First off, thank you!
Second off, sweet CAD!
Third, questions/constructive criticism (see how we can do that now?
)
It appears that there is nothing that retains the ball vertically other than the robot's weight. I can assure you that even with a level field, you'll want to address this. The robot can be lifted up in collisions (not to mention carrying it on and off the field). If this is simply not yet modeled, disregard this comment.
What kind of bearings will you be using to support module rotation?
What kind of traction material will you use on the ball, and ball/roller interface? Fears of slip/low traction have always deterred me from ball drive.
The idea of shifting roller sizes on a ball drive for shifting is very cool, though I agree that as modeled, I don't see the advantage over a swerve with traditional dog sifters. I feel like this may have the potential to be more compact though, maybe by shifting the CIM down low, behind the ball. If there's an advantage I'm missing, please enlighten me though!
The initial module you posted looked like it had a lot of promise as a traditional wheel-switching shifting design. While not totally original, your design work was sound and compact. Maybe pursue this as well.