Our intramural game build has been interrupted by our need to shift to a new build space. It's great news in the long run, as the room will be full-time team space, but we had to move everything over in a very few hours with no warning and then figure out how to re-organize in the space.
As I mentioned above, "team nameless" decided to build a kiwi drive, more for the practice and wow factor than real engineering requirement. Due to the KoP contents and the interesting budget rules (extra items pulled from team inventory count half towards the "extra parts" budget), they decided to build their kiwi using three AM14U2 gearboxes (TB minis with long hex shafts and an 8.45:1 gear ratio) and duraomni wheels mounted on VersaFrame. They thought they would have to make at least one of the motors a mini-CIM in order to fit the 18" square game rules dimensions. Their first attempt was sufficiently disheartening that I decided to "prove to myself" that this could be done. I drew up the following three designs in power point (never learned real CAD). All are pictured from below:
This one seemed most like what the team wanted to do. It is actually 17.5 x 17.66 (9 + 5 √2) inches rectangular. I mounted the TB-minis through the upper pair of holes in the base plate, and cut oversize holes to pass the cap screw heads through. The center really did look like that, telling me that the triangle was essentially equilateral. Do note, however, that the CoG should be located about 10" from the top of the pic, which is a bit over an inch below the geometric center of the rectangle. I also figured that if this were to be a competition robot, I would probably rivet the entire perimeter, but use bolts for the "cross beams" on which the motors are mounted. This is because of the difficulty in accessing the through holes; it was far easier to mount the gearbox to the cross beam, then install it in the chassis than to install the gearbox in place. Also, note that I only needed to drill six through holes in the vertical direction (indicated in red). If you wanted the outer frame to be a round number of inches, you could do the same thing by only drilling the holes for the 30-degree gussets needed at left center and right center. And yes, I was duly impressed with the rigidity of the resultant frame, especially with the triangles in the corners.
I also figured that with the omnis close in to the gearbox, I could get away with no end support on the axle. If a bit of support was needed, plan A was to put a hexagon of thin sheet aluminum and three short pieces of angle aluminum so that the three gearboxes pulled against each other to keep the axles horizontal. If this proved insufficient, I would have had to go with a versaframe bearing support (with the
versaframe side bearing mount being the favorite).
Here was a classic kiwi drive, implemented with versaframe and the same gearbox and wheel. The black box is 18" square, to show it fits within the specification. This one was just a proof of concept; on an actual robot, I would have cut the ends of the long pieces of VF on a 30 degree miter to simplify bumper construction.
This is a design I came up with a few days later. I call it a "U" drive. It's essentially an H-drive (slide drive), but without the belts or chains, because there are only three wheels. The relative loading on the "forward/reverse" wheels vs the strafe wheel is a simple function of center of gravity management; no springs, cylinders, or torque engagers are required. As I worked through the kinematics in my head, the big wrinkle was that if you were to execute a pure strafe, the off-center wheel would introduce a torque that would cause the "forward" wheels to rotate against each other. The bottom line here is that I would not want to run this style of drive train without a gyroscope or other sensor feedback so that the programming eliminated unintended torques.