That's odd. Does it happen when the steering is perfectly straight or only
when it's turned? I have a similar problem where the longitudinal force under
max braking causes a yaw torque when the wheel is turned, even when the wheel
locks, because the longitudinal force is in the tire plane. Probably a good
argument there for Gregor's slip vector usage :-) Seems a "correct" Pacejka
model would take care of this automatically though, since the combined lateral
and longitudinal effects are included.
Maybe your rotation halt stopped working when you switched to quarternions.
>I had a thought the other day about the problem btw; what if:
I tried something like this a few months ago. I forget exactly, but I think
it worked pretty well actually, only it worked in reverse. I think the new
slip ratio was the new calculated slip ratio averaged with the last five or six
slip ratios, with the new slip ratio multiplied into the average a few times,
or something like that. Seemed to smoothen it out a bit, but don't remember
for sure. Actually, I think I dumped it once I split the frequencies.
Something like:
NewSlipRatio = ((OldSlipRatio1 * 10)+ OldSlipRatio2 + OldSlipRatio3 +
OldSlipRatio4) / 14
I know you hate the idea of splitting the frequencies, but your example above
is really what the split frequencies seem to accomplish by anticipating
excessively large slip ratios and preventing them from effecting the car's
movement. The car moves .003 seconds worth, then wheel rotation and new slip
ratio for something like .00001 seconds advancement is calculated, then again
and again, etc.. The slip ratio eventually overshoots where it should be, but
by far less than it would if it was using the same frequency as the rest of the
car, especially when there's a sudden, huge change in wheel torque. Besides,
if there are enough cycles and a little damping (I don't use any, but
calculating critical damping every .003 seconds as the tire load changes and
applying it would be a good idea, probably letting one cut down the frequency a
bit), slip ratio will swing back and forth, eventually settling at the
"correct" slip ratio. If the wheel torque isn't big, it'll slowly climb to
about the value the .003 timestep model would produce, so it doesn't mess
anything up. Anyway, then I move the car with this final slip ratio->torque
value. It really doesn't cost much in CPU power, but it is a pain to go
through the code and try to cut it up like that, as you can imagine. I'm glad
I did though :-) Even with a full size car, when I draw the longitudinal force
vectors in realtime 3-D at the tires, they are a little bit jumpy at only
300Hz. This effects the handling. Crank it up to 3000Hz or more and they're
so smooth you can't see a change.
Just a thought, I know it would hurt to do though :-)
I'm intrigued by your idea too, however. Will give it some thought and maybe
come back to you on it.
lol. This is the "Sim racing and money" thread!
Ok, I'll have to ask the salesperson. Thanks :-)
Todd Wasson
---
Performance Simulations
Drag Racing and Top Speed Prediction
Software
http://PerformanceSimulations.Com
