pieces of fabric: 20 or more sections typically for a pair of jeans. This means a robotic sewing machine also needs to be able to pick up material, feed the appropriate sections to the sewing head and remove them when complete. To do that, the company has developed a materials-handling system which it calls LOWRY. This uses a vacuum grip to pick up pieces of fabric and move them to another machine, which might cut, stitch, add buttons or carry out other finishing tasks. LOWRY is programmable, so can switch easily from working with one size of material to another.
A stitch in time
Although the first LOWRY will be delivered to an American factory later this year, commercial versions of the firm’s robotic sewing machines capable of automating the more difficult tasks in making garments will not be ready until next year.
Frank Henderson, the boss of Henderson Sewing, an Alabama-based textile-equipment firm, and an investor in SoftWear Automation, reckons that robotic sewing will be attractive to American fashion brands wanting to bring production closer to home and produce garments rapidly to catch new trends. Inditex, a “fast fashion” Spanish company, whose brands include Zara, famously does this by making many of its garments manually in Europe to speed up its time-to-market. With designs and samples shuttling to and from Asian factories, it can take months before new clothes finally turn up in other American and European stores—by which time they can be out of fashion.
Jack Plunkett, of Plunkett Research, a market-research company, says pressure on Asian clothing manufacturers to keep wages low while improving working conditions is leading many to look at automation, too. On that point, SoftWear Automation’s Mr Lockhart says a Bangladeshi company has already expressed interest in the firm’s technology.
But it is not only the production of garments which could be increasingly automated. Nike is now using a technology called Flyknit to make some of its trainers. Flyknit uses a computer-controlled knitting machine to automatically weave strands of polyester yarn into the shape of the upper part of a shoe, instead of having it manually stitched together from individual panels, the way most trainers are made in Asian factories. Nike has not said how far it intends to take the technology, but it has the potential to produce customised trainers for individuals and to do so locally—perhaps even within stores.
Shoemakers are already using 3D printers, which build up material additively, to make prototypes of shoes. Exotic clothing and shoes made with 3D printers are becoming regulars on the catwalks at many of the world’s leading fashion shows, although the materials they are printed from tend to be various sorts of plastic, which can make the garments somewhat clunky and shoes a bit clog-like. However, researchers are working on ways to print more flexible materials. One such project involves a collaboration between Disney, Cornell University and Carnegie Mellon University. Their 3D printer uses layers of off-the-shelf fabric to make soft objects, such as cuddly toys.
The real test of how successful robots will be at making clothing and shoes will depend on how efficient and reliable they will be, and how fully they can automate the process. If time-to-market and customisation are priorities, then the robots might win—even if some manual intervention in production is required. But for mass-produced lines, where every cent matters, any human involvement could keep manufacturing offshore. The lesson from industrial automation in other sorts of factories, though, shows that robots keep getting better and cheaper. It may be a while coming, but the writing seems to be on the wall for sweatshops.