Hemispherical Omnidirectional Gimbaled Drive

[Molten]

Originally Posted by Joyride_67_1902

Sooo I went and spent a few hours CADing and went through two iterations of gimbles.

If you can do that in just a few hours, what can you do with the rest of your day? Seriously impressed, can't wait to see how it all turns out.

Jason

[JesseK]

Talk about rapid prototyping ... Bacon seems to be on top of it! John should post this on his blog...

Originally Posted by Lil' Lavery

Do you have any calculations, even first order estimations, to support this hypothesis?

Scaling up to large-scale applications (FRC, heavy-lift mobility, etc) has some foreseeable issues. The first and foremost is the noticeable skidding of the hemisphere as the robot changes directions. Even changing the speed/torque by rotating the gimbal causes localized skidding on each individual hemisphere due to the differences in radii that create the different "gear ratios". I don't know of every way to reduce the effect, but it implies that each individual application will have to account for the effect. The easiest way to account for it seems to be finding the best balance of traction & intentional slip. Another way could be to limit the rotation speed of the gimbal.

I like how it's a 'resurrection' of technology.

[dag0620]

Very, very cool.

I would hope that someone attempts to make this happen at FRC soon. I understand there are some initial issues with making it happen, but I'm sure its only a matter of time before someone figures those out and makes it happen.

[billbo911]

From 1938 (not team 1938).

[SenorZ]

Joyride: very impressed. My only critique (and this is coming from someone who will probably never build a HOG drive) is that the square frame supporting the drive motor seems a bit large. This would limit the total angle of rotation. But you have two hemispheres, so you wouldn't need as much power from tilt from each wheel to get a faster speed.

[Lil' Lavery]

Originally Posted by JesseK

Scaling up to large-scale applications (FRC, heavy-lift mobility, etc) has some foreseeable issues. The first and foremost is the noticeable skidding of the hemisphere as the robot changes directions. Even changing the speed/torque by rotating the gimbal causes localized skidding on each individual hemisphere due to the differences in radii that create the different "gear ratios". I don't know of every way to reduce the effect, but it implies that each individual application will have to account for the effect. The easiest way to account for it seems to be finding the best balance of traction & intentional slip. Another way could be to limit the rotation speed of the gimbal.

Absolutely true. I wasn't questioning the notion of there being issues with scalability of the concept, but rather that the statement was made without even a modest attempt at calculations to prove his (very different) reasoning.

But building off of what you said, I'm curious as to how this would behave with softer tread materials on a soft surface (as likely to be the application in FRC). The deformation of both the tread and carpet under the weight of your robot would result in a contact patch larger than a single point (obviously). Because of that, different points in contact with the carpet are going to be moving at different speeds (and with different torques), meaning that there's going to be some slip in the drive of the wheel on the carpet. Depending on the radius of the hemisphere, the difference in speeds from one end of the contact patch to the other may or may not be marginal. It's possible that portions of your contact patch might interact with the carpet at closer to their coefficient of kinetic friction than their static friction.

In order to mitigate that, larger hemispheres could be utilized. But then the size and range of motion of your gimbal must increase accordingly, as well. Pretty soon that's eating up a considerable portion of your volume, especially compared to traditional drive systems.

[Andrew Remmers]

Originally Posted by SenorZ

Joyride: very impressed. My only critique (and this is coming from someone who will probably never build a HOG drive) is that the square frame supporting the drive motor seems a bit large. This would limit the total angle of rotation. But you have two hemispheres, so you wouldn't need as much power from tilt from each wheel to get a faster speed.

Actually (for the most part) right now the "wheel" contacts first before any metal while its tilting. The current "wheel" does need to be made just a hair bigger which will get rid of this problem entirely.

[Andrew Remmers]

As promised I have pictures. No video since we didn't get to running it.

If you have problems looking at the album just post and Ill upload the pictures!

https://plus.google.com/photos/10171...92528705?hl=en

You may notice some things have changed from the CAD, this is because we didnt have enough of the right metal to build it to scale so we improvised a bit!


Enjoy

- Andrew

[Brandon Holley]

Originally Posted by Joyride_67_1902

As promised I have pictures. No video since we didn't get to running it.

If you have problems looking at the album just post and Ill upload the pictures!

https://plus.google.com/photos/10171...92528705?hl=en

You may notice some things have changed from the CAD, this is because we didnt have enough of the right metal to build it to scale so we improvised a bit!


Enjoy

- Andrew

Very nice thus far! Keep up the good work and definitely post some videos when you get it running.

-Brando

[JesseK]

Originally Posted by Lil' Lavery

The deformation of both the tread and carpet under the weight of your robot would result in a contact patch larger than a single point (obviously). Because of that, different points in contact with the carpet are going to be moving at different speeds (and with different torques), meaning that there's going to be some slip in the drive of the wheel on the carpet. Depending on the radius of the hemisphere, the difference in speeds from one end of the contact patch to the other may or may not be marginal. It's possible that portions of your contact patch might interact with the carpet at closer to their coefficient of kinetic friction than their static friction.

Hmm, you made me think of the AT&T logo when you said this. It's a sphere with ridges. Translate that to this application: if the ridges were rigid, all of the deformation would be on the floor/surface. A hard rubber might work well. Hopefully the builder wouldn't owe AT&T a royalty for every wheel made .

Interestingly for VEX/VRC, if one used a hard/durable material for the tread, without ridges, all of the wear would happen on the rubber tiles.

[JamesBrown]

Originally Posted by davidthefat

It seems that it would be hard to scale up. Scaling up to FRC level would be fine, but to a commercial level would not be all that easy. What I imagine is that the wheel platform would just snap. It's like trying to stand on one leg and trying to rotate your foot. I think there would be too much torque acting on that to be safe.

Commecial robots vary greatly in size, one company, FANUC makes robots designed to handle maximum payloads from 5kg-1200kg, obviously if these were made mobile they would require significantly different Drive systems.

[billbo911]

One thing I would like to know is, where did the gimbal he used come from? Did he have to fabricate it himself, or is it COTS?

[Greg Needel]

And we come full circle again....... Now in the original discussion nobody talked about changing angles to get holonomic motion but now that you see one built it makes complete sense.



http://www.chiefdelphi.com/forums/sh...ight=cvt+wheel

[Aren_Hill]

Whats to say you couldn't plant a ball caster right in the middle for an FRC scale one? to idle freely

[PAR_WIG1350]

Originally Posted by Aren_Hill

Whats to say you couldn't plant a ball caster right in the middle for an FRC scale one? to idle freely

You could do that, or you could just use a nylon/acetal hubcap. But there are still other issues that must be resolved.