[wireties]

This is a good example of a real-world engineering decision. Does the added complexity of a co-processor (2nd development environment, robot<->co-processor protocol, lower system MTBF, extra cables, power considerations etc) outweigh the benefits of the extra processing horsepower? Remember to add in the amount of time it takes to send messages back and forth to the co-processor, you can never miss the reply back to the OI controller! And I think you must process the I/O from the robot controller. So this is not a simple decision, engineers think carefully before doing such a thing in a real-world application.

Personally, I think that only a math-intensive requirement (like the vision processing mentioned in this thread) warrants the addition of a co-processor in these games. Otherwise it is rarely worth it. Implement your math routines using fixed-point or integers and there is considerable processing power on the robot controller.