rec.autos.simulators

at higher speeds my car will always strongly understeer, no matter how
much i push the throttle. if i simulate a locking differential, or pull
the handbrake in curves i can make the car oversteer, however it will
be impossible to get it back to control. instead it will spin around
its vertical axis at least 360 degrees, more at higher speeds. i was
having the same problems with my old tire model, so i assume some other
part of the simulation is causing the problems.

thanks in advance for any suggestions,

jonas

What's the weight distribution, and how do the front/rear tyre
parameters compare?

Jonny

Many things can make that happen. In my case, I have a similar
problem; a loss of grip when the car finally lets go (oversteers). A
possible resolution might be to flatten the Pacejka curve (the height
of the grip once it passed the peak) so the tires don't lose so much
grip.

In my case another thing *may* happen; I use tire rates to get loads
(so if the wheel is 1cm 'in' the track, you get 0.01m*tirerate = ...
Newtons. With a small stepsize, the forces this generates may be a bit
too much for the poor Euler integrator and the tire will quickly
bounce up and down the track, making for at least half the grip you'd
normally have.

What was it again? I have tire rates upto 200,000 N/m (F1 tires seem
to be near that range), so this gives a natural frequency of ? That
multiplied by 10, given Matt Jessick's experiences, should be enough
for a Euler integrator. It's now at 250Hz most of the time in my case,
lol, probably not even close to what is needed at those tires.
Sticking in an inner loop that ups the frequency when doing the tires
may help in this case. I haven't done so yet, so I can't tell from
experience yet.

Ruud van Gaal
Free car sim  : http://www.marketgraph.nl/gallery/racer/
Pencil art    : http://www.marketgraph.nl/gallery/

then I'd use more than 30 to 100 times the *** frequency.
You can also cause major differences in the frequency response
for this simple integrator by exactly how you arrange the
two equations of a second order set.

E.g.:

vel += acc * dt
pos += vel * dt

may be noticibly more stable than the opposite order.

Analyzing the stbility of a digital filter like this
puts you right into digital control theory and
can best be done with digital design methods and
tools like MatLab, etc. (If you are a student, grab
student versions of this kind of software while you can.
It retails for $5000 or more... ;)

With a tire spring, you can estimate the "wheel hop"
mode frequency from the books. It will probably be pretty high.
Above 10 up to 20 Hz or more.

Make sure you do your stabiliy testing with sharp impacts
like steps (curbs, for example) because step inputs have
signal in the high frequency range that can interact
with your wheel hop, etc., frequencies.

--
Matthew V. Jessick         Motorsims

Vehicle Dynamics Engineer  (972)910-8866 Ext.125, Fax: (972)910-8216

  At ras here, some of us have talked about two (or more, depending on how you
break them down) methods for handling tire forces with slip ratio and slip
angle.  I don't use Pacejka yet, but instead use a very simplified curve, but I
think the results would still be similar.  Basically:

  1.  Using slip ratio, calculate longitudinal force.  Then, using slip angle,
calculate lateral force.  If the combined force is larger than the tire can
handle at its current load, trim the lateral force using the Pythagorean
theorem.  (c^2 = a^2 + b^2 for lurkers, basic high school stuff, nothing fancy)
 This means that the longitudinal force stays exactly as it's calculated from
the slip ratio and the lateral force *only* is modified.  This is how I do it,
and the car can be oversteered through high speed corners with opposite lock
and recovered with no problem (provided the yaw axis polar moment of inertia is
big enough, otherwise it happens too fast to control).  The longitudinal force
itself is only trimmed if it's too large by itself.  With this method, a
spinning tire can develop no lateral force, which I don't believe is as close
to reality as you might get with the next method.

  Brian Beckman's, Ruud vanGaal's, and Gregor Veble's methods are different,
however (I think, correct me if I'm wrong, guys)  They do something like this:

  2.  Using slip ratio and slip angle, find the combined force vector.  Then,
if it's outside the friction circle, trim *both* the lateral and longitudinal
forces to fit, rather than just t*** the lateral force by itself, leaving
the longitudinal force untouched.  

  Actually, I think they're t*** the slip vector itself first somehow
before doing this, so my description isn't exactly right.  I still haven't done
much with tire models, so am low-man on the totem pole here on that subject :-)
 Anyway, I have a feeling that this is a tire model thing (even though you've
switched methods), so if you can lay out some details on how you're handling
limit-force behavior, someone might be able to point something out.

  You've mentioned that with a locking differential you can get oversteer.  I
assume that otherwise you're using a fully open differential and the inside
wheel spins, preventing oversteer, right?  If this is the case, try lowering
the center of mass a little at a time and see if you get somewhat controllable
oversteer at some point (or widen the track width.)  This might not help too
much though, as you're saying that you're having trouble recovering the spin.
If you'll describe your limiting-force technique, maybe there's a clue there.  

  It's also entirely possible that the car setup itself is the culprit.  

Todd Wasson
---
Performance Simulations
Drag Racing and Top Speed Prediction
Software
http://www.racesimcentral.net/

Yes, hm, never figured out whether one would be better than the other.
I'll try swapping them, see what it does.

20Hz is ok; that'd mean about 2000Hz tops to integrate, that's
reasonably ok.

Thanks for the info,

Ruud van Gaal
Free car sim  : http://www.racesimcentral.net/
Pencil art    : http://www.racesimcentral.net/

'We' create a slip vector based on optimal slip; so you have a
slipvector that is like:
(currentSR/optimalSR,currentSA/optimalSA)

Optionally using tan(SA) instead of SA itself, but it doesn't do a
whole lot of difference I think. This vector is trimmed to unit size
(1) and then the long/lat forces are combined in ratio to the slip
vector; newLong=currentSR/optimalSR*pacejkaLong. Well, that's how I
think it is in the code, anyway, lol.

Oversteer and understeer are indeed THE parameters that change with
every little bit you change about your sim, be it code or setup.

Ruud van Gaal
Free car sim  : http://www.racesimcentral.net/
Pencil art    : http://www.racesimcentral.net/

yup.

i will try lowering the cm, however the most of the other car
parameters are taken directly from real life, so they can't be that
"wrong" ;-)  this is a '66 dodge charger, btw.

jonas

lowering the cm will make it easier to get oversteer, but still, the
general attitude of the car is to understeer unless a lot of power is
applied on the rear wheels, and once the rear breaks out, there's no
way to recover.

jonas

  I imported a similar dodge charger a few days ago myself :-)  Great car!  

  The yaw axis polar moment of inertia is probably the next thing you could
mess with, but then again, if you've got real data already for that of course
you'll leave it alone.

  I possible, you might try adjusting the Pacjeka values to flatten the curve
after the peak.  It seems that Ruud's sim (Racer) doesthe same thing your
describing, once it breaks loose, you're gone.  That was with version .46
though, haven't tried .47.  My "curves" are actually constant slope lines that
goes from 0 to max grip and then stay there.  It's fairly easy to drive through
corners with controlled oversteer, probably because the curve flattens this
way.

  If that doesn't work, I don't really know what else to say.  The only other
possibility I can think of is to check for vertical force instability in the
tires.  I run my suspension forces directly to the ground so it's very stable.
If you're tires are stiff springs, you might want to plot out the vertical
forces at one of them while kicking the car sideways, along with the suspension
force.  If the tire vertical force is much more jumpy than the suspension
force, that might cause the problem.  The fix for that might be in better
vertical dampening, or increased frequency (possibly greatly increased) for
vertical tire movements.

  I'm running out of ideas here, so hopefully this works :-)  My #1 bet now
would be the Pacejka curve.

Todd Wasson
---
Performance Simulations
Drag Racing and Top Speed Prediction
Software
http://PerformanceSimulations.Com

Try looking in slomo at the lateral forces of the rear wheels when it
breaks out. It's probably low. Then find out the big Why. :)

Ruud van Gaal
Free car sim  : http://www.marketgraph.nl/gallery/racer/
Pencil art    : http://www.marketgraph.nl/gallery/

-Gregor

yesterday i re-enabled some code to use a steering wheel in the sim. i
was using keyboard input before, i know it sucks but under Mac OS X
(the OS i'm using) at the moment i have some technical difficulties,
not letting me use my sound library and input library at the same time,
etc... - you know it's a new OS, many problems still have to be
resolved.

anyways: now that i'm using a steering wheel, the general attitude of
the car to understeer is gone. i suspect that my simulated keyboard
input was just turning the wheel to fast, resulting in understeer (this
is realistic behavior, isn't it?). the car oversteers like hell, like
one would except from an old dodge charger. thr problem of
uncontrollability remains, however. but i have a guess as to what's
causing it. my sim currently doesn't have a constant throttle input to
keep the engine running, so when i'm not on the throttle (which i still
can't control analogue, since i don't (yet) have a good, working
steering wheel/pedals setup), there's a lot of brake force coming from
the engine, which may very well be causing more oversteer when i'm off
the throttle.

i'll look into it, and thanks for the info you got me so far.

jonas

Another issue unrelated to the rear tires that can
make spin recovery difficult is exactly how much
front tire steer capability you give the user.

In video games, without the seat-of-the-pants
acceleration feedback, and because of graphical problems
and others we are probably late detecting the rear end
sliding out. This may mean you just need more front
steer capability. If you don't detect the oversteer
until 20 degrees, you may need 20 + peak slip angle
degrees of front wheel steer to bring it back.
If you only give the user say 20 degrees of steering
it may well be impossible to recover...

It may also be common to reduce the steering gain
at higher speeds to combat the yaw damping reduction
with velocity increase. This also has the possibility
of reducing the front wheel steer capability below
what is required to recover from an oversteer.

Another problem that should show up as sims get
more sophisticated is driver technique. Sometimes,
chopping the throttle is the worst thing you can
do in trying to recover a spin as it starts, because
it reduces the weight on the rear tires.
You have to trade off reducing the throttle induced
oversteer (if that was how you got the rear end
moving out) with reducing the rear weight transfer.

This kind of thing is particularly frustrating for me
to deal with on a daily basis, particularly early
in a sim's development before all the needed feedback
stuff is in place. There are people who will never
agree that driving fast is any different than driving
to the supermarket. And no amount of pointing at tire
marks spinning to the inside off the exit of corners
will convince them either. ;)   My favorite quote
along these lines is:
"We don't need driver adjustable anti-roll bars
because my Jeep Cherokee doesn't need them."

--
Matthew V. Jessick         Motorsims

Vehicle Dynamics Engineer  (972)910-8866 Ext.125, Fax: (972)910-8216

Keep this in mind while reading integrator comparisons.

--
Matthew V. Jessick         Motorsims

Vehicle Dynamics Engineer  (972)910-8866 Ext.125, Fax: (972)910-8216