Interesting, I'm pretty sure I understand what you are saying. Would the wheel assembly look something like this. This is an overhead view with the I's being the big wheel and the -'s being the small wheel. Also, if I kept the robot fairly square in dimension, wouldn't I be able to rotate without friction problems in tank drive? Thank you very much.
II
II
II
II
II----------- (the smaller wheel can be more foward or more back)
II
II
II
II

Check out this thread for many pictures of 'omniwheels.' In the white papers, you can find this AutoCAD drawing of team 226's crab module. The wheel they used is an omniwheel.

Any system that uses skid steering wrestles against friction. Even if you're rotating about a point, with one side in forward and the other in reverse, the wheels are still sliding perpendicular to their intended motion (i.e., not forward or backward). This creates friction. Belted tank drives exhibit the same problem, but there's more surface sliding along the floor, potentially making it worse. (See this thread for a discussion on how surface contact could impact robot performance.)

Many teams choose to live with these losses. It's a balancing act, though, because by adding additional load to the motors while turning, the current draw increases, and you risk tripping a circuit breaker.

Teams choose to minimize these losses in a few different ways. Omniwheels are popular. Casters are popular also, but aren't powered, traditionally. There are some issues with simply strapping a gearbox onto a freely rotating caster. Crab modules are another answer. These are, essentially, powered, controlled casters. They're considerably complicated to build. Tanks can pivot on a single point if the system is designed propely.

Hope that helps.

D.
Using your drawing as looking down from above the small wheels would be the II viewed edge on with the axle parallel to the II and the - would be the drive shaft. If you look closely in the attached pic you can see the little white wheels with "o" rings for traction.
BUPOD=Biased Uni Planar Omni Directional.
Biased meaning they roll in only one direction.

Al, here is an even better picture of your wheels. We loved them and copied them on last years robot. However we were totally unable to keep the O-rings on the rollers with the swerve in full effect so we switched to aluminum rollers with a heavy knurl pattern on them instead and that proved very durable.

Very great wheels, I recomend that any team with a fixed tank drive setup uses them on the rear wheels.

I would've automatically assumed to put them in the front, as that's where cars steers and all.

But, then again, I've never driven these robots in competition.

Why would you recommend putting them as rear wheels over front wheels? Or, am I reading into it too much?

Well I guess we don't have a totally great answer for why except that when we used them on the front the robot steered 'about' the rear wheels which made it hard to grab the goals last year.

When they were on the rear wheels the robot spun about the front wheels similar to a fork truck.

We also tried them on all four wheels and the robot spun about its center but it didn't have as much pushing force as dedicated traction wheels on two of the four.

Oh, I think I get it now, those white roller thingys alleviate the side friction problem. How did you guys go about manufacturing this and what exact materials did you use.

Matt, thanks for the better picture. I am sure I have one some where but that was the best I had at work.
Construction... The large wheels consist of a sandwich of 2 aluminum pieces in which areas are relieved to hold a small axle for the white rollers to turn on. The rollers had slots cut in them for the o rings and then the o rings were glued in place. Each drive module had two assemblies because we found that the rollers were not in contact with the floor long enough to grab. Two assemblies next to each other but offset, allowed at least one roller to be in contact with the floor at all times. The rear drive set used normal wheels and performed the drive for steering. Note that a normal tank style drive with this arrangement allowed responsive steering without the side friction problems.
This robot was for the ramp competition and the drive was strong enough that we were able to drive around with another robot on our back and drive across the ramp.
We have discussed various designs to replace the roller/o-ring arrangement but have not used this design again. A roller made from aluminum with some kind of tooth cut in the surface would be ideal though.
I should mention that our mechanical engineer, Raul, is the one who is responsible for this design. It was in direct response to the electrical team's research on high current demands in tank drive systems.
Just a few more hours...

2001
 

Here is a design for a trick wheel that we used on our 2000 robot.

http://www.theforumisdown.com/uploadfiles/1102/trick%20wheel%201.jpg

(copy & paste the link)


edit: fixed link
edit#2:fixed url parse

Look in the white papers under team 226 crab drive for the plans on how to make the plates for the omni-wheels.

For what it's worth, I've 'adapted' those omniwheels into an Autodesk Inventor assembly (R5). I changed the bolt hole pattern around shaft so it would meet my preferences. Anyone with a little working knowledge of Inventor can change that easily enough.

Forgive me, Matt, if I bastardized your design. If I did, please let me know. I want to know the wheels will work if we decide to use them :)

(The file is a little too large to be an attachment. If you're interested in it, PM me and we'll figure out some way of getting it to you. It's really nothing all that exciting, though. It took me about 20 minutes to make.)

Not my design M ;) it was lifted from Wildstang and I think it is yours to do with as you desire.

Well, I lifted the design from your team's CAD drawings in the White Papers section. So, thanks for those, and thanks to Wildstang for making them in the first place, and thanks to cavemen for inventing the wheel :)

I managed to upload it to FTP.

http://www.magenet.com/~imagination/...Wheel_INV5.zip