Slashdot Mirror
Making Robots Mimic the Human Hand
RougeFemme writes "As part of a research project to develop low-cost artificial hands, DARPA has developed a two-hand robot that can almost change a tire. Research has mastered grasping objects with robotic hands; the next objective is to manipulate the objects once grasped. Research also continues on a neural interface, a direct link between a robotic arm and the human brain. The ultimate goal of the research project is to develop prostethics and robotic arms for wider use, by reducing cost and improving dexterity and machine vision."
Almost drive a car like autopilot can almost fly on it's own.
Just the thing I need when I almost need a tire changed.
The shepherds did so well protecting the flock that the sheep no longer believed that wolves existed.
I can almost watch the video. Come on ny times is an mp4 really that hard? 30% of all Internet traffic is on a mobile device.
A fool throws a stone into a well and a thousand sages can not remove it.
DARPA has had some of those two-handed robots for a few years now, and has researchers working with them. So far, the results have been disappointing. They can pick up a screwdriver, but not use it. Robots manipulating a powered socket wrench in a well-defined environment aren't a big deal; auto assembly lines have had those for years. Putting a key in a lock, slowly, is about the upper limit of that project.
Part of the problem is that simulators for manipulation aren't very good. Willow Garage is funded to take the current version of the Gazebo simulator and make it work for manipulation. But I doubt they'll be able to make it good enough to do manipulation using force feedback. The physics engines they're using are for game physics; they cheat on frictional contacts. You need a really good simulator to debug control software for putting a nut on a bolt by feel.
I do a lot of actual work these days (as opposed to key-punching) and you might say I am highly in touch with the amazing abilities of the human hand. Sex life aside, the sheer sensitivity and flexibility of the hand is awesome to consider. There's no robot yet that can pick arbitrary parts out of a bin (or, say, out of some gravel) and hold them in one part of its gripper while another part feels for the part's mounting location, then advance the parts into place (rotating them into the proper position as they go) and finally be able to push, twist, etc. in such a broad variety of fashions in order to install it... and then follow that trick up by picking up a tool to finish the job.
We're a long, long way from a good mimic of the human hand. It's not clear that anything in between a good one and what we have now is particularly useful as compared to purpose-made manipulators. Robots aren't that great at navigating varied terrain anyway, so using purpose-built robots isn't a big problem. We have to deliver them, regardless.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
see: http://www.hizook.com/blog/2009/08/03/high-speed-robot-hand-demonstrates-dexterity-and-skillful-manipulation
I agree the human hand is a marvel, along with hand-eye coordination. But these sorts of technologies are rapid displacing the value of much paid human labor in different areas of the economy. Creating factories generally means re-engineering most tasks so they fit what machines can do. Service industries will also do that more and more (like the US post-office automates, fast-food places automate, hospitals automate, people get household vacuuming robots or buy prepared food, etc..) This means our economic social contract is breaking down where an adult's right to consume was linked to selling his or her labor in the market -- as predicted decades ago:
https://en.wikipedia.org/wiki/The_Triple_Revolution
A basic income is one way to address this, but there are a mix of others.
A 21st century issue: the irony of technologies of abundance in the hands of those still thinking in terms of scarcity.