[Rickertsen2]

The idea solution would be to record distance in 2 dimensions as a series of waypoints. This is complicated by the fact that the robot does not travel in perfectly straight lines at 90 degree angles.

You could record time vs speed and then use closed loop feedback to stick as close as possible to that speed. Feedback seems like it should work better than the normal way but like you say something time independant is far better.

Perhaps distance vs. speed? IE: when we had traveled 815 ticks we were going 65 WheelTicks/Sec. Distance would be defined as how many ecoder ticks since the start. This still has a problem of the two sides being out of sync. They may reach their relative distances at different times cusing improper turning.

Somehow special attention needs to be payed to the phase of the left side and right side.