High Speed Robot Hand Dexterity [video] | Hacker News Reader

High Speed Robot Hand Dexterity [video]

youtube.com

61 pointsdirtyaura10 comments

I watch a lot of How It's Made. On many episodes you'll see an assembly line that is almost entirely automated, except for some small section. Sometimes that section involves (A) precision work that we haven't yet figured out how to automate. Most other times it seems to be (B) extremely "automatable" work (e.g. grabbing widgets six at a time and putting them in a package) that leaves me scratching my head as to why the robots aren't there as well.

I watch these episodes and think to myself that these workers are just the last portion of the line to be replaced by robots, and it's only a matter of time, capital, or both. Demos like this move the line between (A) and (B). The Industrial Revolution was only a half-step--just the "brawn" part. Now we're in the middle of the "brains" part, and the outcome is both exciting and scary.

My presumption was it came down to money.

Machines have a lot higher initial one off costs. The repair costs of some robots might make them more expensive per hour. Humans are tried and tested, they don't break-down and stop your production line.

When designing the production lines they must weigh up the cost of a machine at each stage (initial cost, elec, repairs) vs workers and decide a time-frame they want to project and compare the costs over (because of their higher initial cost the longer period you do your projection over the more likely a robot will ever be cheaper).

There are already factories that can run unsupervised. It's called "lights out manufacturing" and one of those factories is building robots:

"Robots are building other robots at a rate of about 50 per 24-hour shift and can run unsupervised for as long as 30 days at a time"

http://en.wikipedia.org/wiki/Lights_out_%28manufacturing%29

This is a fairly old video, I first saw it about 8 years ago. The interesting thing for me was the motors. Modern pick and place robots (surface mount circuit assembly) operate at speeds that make them invisible to the unaided viewer, and they are manipulating parts that are the mass of rice grains or smaller as well, with precision placement.

I kept hoping they would do a follow up which went beyond holding and went to multi-arm manipulation.

I came to comment to point that this demonstration is from may 2009 at ICRA[1] which is getting old by now (though not enough that you had seen it 8 years ago). It's the work of a research lab of university of tokyo called the sensor fusion project[2] and more videos of their work are available on the official page [3].

[1] https://ras.papercept.net/conferences/scripts/abstract.pl?Co... [2] http://www.k2.t.u-tokyo.ac.jp/fusion/index-e.html [3] http://www.k2.t.u-tokyo.ac.jp/movies/fusion_movies-e.html

Well I was doing mental math, a coworker at Google forwarded it to me while I was there, and I didn't think I had been there all that long ('06 - '10) so guessing mid-tenure at '08 and subtracting off that from '14 should have given 6 years but given [1] above I'm sure it was after the conference in '09. (Now added to my Evernote database so I won't make that mistake again :-)

That dribbling and re-grasping? oh man. Ugly bags of mostly water should be very afraid of the future. Meatbags are able to achieve similar level of skill, but not repeatability and endurance, not to mention the scale. Think about any Jackie Chan movie. We have one Jackie, and it takes him >20 takes to get some things right, but in the future we can have an army of robot Jackie Chans getting every single movement right at the first try.

Forget Bigdog or Atlas, turns out future will be exactly like AMEE from Red Planet.

FoxConn is trying to get robots to make iPhones. They are close but the precision isn't quite there. Will millions of workers be replaced in the next decade?

http://www.extremetech.com/electronics/195556-foxconns-robot...

https://www.facebook.com/video.php?v=982690585087334

My first HN LOL thanks!

Reply on news.ycombinator.com