This is an improved version of a previous design, http://www.chiefdelphi.com/forums/showthread.php?t=147276

Most changes involve trying to solve the problems stated by this extremely helpful post by SoftwareBug2.0, http://www.chiefdelphi.com/forums/showpost.php?p=1574346&postcount=14

The design uses ball caster wheels a couple inches from the center of the top of the basketball. This wedges the ball into the 6" wheels in the center of the robot. Also, ball wheels designed to hold the basketballs in the robot have been added.

What was the philosophy behind caster wheels over something like omniwheels? Wouldn’t omniwheels be better for rotational friction reduction?

Packaging, I’d say. Ball casters are very compact; omniwheels… finding anything smaller than 4" is very very difficult. And if space is critical, the less you use the better. I’d call space critical on this drivetrain.

Agreed, but I don’t see how this would avoid scrubbing on the higrip wheels while moving without using omnis.

There isn’t any scrub. The wheels are in the middle of the ball vertically, so if one wheel spins, it rotates around the other. And if both are spinning to go diagonally, the axis of rotation lines up so both wheels have no scrub. I’m not sure, though, how it would work out if it wasn’t a 45 degree diagonal, but rather something like a 60 or 30 degree diagonal.

Depends on what you mean by “scrub”.

It looks like there are 2 traction wheels pressed against each basketball.

Consider the simple case where only one of those traction wheels is spinning.

There will be considerable pivoting friction between the ball and the other traction wheel. It may even bore a hole in the basketball.

There axis of rotation of the ball would not line up with the wheels when both are spinning though. Here is a quick sketch of where it would be in relation to the wheels: https://docs.google.com/drawings/d/11VqwdAdUHyRPn18JtLDMhWrUNKzKlGVhxnDOI4KIXro/edit