Sewbo Robot Sews Up Automated Garment Manufacturing

While robots enter other industries in herds, the assembly of garments has long been a tedious, human privilege. Now, for the first time, a robot has sewn an entire, wearable piece of garment. Sewbo, an industrial robot programmed to tackle the tricky task, assembles clothes and makes it look easy.

Despite their precision, robots still lack the cognitive abilities and motoric skills to process soft fabrics reliably. Sewbo tackles this by impregnating the fabric with PVA, a non-toxic biodegradable polymer. The temporarily stiffened fabric then can be processed as if it were sheet metal. It can be welded, molded, and most importantly, grabbed and sewn by the robot in a repeatable manner. From the finished garment, the PVA is removed by simply rinsing it with warm water.

To sew two pieces of impregnated fabric together, they are first tacked together using an ultrasonic welder. Other than that, the process is surprisingly similar to manual sewing: Sewbo simply grabs the assembly and runs it through a standard sewing machine to put a permanent seam in place. Due to the fabric’s increased rigidity, this step is repeatable enough to let the robot work blindfolded, with no CV magic required.

Naturally, behind every smart robot, there’s a smart human. In case of Sewbo, it’s [Jonathan Zornow], who invented the new process. For his proof-of-concept, he rented a Universal Robot UR-5 for 30 days, equipped it with a vacuum cup end effector (kindly borrowed from Axis Automation in New Jersey), and bolted the whole thing to his Ikea dining room table. A consumer-grade, Brother CS6000i sewing machine was then hacked to be handled by the robot’s motion controller.

On the way to filling the automation gap in garment manufacturing, Sewbo marks an important milestone and we are curious to see this concept being taken further. Impregnating fabric with polymers may raise environmental concerns, but it’s already common practice in textile manufacturing to strengthen yarn with PVA for easier processing in weaving machines. It probably could be washed out a few steps later in the process chain.

Does anybody here have experience with kitting holes in dining room tables? Let us know in the comments! Enjoy the video below, where you can see how Sewbo assembles a T-shirt from start to finish.

42 thoughts on “Sewbo Robot Sews Up Automated Garment Manufacturing

  1. “Now, for the first time, a robot has sewn an entire, wearable piece of garment.”

    Thats a misleading statement. It should read “robot arm sewing machine feeder”. This is more on par with a “self parking car” and not the “fully autonomous vehicle” I was expecting to see.

    1. Definition for ‘robot’ from Merriam Webster
      ” a machine that can do the work of a person and that works automatically or is controlled by a computer”

      So, from that definition the sewing machine is considered a robot, and it’s been around a lot longer….

    2. I doubted this was even real, not because of what it was doing, but because I couldn’t believe anyone would bother making it for as little as it does. It takes only a few hours of sewing by hand to get used to the feed process, and often feeding techniques require tensioning the cloth variably as one shifts from on-bias to off-bias for a curve or wants a tuck or ruffle, etc. This robotic arm seems like a modern take on something that was outdated by 1900. Would have much more use for a free-arm quilter or the like. But then, I’m just a hobbyist, so maybe there’s something I’m missing.

      1. No, I’m afraid in the needle trades it’s all about speed, and at any rate they pay these girls a pittance per unit anyway. While there may be a market in the lower end bespoke sector, it will be a while before this will close down Asian sweatshops for prêt-à-porter.

          1. “After setup” means after initial investment and at this point, given the cost of labor in places like Bangladesh and the margins this sector commands, is going to be very high. At this point you cannot make a business case for automation in this industry unless there is a huge shift in the way international trade is done. In these Third World sweatshops, people are simply cheaper than machines.

          2. Are they though?
            UR-5 has a useful life of 36k hrs (2yr warranty) and initial investment of 33k USD, 4k USD for 3 training classes on maintaining the arm, and lastly a 50W PSU.

            I couldn’t find anything on how long it takes to sew a t-shirt on one of these but it (apparently) takes a human 30-60 min once all the material is cut to size. The actual stitching is probably already done as fast as it can be done, so the only time you’re saving is in alignment. Which is likely more to do with the stiff fabric than the UR-5, in which case, why bother? Just use the same PVA sheets & human labor.

            The UR-5 costs 22.08 USD/day (ignoring power consumption & training) it may beat out a first world seamstress/ter but lets be realistic, no one makes t-shirts from scratch in the first world. A lot of clothing is made in Bangladesh & Vietnam which have minimum wages of about 3.40 & 7.85 USD/day respectively. This is a great idea to pursue, but there needs to be massive changes to the garment industry for this to be in any way competitive.

          3. On an assembly line with different people doing each task it takes less than five minutes from first fabric cut to packaging and shipping to do just about any common item of clothing. On top of which there’s likely several hundred items being made at the same time

  2. maby it would be possible to sepperate the polymer from the rinsing water and reuse it..
    or better use bio degradable polymers made from plants.

    now the tricky part is to get the robot to; turn it inside out and fold it precisely, when it have been washed.

  3. This and all work saving automation is amazing and will benefit greatly someone. It can be amazing for everyone where we can enjoy a robo-communist economy in an advanced economy where a Marx-Engels classless society, without the deformed bureaucratic ruling class worker state of Lennin and Stalin, can actually work. Where scarcity is limited to energy production and pollution, where we can peruse infinite catalogs of makey stuff or custom made recyclable materials. We can choose the food and stuff we need and act wealthy because as a community we have made it so. Or the means of production can remain in the hands of the few who can afford to own and maintain the machines, maybe they can throw plastic tents and protein sludge laced with contraceptives to the unemployed dreck while they enjoy a Roman generational holiday on the few normals they breed to remind themselves how much better they are then slaves. For now at least poor people in cheap economic zones have some hope working in a garment sweatshop, what when it is automated cheaper than them? This story changed a big part of my world view on the automation singularity http://marshallbrain.com/manna1.htm

    1. Read 1984, It doesn’t matter how much stuff you have, what matters is how much more stuff you have than your neighbor. If we have a robo-communist economy, the people who first purchased the machines will have power because they control production. They wont give power to the masses, because then they won’t have any. I guess that I think and fear that your second option is more likely.

      1. Not necessary. Keep in mind that it was only in the last century that clothing started to be made in factories in any significant amount – most were made at home, and if robotic seamstresses are coming it is likely that they will enter by the same path as sewing machines, that is as a home consumer product, rather than one for bulk manufacturing.

        1. Doubtful.
          Consider the roadblocks to entry for a robotic system. Capital investment, intellectual investment all with a low rate of return (margins on garments are very low for everything but designer clothes). Until robotic arms are commoditized and are able to be programmed by gesture this will be a rich man’s pastime.

          Nice to see some discussion about the needs of common laborers.

          1. Margins on prêt-à-porter are huge Gross Margins in the apparel market run around 50% average. Designer clothes on the other hand have tiny margins, as they really exist to support the label.

        2. The problem with any kind of manufacturing is waste…using water to rinse out PVA pollutes water, land, and ultimately the food chain, outdoing the huge ocean plastic pollution impacts. This technology fails to consider this and recycling issues.

      2. The trick is the population all needs to have equal(yup, it will require a short period taxing the rich to equality) share in the automated means of production, and their dividend is the combined output of that production. The way to do big projects is simple pooling shares in affinity groups for stuff like a space program, beanstalk, or supercollider. I suppose position in society would be by fame or contributing popular IP from designs to multimedia where people contribute production shares to your project as long as it interests or otherwise benefits them.
        It is like how the pre-Columbian Native American economy distributed buffalo, stone, and wood; there was enough and people just consumed what they needed or could make. We can operate like that but have modern medicine and almost unlimited recyclable resources. So much to go wrong but without artificial work, there are only so many slaves useful to one family, 90-99.99% of the economy becomes something that the feudal or slave owner lords will wish to expire quietly.

  4. This excites and terrifies me. Will we live in a world where we have the automation and advanced enough economy to live in a Marx-Engels (no Stalinist bureaucratic ruling class) Communist utopia without artificial scarcity or a Fuedal cesspool where even near-slavery is not a way to step up; where only the strongest and wealthiest have kept wrested exclusive control to the means of production. Can we all live wealthy sustainable lives or can a few families live in orbital palaces while they warehouse the poor out of sight on a few square km of waste land?
    I love the first few chapters of manna, describes the problem perfectly. http://marshallbrain.com/manna.htm

  5. i was excited by the headline, disappointed by the PVA hack. i believe a real sewing robot (and a real knitting robot and so on) is on the horizon, but the sort of things coming out are kind of limited so far.

  6. I don’t mean to be rude, but the amount of misinformation is these sewing related stories is pretty staggering. Go to YouTube, search for “automated sewing” and get some perspective on how this stuff is already being done.

    This example is a fun “home” project, and a cool hack, but the industry is simply at a different level and I don’t think its sophistication has been properly represented.

  7. I’ve been waiting, for a long time, to see the day that robots can accomplish tasks like this. The key isn’t that they’re cheaper than kids in china, the key is that they’re cheap here in the USA, etc. Currently we ship cotton over the ocean and they ship clothes back. Huge waste. Soon we’ll just keep the cotton here and make our own clothes.

    1. Not going to happen. Those “kids in China” can also build robots and it will always be cheaper to make and run them there. Or in some other low wage country with flexible laws and taxes.

  8. Sorry, not impressed yet. The video skips over the numerous human involved steps to produce the shirt in the video. The arm is only providing a single function and that is positioning the stiffened cloth through the sewing machine. Prepping the segments of the shirt so the arm just has to move the material through is all edited out. The design is dependent on stiff and properly positioned material which is a variable that will not be perfect (notice the lateral sag when the full shirt is picked up for side sewing at 0:31) , thus leaving tolerances in the quality of sewing which are clearly evident in the left shoulder in the final shot.
    I am impressed with the idea of stiffening the fabric for handling purposes, however. That would make lots of materials easier to handle for many hand sewn projects and for use on a laser.

  9. The funny fact is that many pieces of clothing are done by machine only, though, sewing is not like CNC machining by any stretch of the imagination.

    Though, I would rather say that just tossing a robot arm at a problem is far from a logical solution…
    (It’s like some people toss encryption onto data security and leaves it at that.)

    Simply, this machine could be far more efficient if one stops using the arm, and replaces it with something more suited for the job. And for those thinking replicating how humans do things must be most efficient, this is far from the case.
    (Take the wheel for an example, it’s far more efficient at moving from A to B then legs. (on relatively smooth surfaces))

    Now if this were developed for the task, like a knitting machine is, then it would probably be a lot faster, and maybe even mostly self running. (And never heard of a knitting machine then go look at one!) http://hackaday.com/2015/06/29/hackaday-prize-entry-circular-knitting-machines/

    Now I’m not thinking of designing a T-shirt machine, but this method presented here looks rather shortsighted to be honest.

      1. That surly is true. It would be interesting to see someone fully automate a real sewing machine.
        I though a little on the subject, as why not. And it wouldn’t be too hard to make a reasonable machine.

        The thing that the machine here does bad is that it’s not holding the fabric at the seam and aligning it to make a decent product.

        So what is needed is practically two rubber “fingers” on each side of the needle, a camera doing the aligning, and something that feeds over the fabric to the machine, on top of this we would only need some form of easy way of placing two pieces of fabric on top of each other to feed in under the needle, this could be done simply by having a second slightly elevated table, then drag the top one onto the bottom one. (The elevated table is to get rid of the problems of the fabrics folding onto each other, as friction is apparently real.) Here in this step were the top fabric is dragged onto the lower one is were we need another form of machine doing a bit of alignment, this could probably be an arm, then when it comes to unfolding fabrics and keeping things ordered as we move them around is where the Camera and the robotic arm shines. (Not in the actual sewing…)

        If the hack were something along that line, it would maybe have been more interesting, and expensive…

  10. All comments above make many good points…..but….. THE STITCHING / SEAMS ARE HORRIBLE…PERIOD i would not use it to make any garment for any reason, a person with a needle and thread could do better even if they had never sewn before, it does not matter if it could make a 1000 to 1 at lease the one would be wearable or more importantly marketable

    1. I love this! I only read a handful of the comments but could not believe how much people are giving you a hard time for something I doubt they could create if they tried. Sure it is not perfect it needs fine tuning but for a beta version or a proof of concept it is wonderful! I would love an automated sewing machine that could see a pattern, get data for measurements, and then do what your machine just did and I don’t think it is that far from connecting with a user interface that allows someone to plug in what they need, pick the stitches, and go.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.