105Views15Replies

Author Options:

Next step to the robotic house-maid... Answered

OK, so it's a fairly simple task for homo artifex, but a robot is now capable of folding towels.

It takes two minutes of sped-up video to fold the five towels (100 minutes in real time!!), but this is a huge breakthrough in robotics, because towels are not rigid, predictable objects.

The towels are different sizes, different colours, and start off screwed up.

You may be bored, but I'm impressed.



Berkley Website article

Still image from dot.Maggie

Discussions

That's quite a feat of design and engineering!

See my discussion below.  The slowness is there both due to the complexity of the visual analysis, and to allow the researchers to closely monitor the motor sequencing.

I fully agree with you.  As hobbiest robot designer, (as of right now)I under stand how difficult, and thus how amazing it is to be able to accomplish such a feat. I was mearly pointing out that it would not ba a time saving device for home use. (yet!)

What next?
Sorting whites from colours would be a good one?

L

I'd like it to be able to clean up a large table full of dirty plates, clean plates, uneaten food and spills!

Big dramatic table-cloth pull, the china stays where it is and everyone claps?
I would post you a "Bender get's made" clip, but YouTube is too busy right now...

L

That's pretty busy. Possibly DDOS attack by some forum?

20 minutes per towel!!! I think I'd rather just stick to doing the work myself. Maybe I'll get one of these robots once it can fold a towel in a minute/less... 

Cool! I hate folding laundry. Make my robot slave get here faster!!

ok...it's a robot doing a human job...can we anthropomorphise it yet?

i think it's a he....i may be biased since i'm a he and i do all the laundry in my house.

That is really impressive.  The 50x speed-up video make the robot's operation look almost like a human towel-folder, and it shows very clearly just how complex the visual analysis problem is! 

Running the actuators slowly allows the researchers to capture, analyze, and validate the detailed motor-control instructions generated by the robot.  If they ran the robot at human-speed, it would be much harder to verify that it was behaving as intended.

Finally, I love the dry academic title, Cloth Grasp Point Detection Based on Multiple Geometric Cues.

almost 2 hours for 5 towels? jeebus...hehe

it looks almost inquisitive while it's inspecting the towel for corners. like it's really really trying to figure out what it's doing.

it looks like a relatively simple pattern as far as what it's doing (not saying it's simple to make a robot do that). grabs the towel as a pile of something, inspects the edges for a corner, grabs the corner and looks for a second corner (which now should be easier to find since the towel has draped out to full length), checks to make sure that it's got both corners held the same, then does a simple fold.

if you really break it down, that's how most people actually fold towels or sheets, blankets, or any other rectangular foldable. of course we do it faster and mostly from trained actions. but i know when i'm folding laundry i basically follow the same steps, find a corner, find another corner, make sure it's not twisted, then fold it.

I don't know why, but this made me giggle uncontrollably. SO AWESOME!