Gerhard Wesp wrote: > On the contrary. I haven't followed this discussion too closely and > I'm no physicist either, but this sounds to me exactly like static vs. > gliding friction.
Yes, there are separate coefficients of friction for the static vs. dynamic case. But these are only different numbers. The problem is that the *model* (the algorithm used to compute results) must change in the static case. In the dynamic (skidding) case, it is enough to know that the coefficient of friction will be proportional (by the dynamic coefficient of friction) to the normal force into the ground, and be in a direction opposite the velocity. This works. The problem is that in the static situation, this model breaks. It's still physically correct, mind you, but it leads to numerical instability in the algorithm because of the very high velocity derivatives of the force function (i.e. the problem is "stiff" in the differential equation sense). Fixing *this* by interpolating the force function over small velocities leads to a stable but non-physical solution that exhibits the "drift" problem that was talked about. > Isn't that a bit of oversimplification? My car will skid in at least > two dimensions if I try hard enough :-) And a small amount of skidding > will always be present. It's not a simplification, it's a complication. A static spring in two dimensions requires storing a location and computing a distance. What do you do when the location has changed becaues the tire was rolling? You need to use the velocity and compute a 1D distance from where the tire "would" have been. Ick. Andy _______________________________________________ Flightgear-devel mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-devel