Wrza wrote:
> Hi!
>
> This question is more about simulating physics than simulating exactly
> cars. We have lateral force acting on a car's wheels. Let's assume the
> car is moving sideway with the constant velocity so the fricton acting
> on the wheels is very high and causes the car to slow down. As much as
> I know this friction doesn't depend on car speed only on coefficient of
> friction (which depend on side slip) and load on wheel. My problem is
> that when the speed of the car is very low this friction force still
> acts on the car and finaly applies acceleration that moves my car in
> the opposite side it was moving before and then once again and the car
> starts to oscillate.
>
> The question is: how to simulate this lateral force on wheels on low
> sppeds? How to limit this force on wheels and not let it act like it
> was dynamic force (not friction).
>
> Thanks.
>
> Wrza

Low speed tire modelling is a tricky thing indeed. I had the same
problem in Virtual RC Racing's development (or should I say, "have," as
the problem is still there). It's a bit trickier to solve with the
little cars than bigger ones, but the basic problem is the same. The
car is barely moving sideways, but the slip angle is still +90 degrees,
so you get a kick back the opposite direction, now it's -90 degrees
slip angle, so you get another opposite kick. Back and forth, back and
forth... Same thing in slip ratio. Either way, the result is that the
car floats around instead of stopping completely and will slide down
the most gentle hill even with the wheels locked.

If you haven't heard of slip angle/slip ratio and aren't using them,
the problem is the same with a coloumb friction model in that the force
acts opposite the velocity vector, which itself can change signs every
step. Big force to the left, then one to the right on the next step,
back and forth, back and forth... Unfortunately people usually
interpret this to mean the "physics are wrong" and anything that
happens over 1m/s must be equally screwy :-P Not true of course, but
they don't know that so we're stuck spending a bunch of time trying to
make a realistic "sitting in one spot" simulation that is convincing.
Oh well :-)

If you're using a slip angle/slip ratio based model this can be fixed
up by incorporating a relaxation length into the model. A more simple
way is to make the slip angle/slip ratio that you feed into your tire
model a state variable that progresses slowly toward the "real" slip
angle/slip ratio depending on wheel rotation and movement. Another way
would be to switch to a spring type of model at very low speeds.

You might try googling here at RAS for "Car physics" + "tire" and so
on. There have been plenty of discussions on this topic and many other
vehicle dynamics modelling topics here over the years. Should be
plenty to peruse

Todd Wasson
www.PerformanceSimulations.com
www.VirtualRC.com
http://www.PerformanceSimulations.co...ToddSim17a.wmv