Berkeley Gets Willow Garage Robot To Fold Towels 99
kkleiner writes "Researchers at UC Berkeley used Willow Garage's PR2 robot to fold towels. The UCB programming used some innovative visual scanning techniques, allowing the PR2 to pick up a towel, find its corners, and fold it on a table perfectly. According to the paper presented at the 2010 ICRA (PDF), the robot successfully completed 50 out of 50 attempts to fold a single towel, and also folded 5 out of 5 towels when they were presented in a group. Is watching a robot do laundry really that exciting? Hell yes — wait until you see the video! UC Berkeley used a Willow Garage robot to develop their own sophisticated robotics program. That validates the whole premise of the PR2 — faster development by letting researchers use a common platform. Score one for open source robotics!"
Hotels (Score:5, Insightful)
I'm sure there are some hoteliers that will be excited about reducing their staffing for for washing and folding all the towels and sheets they go through. Hospitals likely would love this too, since it wouldn't show up sick and help spread diseases on clean linens.
Re:Hotels (Score:5, Insightful)
You don't need this level of sophistication just to fold towels. The robot could just kind of flatten it out by tugging on different corners until it's flat, and then grabbing two corners and folding it over.
Even easier would be just a big machine that pulls in a bin of linens and separates them onto rollers, which deposit them in piles.
TFA states that the robot is not the best model to fold towels. It does demonstrate "open source" robotics. At one time, computers were dedicated, single use, machines. Now, an off-the-shelf robot can be programmed for various tasks. That's the point.
Re:The video seems impressive... (Score:2, Insightful)
Eerily Creepy (Score:2, Insightful)
I found it very creepy. The way it handled the towels and turned them while 'looking' for the next step. It was reminiscent of what I felt was a child learning to fold towels (although, I'm fairly certain the robot wasn't doing any learning). For whatever reason, and despite it's appearance, this robot seems more human than any other robot I've seen previously.
Fitted sheets. (Score:1, Insightful)
If it could fold fitted sheets I'd be impressed.
Re:Hotels (Score:3, Insightful)
Yet the robot would still need to find the corners.
Whoever doesn't think this is amazing needs to pay attention to a young child sometime. This thing has more programmed dexterity than a 3-year-old: my daughter isn't stupid or anything, but I doubt she could neatly and consistently fold a towel or washcloth. Ask any parent: having young children "help" with the laundry ends up being more work.
Re:Excellent (Score:3, Insightful)
Don’t you mean a 25% chance? C programming and folding socks are uncorrelated skills and if we assume there is really a 50/50 split with no gender bias that would give a 25% chance that she is both a better coder and worse at folding socks.
Or are you implying that C programmers can’t fold socks? As someone who has written a few C programs and folded a few socks, I find that outrageously offensive.
Re:Foresight on the towel choice (Score:3, Insightful)
Well, sure, it'd make it easier, but now that this technology has been shown - technology that can fold a solid-colored towel or a multicolored towel or anything - it must be developed and furthered as-is. If companies were to try to lock us into their towels so that our robots and towels would be compatible, we'd have comparisons to Microsoft and complaints about technology being held back and whatnot in an instant.
If we know that the robot can fold any towel, any color, any pattern, then that's what has to be developed in order to look good, impressive, or any other adjective that would be favorable to a manufacturer.
Re:open source...fast? (Score:2, Insightful)
Your linked videos are all examples of extremely impressive hardware and motion control. You don't want a PR2 if you're interested in motion control and dynamics problems, you want one to interact and manipulate in a similar manner and with similar capabilities as a human torso and arms. The PR2 platform isn't for industrial use, and it's not supposed to walk. It's to give researchers a common-ground solution with extensive software to work on AI problems with, not traditional control problems.
The industrial arms are very accurately playing back a path that was programmed in by hand - I certainly don't want to downplay how impressive that demo is, but there's not a lot of 'intelligence' about 3D motion control. The pick and place machines do need some very basic vision tasks to identify and track the targets on the assembly line, but in general the problem is solving motion control (plus with the Flexpicker I think a terrific mechanical design helps.) Likewise bipedal locomotion is a difficult problem, and the dynamic stability of the humanoid robot is a great feat, but again a different field from the towel folding task. To balance/run as in the video requires no external sensing, only internal inertial sensors and a pre-programmed gait (technically it's not running, to run both feet need to be off the ground at the same time - watch closely, it's actually a fast walk.) I think they are all very impressive, and I have no wish to imply that they are any more/less impressive than towel folding, just that they are different problems.
Yes, I agree, a 50x speedup is painful and completely impractical. The cool thing here is the object recognition and manipulation, especially with soft, flexible objects. Identifying a random towel, picking it up and finding all four corners, then grasping two adjacent corners and folding twice, all dynamically is not an easy task. There are tons of easier and faster ways to fold a towels automatically, but none of them could work from an untidy pile of different types of towels piled on a table - they'd all need some special loading mechanism and take up half a room. In theory this robot could wander around a room actively looking for towels people had left around and do the same thing - that's what makes it interesting. And no, I'm not saying we should ever plan on using expensive humanoid robots solely to pick up and fold towels, it's just cool to know we're one step closer to the day when they -can- do it.