The Difficulty of Creating a Laundry-folding Robot (npr.org) 75
"It might be a while before you can buy a 'Roomba for laundry'," jokes Slashdot reader Tony Isaac, pointing out that "while robots have been developed that can fold specific types of laundry, there's still not a good robot that can do the job quickly, or for all types."
But NPR reports laundry-folding robots are getting closer: As NPR has reported, machines need clear rules in order to function, and it's hard for them to figure out what exactly is going on in those messy piles That's not to say that it's completely impossible. University of California, Berkeley professor Pieter Abbeel spent years teaching a robot how to fold a towel, eventually cutting that process down from 20 minutes to a whopping minute and a half.
And Silicon Valley-based company FoldiMate raised hopes and eyebrows when it showed off a prototype of its eponymous laundry-folding robot at the Consumer Electronics Show in early 2019. It said the machine could fold some 25 pieces of laundry — except for small items like socks and large items like sheets — in under five minutes, with an estimated price tag of $980. It's unclear what happened to that company — its website is down and it hasn't tweeted since April 2020. Its sole competitor, a Japanese company with an AI-powered prototype, filed for bankruptcy.
In sum, most robots have not generally been equipped for the task. But an international group of researchers say their new method could change that — or at least speed up the process. Researchers are calling the new method, SpeedFolding. It's a "reliable and efficient bimanual system" — meaning it involves two hands — that's able to smooth and fold a crumpled garment in record speed (for robots, that is). SpeedFolding can fold 30 to 40 strewn-about garments per hour, compared to previous models that averaged three to six garments in that same time span, according to researchers. They say their robot can fold items in under two minutes, with a success rate of 93%.
"Real-world experiments show that the system is able to generalize to unseen garments of different color, shape, and stiffness," they add.
According to the article, the team will be presenting their paper at a robotics conference in Kyoto this month, and they've also posted a one-minute video on YouTube. (Their solution involves both an overhead camera and a novel neural network called BiManual Manipulation Network that "studied 4,300 human and machine-assisted actions in order to learn how to smooth and fold garments from a random configuration."
"While researchers describe SpeedFolding as a significant improvement, it's not likely to hit the market anytime soon," notes NPR. "For one, Ars Technica tracked down a robot similar to the one they used and found that it retails for $58,000."
But NPR reports laundry-folding robots are getting closer: As NPR has reported, machines need clear rules in order to function, and it's hard for them to figure out what exactly is going on in those messy piles That's not to say that it's completely impossible. University of California, Berkeley professor Pieter Abbeel spent years teaching a robot how to fold a towel, eventually cutting that process down from 20 minutes to a whopping minute and a half.
And Silicon Valley-based company FoldiMate raised hopes and eyebrows when it showed off a prototype of its eponymous laundry-folding robot at the Consumer Electronics Show in early 2019. It said the machine could fold some 25 pieces of laundry — except for small items like socks and large items like sheets — in under five minutes, with an estimated price tag of $980. It's unclear what happened to that company — its website is down and it hasn't tweeted since April 2020. Its sole competitor, a Japanese company with an AI-powered prototype, filed for bankruptcy.
In sum, most robots have not generally been equipped for the task. But an international group of researchers say their new method could change that — or at least speed up the process. Researchers are calling the new method, SpeedFolding. It's a "reliable and efficient bimanual system" — meaning it involves two hands — that's able to smooth and fold a crumpled garment in record speed (for robots, that is). SpeedFolding can fold 30 to 40 strewn-about garments per hour, compared to previous models that averaged three to six garments in that same time span, according to researchers. They say their robot can fold items in under two minutes, with a success rate of 93%.
"Real-world experiments show that the system is able to generalize to unseen garments of different color, shape, and stiffness," they add.
According to the article, the team will be presenting their paper at a robotics conference in Kyoto this month, and they've also posted a one-minute video on YouTube. (Their solution involves both an overhead camera and a novel neural network called BiManual Manipulation Network that "studied 4,300 human and machine-assisted actions in order to learn how to smooth and fold garments from a random configuration."
"While researchers describe SpeedFolding as a significant improvement, it's not likely to hit the market anytime soon," notes NPR. "For one, Ars Technica tracked down a robot similar to the one they used and found that it retails for $58,000."
Laundry-folding robot (Score:2)
One thing you can be sure of is that the company that will makes a Roomba for folding laundry will get bought out by Amazon or Google for the express purpose of having yet another device in your house to keep tabs on you.
Re: (Score:2)
Folding laundry never made sense to me. I put my shirts and pants on hangers. It is faster than folding and leaves them wrinkle/crease free.
Re: (Score:3)
Folding laundry never made sense to me. I put my shirts and pants on hangers. It is faster than folding and leaves them wrinkle/crease free.
This works if you only have a few shirts and pants. For people who have real wardrobes and multiple seasons of clothes there wouldn't be enough closet space in most apartments and even some houses to do this.
Re: (Score:2)
No to mention, it works if you work at McDonald's or at Google. If you have a job that requires you to dress sharply, you need to iron and fold your shirts and pants.
My solution to this is to bring my stuff to the dry cleaner. It's not cheap, but it's cheaper than my time if I did it, I hate doing it so it's worth the money to me, it'll probably be cheaper than a robot for a long time to come - and better done - and it keeps two people employed.
Re: (Score:2)
Exactly. Anything that needs to avoid creases goes on a hanger. Everything else gets stuffed into or dropped into bins. If it refuses to be crumpled, the punishment is hanging.
The Difficulty... (Score:3)
...in replacing all human jobs.
Will we as a species feel any better about ourselves when that headline reads how easy it is to replace all human jobs?
Just curious. You know, for planning more than 5 tweets ahead in life.
Re: (Score:3)
Will we as a species feel any better about ourselves when that headline reads how easy it is to replace all human jobs?
Replacing all human jobs will require human-level general artificial intelligence. Machines that can think as humans do will be a profound revolutionary change comparable to humanity taming fire. Keeping your job will be the least of your concerns.
Re: (Score:3)
In old science fiction, low skilled jobs are done by robots and higher function jobs like plotting navigation is done by humans. We see the opposite is true. Maths fell easily to computers. No one solves systems of equations by hand. Humans set the up, machines do the heavy lifting.
Those who can leverage the machine to create value earn the money. Those that depend on their wits and physical strengths tend to be low paid gig workers.
Re: (Score:1)
In old science fiction, low skilled jobs are done by robots and higher function jobs like plotting navigation is done by humans. We see the opposite is true. Maths fell easily to computers. No one solves systems of equations by hand. Humans set the up, machines do the heavy lifting.
That's just down to the idiosyncratic way human bodies and brains work. Many things we do instinctively and don't think about at all actually turn out to be pretty complicated. A while ago there was an article about a company showing off a robot that can run on two legs, the way bipedal animals do. It was quite a bit slower than the fastest human runners, even though you'd expect a machine to be better at this. Otoh many things we consider to be hard and to require intelligence, like maths, turn out to be e
Re: (Score:2)
Is it possible that there are more fulfilling activities for human beings than folding laundry? Why must we reserve every job for humans as if robots are going to push us out into the street and live in our home. Is it some kind of misplaced fear of immigrants applied to robots.
Of course when robots can do everything our economic system will be completely different. Hopefully we catch onto that fact early enough to eliminate all the rich people before they can own all the robots and all the police.
Re: (Score:3)
Of course when robots can do everything our economic system will be completely different. Hopefully we catch onto that fact early enough to eliminate all the rich people before they can own all the robots and all the police.
*laughs in predator drone*
Re: (Score:2)
Is it possible that there are more fulfilling activities for human beings than folding laundry?
That question has been asked and answered by Generation Ritalin, who manages to find a comb once or twice a week. Clothing iron? What the fuck is that? Sounds like some kind of torture device.
Of course when robots can do everything our economic system will be completely different. Hopefully we catch onto that fact early enough to eliminate all the rich people before they can own all the robots and all the police.
Why, because we caught on early enough to prevent the Donor Class from gaining complete Control over Government? Our economic system will remain the same because humans will remain the same. Yes. That includes slaves. Just as it has been for thousands of years.
Re: (Score:2)
Clothing iron? What the fuck is that? Sounds like some kind of torture device.
Synthetic and synthetic-blend textiles often don't benefit significantly from ironing or steam pressing.
That includes slaves. Just as it has been for thousands of years.
Luckily for me, wanting to kill all the slave owners still puts me on the right side of history.
Re: (Score:2)
History is written by the winners. There is no right and wrong. Might makes right. It sure was wrong of western Europeans to come over and start USA and yet here we are. I don't see us all saying, yeah we should return to Europe and give back the land.
Pretty sure every country in the world likely has it's own story and in every one of those countries someone is going to feel wronged by the status quo.
too easy (Score:1)
Re:too easy (Score:4, Insightful)
i can see the folding being hard. clothes could be inside out. Or they could be knotted. And fabric is hard to work with in general because it is too fluid in a sense.
And if the clothe has cord dangling that could get stuck in the machine. It seems like a non trivial problem.
Re: (Score:3)
Jst seperate the fibres and use the raw material to 3D print a pre-folded item.
Get machines to do all the work. (Score:2)
It would be interesting to have Slashdot stories about new successes in automation.
Re: (Score:2)
Humans don't share very well. We can all dream of a utopia where robots do everything and everyone gets all their basic needs met while not having to be at all productive but even that has it's own flaws because people need to feel useful and many of us like to feel like we've worked hard to achieve whatever we have achieved.
We like to be better then each other and that doesn't work well with equal outcomes.
Congrats Pieter (Score:2, Interesting)
Met him in person at the lab after finishing CS 188.x Amazing achievement. Hoping all the work into this will improve all sorts of robots in the future and make the world a better place. It's wonderful to see someone met in person on a slashdot piece.
Let's celebrate! WOOT. Go Pieter!
But can it fold fitted sheets (Score:3)
Re: (Score:2)
this
Shirts and pants are hung, underwear are not folded, bras are festooned on doorknobs, door edges, draw pulls and any other projection, socks are paired. Towels and Tees are easy to fold and it's not so bad while watching TV.
But fitted sheets, Even though I've read how to fold them "properly" I think it's easiest to just roll them up and put them in a pillowcase along with the rest of the matching linen to make up a bed. Maybe they should change the design of fitted sheets or make a specialized robot to
Re: (Score:2)
Yeah, I had to try many methods of folding a fitted sheet. The only one that seems to come together in a reasonable fashion is to take the corners along the long side and put them together inside to inside then fold it over so you have one outside corner inverted and one outside corner the right way. Repeat on the other side (the same way), then combine the two corner-corners so all 4 corners are together, inverting the sheet as neces
Re: (Score:3)
Ignore the corners of the sheet. Fold it from the innermost points of the corner seams, as if they were the corners of the sheet. It becomes much simpler then.
Re: (Score:2)
According to the article, it can't fold any kind of sheets, or anything larger than clothing that a person would wear.
Re: (Score:1)
Re: (Score:2)
haha, tricks on you. You just ball that shit up and toss it in the closet with all the other properly folded towels and sheets. Trying to fold a fitted sheet just doesn't work well.
Roomba (Score:3)
It's going to be a long while before you can buy a Roomba to clean your floors, either. The best Roombas can do is speck patrol, or hair pickup.
Re: (Score:1)
Fear (Score:2)
A Roomba just keeps the dog cornered in one small part of the house.
Re: (Score:2)
Actually the floor cleaning roombas are not bad. Clearly the roomba vacuums are great and the cleaning ones aren't as good. But they work decently actually.
Re: (Score:2)
It only takes one cat vomit while I'm not home or otherwise a sleep and roomba is gonna ruin whatever area the cat nailed.
How about a robot that can vacuum a couch? (Score:3)
Re: (Score:2)
This is a great example of why robots aren't coming for our jobs any time soon. They can be built to perform one specific job (i.e., Roomba), but ask it to do something similar in a slightly different context, and no can do.
Oh, consumer robots (Score:5, Informative)
When I quit the laundry industry 20 years ago, we had robots to fold most anything that needed folding. A "French folder" for towels and uniforms that was 40 years old (at the time) did 6-1200 per hour.
But this are "general purpose robots"... Different than what was standard, "off-the-shelf" tools.
Folding (in general) has been well-known and "dealt with" for decades, where is needed to be done. They're just trying to make is "general purpose" for people who didn't need to do it before.
Re: (Score:2)
Re: (Score:2)
Usually (at that time) goods were hand-picked from the dryer loads and fed to the machines. Some took output directly from the ironers as input. But humans sorted everything (still do in industrial laundries).
We're talking small-batch loads here, from what it published. The "folder" determines what it's folding and works appropriately.
Got a dress? Fold dress. Got pants? Fold pants. Got baby's blanket? Fold blanket.
When dealing with the same mix of loads for 8-24 hours a day, it's (relatively) easy to make a
Re: (Score:3)
I found some footage from 3:50 to 4:30 of this video [youtu.be]. Are these machines similar to the ones you're familiar with?
I wonder if the first actual buyers of parts of this new technology, if it ever matures, will be industrial laundry facilities like that. If you could build a robot to sort clothing, or a robot to identify and grab two corners of a sheet - the parts where humans are still involved - could they make sense as replacements for those workers in a facility like that?
Re: (Score:2)
Ah, having modern equipment is nice. These machines take up a LOT of space though, but they work quickly.
You simply would not use anything like this if you were doing a household's laundry; the difference between items is too great (you don't fold baby clothes like a baby blanket). That's where the intelligent equipment will "do a better job", but will also have an issue with costs.
It will take a LOT to reduce the costs to where people would consider buying it. A folder like the first unit show will set you
Re: Oh, consumer robots (Score:2)
In my opinion the cost wonâ(TM)t reduce much (some due to economies of scale, but thereâ(TM)ll always be a lot of fancy hardware costing money in there). The value proposition will become good not from the cost reducing, but from the functionality increasing. Iâ(TM)m sure there are lots of people who would be willing to pay £20,000 to have a machine that did all their laundry. Maybe even £50,000 if it does the washing up, the vacuuming and the cooking too. There ar
Re: (Score:2)
When I quit the laundry industry 20 years ago, we had robots to fold most anything that needed folding. A "French folder" for towels and uniforms that was 40 years old (at the time) did 6-1200 per hour.
But this are "general purpose robots"... Different than what was standard, "off-the-shelf" tools.
Folding (in general) has been well-known and "dealt with" for decades, where is needed to be done. They're just trying to make is "general purpose" for people who didn't need to do it before.
Something like the machine the guy is feeding here [youtube.com]?
The folding robot isn't actually folding clothes as much as going through a predefined series of motions, and a towel caught up in those motions ends up folded.
Just like a conveyor belt is a robot for carrying items from one end of a room to another.
The difficulty comes when the robot needs to adapt those motions to a specific task and you suddenly need a huge pile of AI to do that adaption.
You're correct there may not be a big industrial application for ar
Re: (Score:2)
That video (first one) was showing off the unit picking up the towels and handing them to the operator. I've seen an operator match that speed with "just" the folding portion of the unit 30 years ago, with less movement (twisting to pick up next towel). But yes, that's kind of what it's like.
I'd be interesting what what they come up with. It's not going to fit the industrial market, and it has a LONG way to go to match the consumer market. It's certainly not going to fit in the common laundry space of the h
Re: (Score:2)
That video (first one) was showing off the unit picking up the towels and handing them to the operator. I've seen an operator match that speed with "just" the folding portion of the unit 30 years ago, with less movement (twisting to pick up next towel). But yes, that's kind of what it's like.
I'd be interesting what what they come up with. It's not going to fit the industrial market, and it has a LONG way to go to match the consumer market. It's certainly not going to fit in the common laundry space of the home, already filled with a washer and dryer (that seems beyond many people's operating skills).
So my line of thinking is that the only real market is the industrial market where the items are generally identical and specialized machines already do a fantastic job.
So there might not be a market for an awesome generalized laundry folding robot even if you could make one.
But for me, the more interesting question is how difficult a problem a generalized laundry folding robot is. And from this story it does seem like it's quite difficult to build a robot that can transform a basket of arbitrary items of c
Re: (Score:2)
I've heard of those... https://www.youtube.com/watch?... [youtube.com]
This highlights why self-driving cars FAIL.. (Score:2)
We humans take our cognitive capacity, the innate ability to reason, to 'think', for granted. We furthermore take for granted that somehow it should be trivial to build machines that have this same innate ability. Wrong.
WIthout understanding how our human brains, which have evolved over hundreds of thousands of years, have the capacity to think, we have no chance of writing software or building machines that posess that same cognitive capacity.
Meanwhile companies attempting to dev
About your sig... (OT) (Score:2)
Seriously, why do you think that?
Even assuming he deserves to, there is no realistic reason to expect that will actually ever happen.
Re: (Score:2)
So wait, are you saying you believe in intelligent design?
Because otherwise you have to accept the possibility that just like evolution, it will be stumbled upon by mere chance. Do you have any treason to think that something that we believe occurred entirely naturally cannot
Re: (Score:2)
Rather, he seems to believe in UNintelligent design: That humans cannot create anything novel, but only imitate something that already exists.
Re: (Score:2)
Which is pure bunkum if you think about it. Humans invented and/or discovered math, then went forth and attempted to do a bunch of stuff mathematics said was possible. Only when that started to peter out did humans then go back and take a serious look at doing things the way nature did them, which is sometimes elegant. (And sometimes seriously not, but that's what discernment is for.)
Re: (Score:2)
Nevermind trying, you can't do it, and neither can 'AI' companies, they have no idea either. Stop acting like it's trivial, and stop believing that what they produce right now is capable of that, it's not, and throwing more hardware and 'training data' at it isn't going to make it 'wake up' either. We aren't there yet and I don't see
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
I don't know, obviously... but that doesn't mean it cannot be done.
Unless you are alleging that nature somehow "knew" how to evolve mankind into something intelligent (ie Intelligent Design), there is no reason to assume that the same thing cannot be achieved artificially.
Re: (Score:2)
If nature can do it entirely by coincidence, what reason is there to think that human scientists could not stumble across the same thing, even before they necessarily understand exactly how what they've discovered actually works?
Your implication is that nothing that humans can ever do is accidental or unplanned.
Re: (Score:2)
Re: (Score:2)
While intelligent design theory is often said to be creationism in a thinly veiled cloak, the scientific failing of intelligent design is that like the simulation hypothesis, it lacks testibility, and can never follow the rigors that scientific theories undergo. It is not entirely orthogonal to belief in god in that if one believes that a god made everything, then one must necessarily believe in intelligent design, whereas if one believes in intelligent design, one merely only believes in the existence o
Re:This highlights why self-driving cars FAIL.. (Score:5, Insightful)
The people who work in this field don't assume any of those things. They know better than anyone how hard the problems are. But they also know something else: every last one of them is solvable. We're making steady progress. We've been making steady progress for years. It shows no sign of stopping. If anything, progress is speeding up.
They also know that general intelligence isn't needed for most real problems. Most real problems can be solved with specialized AI. Certainly folding laundry. Certainly driving a car. Ten years ago (probably even two years ago), I bet you would have insisted general intelligence was needed to create a program that parsed a text description and created a convincing image of what it described. I'm sure you would have offered lots of arguments for why that's impossible without understanding the text and the concepts that relate the text to the image. But here we are, with programs that produce amazing results.
At this moment, there are self driving cars on the road, carrying paying passengers without a safety driver. Not a lot, and only in a few places, but three years from now there will be more. Three years after that there will be still more. When a field is making steady progress, betting against further progress is a bad idea. You don't want to be one of the people explaining why something is impossible while other people are busy doing it.
Re: (Score:2)
Ten years ago (probably even two years ago), I bet you would have insisted general intelligence was needed to create a program that parsed a text description and created a convincing image of what it described. I'm sure you would have offered lots of arguments for why that's impossible without understanding the text and the concepts that relate the text to the image. But here we are, with programs that produce amazing results.
Have you actually tried any of these generators you are hyping so much? They are excellent at very well-defined classes of text-to-image. But for many queries they're quite awful and makes it completely obvious are doing little more than word association (essentially taking all words in the query, doing an image search and then smashing the results together using ML). For example, if you provide even a simple query where Person A VERB with Person B, the generators almost always give an amalgamation of both
Re: (Score:2)
If our state-of-the-art ML can't even understand basic queries, what chance does it have to fully self-drive a car with the infinite number of scenarios that might happen on the road?
Exactly my point. If it can't tell that a STOP sign with a sticker on it or some graffitti is still a STOP sign then it's not capable of 'reasoning'. Meanwhile a small child can tell the difference without any training at all.
Re: (Score:2)
Show me ANY research that shows they're getting a handle on how 'thinking/reasoning' actually works, and if you're going to point me at 'deep learning algorithms' or 'neural networks' then don't bother, those are at best just tiny parts of the overall picture.
Re: (Score:2)
I would argue
Re: (Score:2)
Re: (Score:2)
I believe I'm doing the opposite - not taking my brain for granted.
Most people don't question how they perceive the world, or how they process the input.
Turns out that humans have a poor mental map of the world they perceive.
Re: (Score:2)
All the DATA in the Universe is useless without a way to process it in a meaningful way.
That's what 'cognition' is all about -- and again: we don't really understand how that works. If we did, we'd have full-on General AI, walking and talking like the androids in I, Robot, 'positronic brains' and all.
Meanwhile all we have is just barely scratching the surface of how a brain -- any brain, human or otherwise -- actually works.
We don't even really have the proper tools to res
Fitted Sheets (Score:2)
The AI required for folding fitted sheets will only available by monthly subscription.
Not enough hands? (Score:2)
It would seem a two-armed robot would be tricky as a general purpose folder; you at least need a chin but three arms for a t-shirt (and four for sheets) with dexterous fingers should be fast. Using The Claw and a table can be done, but obviously it will be slower than a human doing it with five fingers and two arms.
Re: (Score:1)
Tesla (Score:2)
This should be a wake up call for all those people who think that Tesla is just a few years away from making a general purpose robot that will automate a bunch of human tasks.
Even folding arbitrary laundry is beyond the state of the art.
The first 50 years of AI/Robotics .... (Score:2)
....proved that all the things we thought were simple are difficult and all the things we thought were difficult are simple...
Chess, is simple ...
Folding clothes is difficult
note : Train yourselves for the jobs robots/AI can't do ... where humans are involved, or where the environment around the problem is most of the problem not a small compartmentalised task
Alternate headline (Score:2)
Alternate headline: Computers good at folding proteins, bad at folding clothes.