[Jared Russell]

I believe the intention is that the center wheel on each side (where the chains meet) uses a ball differential. Thus, the mecanum wheels could still be independently powered and the center wheel gets the "average" speed. Likewise, the second and fourth omni wheels appear to also be linked between the left and right side with a ball differential in each wheel. These wheels would receive the "average" of the two front and two rear motors' power, respectively. When strafing none of the omni wheels would turn.

So while the kinematics work, I too have to question the benefits of this arrangement (other than being a novel use of ball differentials). More wheels != more traction. In fact, in this case, I believe that the reverse is true.

In the forward direction, the mecanum wheels will be spinning at the same rate as the omni wheels despite the fact that their rollers are offset by 45 degrees. Thus only 71% of the torque applied to these wheels is applied usefully (assuming free spinning rollers; the loss of torque, of course, is an issue with all mecanum/kiwi designs). In the lateral direction, none of the omni wheels are spinning. This means that all of the weight over those wheels (and in the center of the bot, I'd have to imagine that that is most of the weight) is not aiding in traction. Worse, this means that a bot with this drive system will be especially susceptible to pushing from the sides, as only four of the ten wheels can offer any resistance whatsoever.

joeweber - please correct me if I have misstated your design intent or made false assumptions about what you were trying to show with this drawing.