Fluid Actuators from Disney Research Make Soft, Safe Robot Arms | Hacker News Reader

Fluid Actuators from Disney Research Make Soft, Safe Robot Arms

spectrum.ieee.org

110 pointsspectruman24 comments

The section of the video where the robotic arm "plays ball" really impressed me due to the kind of playfulness which it exhibited. Beyond being impressed by the fluid actuator system I was just starting to be amazed at the programming involved with making a robotic arm that can show such accuracy to "catch" the ball and roll it back but also do it in such a playful manner..... then the camera zooms out and shows that the arm was being controlled by a human. Damn.

* I'm at working and watched the video with out sound, so maybe I missed as it was explaining what was going on.

The fact that you can have something responsive enough to mirror human movement is very impressive.

You'll be amazed when you see the video of this device. It's accurate enough to mirror human hand writing!

http://en.wikipedia.org/wiki/Pantograph

I was surprised by that too, but then later in the video it shows the other arm that was being controlled by a human hand. It appears to mirror that arms movement, which is why it looked so human.

My take on it, which might be wrong, was that the units can learn repetitive motion from another tracking bulb (in which case, it was learning from the girl playing the musical toy, copying her motions).

No, they use zero electronics it's a pure analog system for now.

That's a nice way to make a low-friction pneumatic or hydraulic cylinder. It's not original with Disney. Rolling diaphragm seals for pneumatic and hydraulic cylinders have been available commercially since the 1950s. ("http://www.marshbellofram.com/belgas/files/2012/05/design_ma...) It was used for humanoid robots as far back as 1985. ("http://cyberneticzoo.com/robots/1985-mcandroid-jon-barron-et...)

There's no power source in that demo. That's an un-powered Waldo, with the manual operator providing all the power. If you put that mechanism in a robot, you still need a motor, in addition to the four pneumatic cylinders and plumbing for each joint. For Disney's anamatronic applications, where the character is fixed in place and can have lots of plumbing to an external control unit, it makes good sense. For working robots, not so much.

Your getting a 404 from an extra " at the end.

http://www.marshbellofram.com/belgas/files/2012/05/design_ma...

http://cyberneticzoo.com/robots/1985-mcandroid-jon-barron-et...

This pneumatic power design could improve working robots. Yes, you would still need a motor to actuate each joint, but now you can put the motor anywhere. Placing the motor in the torso removes weight from the arms, which means you need less power to control the arms and the arms have less momentum, both of which greatly increase safety.

True. You can also get rid of the motor entirely and just use proportional valves and compressed air. Festo's robots work that way.

A common model of human and animal muscles is to treat them as two opposed springs, where the spring constant and neutral point are both adjustable. Mechanical systems for that have been built for research robots, but they're mechanical nightmares of cables, pulleys, and springs. Two opposed air cylinders, though, can directly implement a pair of adjustable springs. That's been done a few times on research robots. This gives you muscles of adjustable softness, just like humans. A running robot with spring-type energy recovery could potentially be done that way. (Humans recover about 60-70% of energy from one stride to another when running. Cheetahs, 90%.) Mechanical running machines with springs for energy recovery have been built. With fixed springs tuned for running, all they can do is run.

To make soft air springs work, you need position and pressure feedback on everything, and small, fast servovalves. Fast servovalves tend to be very expensive. They also have to be near the cylinders for good position control - too much air line and there's no way you can make the system stiff when you want to. Of course, you need compressed air.

Meanwhile, brushless DC motors have improved so much, and are so easy to control now, that the trend is away from pneumatics and hydraulics. Boston Dynamics' humanoid is all hydraulic (it's really a modified Big Dog), but Schaft's is all electric. Google now owns both, of course.

If you liked this, check out another project from Disney labs called "Touché" -- http://www.youtube.com/watch?v=E4tYpXVTjxA

For some reason, the girl's reactions to the 2 bzzzt-wrong sounds @ 2:55 are hilarious.

What is Disney's interest in this type of technology?

Disney has a whole R&D division to develop theme park attractions: http://en.wikipedia.org/wiki/Walt_Disney_Imagineering

Alan Kay and Danny Hillis both used to work at Walt Disney Imagineering

Disney has always done a lot of work in animatronics for their movie special effects as well as park attractions. Being a current or former imagineer is quite a badge of honor in many circles. https://en.wikipedia.org/wiki/Walt_Disney_Imagineering

This was done by Disney Research http://www.disneyresearch.com/ . A lot of their work relates to robotics and display technology, as well as user input. I suppose the idea is that this kind of thing could be ultimately used for entertainment robots and exhibits like they have in the theme parks, or possibly movie special effects.

That was truly bizarre. Why would you use this device to make sure someone used the correct method of eating food?

The big deal appears to be that the're using hydraulic cylinders with what's called a rolling diaphragm: http://machinedesign.com/hydraulics/rolling-diaphragm-seals-...

It doesn't look like a particularly new innovation, but perhaps something that's not commonly used in robotics?

I've used them in robot arm prototypes, but they're basically 3x as big as regular cylinders (1.5x for the rolling diaphragm overhead, and another 2x because you need two single-acting cylinders instead of a double-acting one).

If you're interested in this, I highly recommend looking at the stuff that Festo (a german industrial automation company) is doing: https://www.youtube.com/watch?v=u7KIp7A8iqs

(This is from 2008).

And not necessarily related, but really cool anyway: https://www.youtube.com/watch?v=nnR8fDW3Ilo

It’s not a robot, it’s a Waldo¹ – a puppet, essentially. Which is fine for Disney, which wants to make puppets for puppeteers, but this has basically nothing to do with new developments in robotics.

① https://en.wikipedia.org/wiki/Remote_manipulator

What, aside from the reduced mass of the robot arms, makes it safe? How, aside from operating pressure, is this different from traditional hydraulics?

I'm afraid I just don't see the novelty of a low pressure hydraulic system.

I believe the new part is the "rolling diaphragm" actuators, which have a high torque density.

And very low stiction compared to pistons.

FTA:

"With pneumatics and hydraulics, you can certainly get power and speed, but you often need to deal with leaks and friction and elaborate arrangements of pumps and valves and whatnot."

Is Disney Research really making cast parts for their prototypes, or are these is this off-the-shelf technology from somewhere else?

http://spectrum.ieee.org/img/fluidtrans2-1412583381292.jpg

[edit] Nevermind, in other photos it is clear that the parts are machined. I'm still surprised they've gone to such lengths to remove weight in the prototype, but that's more of a cool thing than something truly unexpected.

Reply on news.ycombinator.com