[BenB]

If the wheel was mounted so the axis of rotation was fixed, it would provide lateral traction. Even if the wheel was mounted so it could swivel, it would provide an instantaneous amount of traction while the wheel swiveled to its new position. The only way I can see this feasible is if you mounted a sphere in the middle, like the ball for an old mouse. Even then you would have to prove to FIRST that it provided no additional traction. Unfortunately, I don't think they would accept this.

You could mount a wheel that was not powered. As long as you have good bearings that provide negligible friction in the axle, it won't slip. An encoder could be attached to the axle. The issue with this is that you loose precious traction with each wheel that touches the ground that is not powered. If your willing to go this route however, you could mount a wheel in a perpendicular direction to receive encoder values for both the x and y axis. The wheels have such little traction, a wheel in the perpendicular direction won't resist motion provided by the other wheels.