[EricH]

Originally Posted by TheKeeg

Also, think of the grooves in your tread as teeth and the carpet as another set of teeth. As your wheels try to spin the teeth on the wheels will dig in and push against the teeth in the carpet which will move the robot. So the more teeth you have pushing against each other, the more "traction" you have. Lets say there is only one set of teeth, i.e., one groove in the tread and one in the carpet. Now, since the tread is a compressible material, it will not take much load to deform the tread and cause it to slip. If there were 100 pairs of teeth it would take a lot more to deform all of the teeth and slip the wheel.

This is actually pretty close. To summarize previous discussions:

There is a small-scale interaction between the tread of the wheel and the carpet, think of it as "mini-Velcro". Nominally, the wheel has a certain coefficient of friction, thus a certain maximum force that can be exerted either by the wheel or on the wheel (which we generally think of as "traction", F=mu*N). This small interaction will slightly--oh-so-slightly--increase the apparent coefficient of friction, and thus increase the traction very slightly.

Here's where it gets tricky. As you add more wheels, N goes down per wheel, something about more points for N to act on. But... you're also adding more interaction points, which will slightly increase mu overall... There is a limit, I think, to how useful it is to add width and area, but I'm not sure where it is.


Someone managed to kill the discussion entirely by noting that in 2009, where traction was limited, their team got better traction by doubling up some or all of their wheels. Anecdotal, to be sure, but still worth considering.