High-Speed Robot Hand Shows Dexterity and Speed

An anonymous reader tips a blog posting that begins "A few blogs are passing around videos of the Ishikawa Komuro Lab's high-speed robot hand performing impressive acts of dexterity and skillful manipulation. However, the video being passed around is slight on details. Meanwhile, their video presentation at ICRA 2009 (which took place in May in Kobe, Japan) has an informative narration and demonstrates additional capabilities. ... [It] shows the manipulator dribbling a ping-pong ball, spinning a pen, throwing a ball, tying knots, grasping a grain of rice with tweezers, and tossing / re-grasping a cellphone!"

Comments

holy crap!

[ILuvRamen]

Watching that, I can't help but think that the math/programmer behind it is fairly simple with a basic knowledge of physics and the only reason this wasn't around 10 years ago was chips not being fast enough. Now that the technology exists, all the theoretical robot motor function logic can actually be put into practise. If you think about it, the dribbling code must be really simple. Run the numbers while the ball is in the air about precisely where and at what angle the ball impacted the first finger at and position the third finger accordingly and make it swing with the proper force and angle to bounce it back to the first finger. It's like programming pong with a few extra lines of code and raw date from sensors lol. Okay that was a bit too simple as a comparison.
By the way, is it just me or does anyone else think that it failed some of those test on the first fifty takes lol. I think it's not quite as accurate and perfect as the video makes it out to be.

Re:holy crap!

[sexconker]

No. A closed loop can be perfectly fine forever.

If the loop is:

Track ball
Hit ball when in range

Then as long as all hits result in the ball being in range, you're good.
You don't need to measure the error of the last hit - just get the next hit right. Always aim for the most optimal spot for the next hit.

If the ball ends up further and further away from it's optimal spot with each hit, then the hits are in error (in terms of aiming, timing, speed, etc.).

You don't need to know what happened last time to be able to correct an error. Knowing what happened last time won't necessarily help you either.

If the required correction is outside of the robot's (physical) capabilities, you will always fail.
If the required correction is within those capabilities, a feedback loop can help you recover, but the original error still came about due to non-optimal hitting/tracking algorithm. If you can't trust your algorithm, you can't trust it to fail in a correctable way.

Fix the hitting/tracking algorithms (and all physical procedures they extend to) and avoid the errors altogether. No need for a feedback loop, and you get a much more reliable robot.

Dribbling demo?

[QuoteMstr]

Very impressive. I wasn't able to quite tell from the video, though: was the end of the dribbling demo planned, or did the robot lose control of the ball after a few seconds?

Very nice. Some things are easier when done fast.

[Animats]

This is very nice work. The most interesting result is that some manipulation problems become easier if done fast. In the short term, inertia makes the motions of objects very predictable. With millisecond reaction times, that can be exploited.

Fast machinery isn't unusual, but it's rarely that smart.

Re:Skynet

[93 Escort Wagon]

However, if and when a robot learns to play Eddie Van Halen's "Eruption", then I'll worry.

What if the robot can do something that Eddie can't, such as keeping Valerie Bertinelli happy? Would that make you worry?

new experiment

[StripedCow]

i have an idea for a new experiment: fire a bullet at it, and see if it can catch it :-)
it would be the ultimate body guard :-)