rec.autos.simulators

Hi

I implement simulation ( used in game like environment but for other than
game experiance purposes - mainly for research studies in perception ( but
driving is essential) where I use combined lateral and longitudal forces
using approach described at Tyre and Vehicle Dynamics by Hans Pacejka
Chapter 4

I have implemented non - related force computations and this worked fine ..
but moving one step forward....
 I got to the point on page 189

where the formulas for term Eyk reads

Eyk=rEy1+rEy2*dfz                  (4.E64)

and for term Exa reads

Exa=rEx1+rEx2*dfz                 (4.E56)

but  though other coeffs for all formulas used befire are provided as sample
data on page 614 in appendix 3  the values for these coeffs
rEy1,rEy2,rEx1,rEx2 are not provided :(

as seems it is not matter of great importance... but
I wonder is it possible to use coeffs pEy1, pEey2 (4.E24) pEx1,pEx2 (4.E14)
page187 ( or maybe  though the coeffs are somewhat realted they have
essentially  different  in their numerical values?)

Maybe   there is a place to ask for book errata. as seems these coeffs are
ommitied unintentinally?

there are other minor issues ( unused expressions etc..) so maybe
 Prof. Pacejka has a support page where additions to his work could be
found?

or for ex there are number of works on the net using his previous
formulas... are their newer publications whihc migh shed light on these
coeffs?

it is just really good to try his apporoach for simulationg dynamics....
but this is a sort of
- there is almost all to move with it.. except some minor ommitions :)

Regards
Serge

You might want to try Ruud Van Gaal's computation of Pacejka to see how
Racer (www.racer.nl) does it, the source code is freely available there, and
it uses Pacejka, I believe he has written a Pacejka class. I'm afraid
although I have used G. Genta's book on Motor Vehicle Dynamics which covers
some of Pacejka, I am not familiar with Pacejka's books, or the coefficients
offhand.

"Alex Smith"

Thanks Ruud really made a great work of  popularization of Pacejka
approach.
There are several Pacejka formulas each is an improvement of previous.
Ruud uses  older approach that described in new book.
for my purposes older Pacejka formulas are not sufficient (  or gives
some head ache.. while newer approach solves these my probs quite
elegantly)

so thanks for your hint I would rather  think on coeefs ( as they have
meaning ( they affect on some properties of friction circle ( as all
the story about  combined forces))and it is possible to draw friction
circles and see how coeffs change the shape of the circle  ) so
eventually just a bit of my efforts :)

Thanks once again

Regards
Serge

Hehe, nice to hear, although I don't know if I can be credited for
that. :)

What problems do the new Pacejka formulae solve? Combined Fx/Fy
forces?

The problem is always the coefficients. I might move to a later MF
version (MF=Magic Formula), but only when I know how to translate the
older versions into the newer.
I have done combined Fx/Fy a la Gregor Veble/Brian Beckman, which is a
friction circle approach.

There are professional tools to create Pacejka coefficients from
experimental data, but then you get very deep in all this stuff (most
is proprietary). I'm not so sure I even *want* extra coefficients. ;-)

I haven't read the Pacejka book yet, so I don't know any nice example
values. Mostly starting out with id values (so that nothing really
happens) may be the best way to avoid getting too much into the dark.
Ofcourse, then you don't really use the formula to its full potential,
but on the other hand, for a convincing MF you don't really need it
all. A normalized approach (Ch. 14 in RCVD I think, just dividing Fx,
Fy by Fz and using that as a base curve) does nicely as well, and a
modification of the forces by rational reasoning works ok quickly.

With normalization it is only the general notion of 'less tire
pressure gives grip increase/decrease in such a curve' is important.
And it gives you something to tweak bit by bit, and improve by itself.
I think the MF is actually a bit too complex (the tan etc); a base
curve would do, with modifiers that all add up, but in a more direct &
independent way.

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

first  for me it is important that now friction coeffs are easier to
add ...
so they have meaning in the new MF ( not like the older approach some
coeffs  which are  looking- like -friction coeff :) )
also it is possible to try more different properties of tire such as
effect of turn slip etc .. also seems that camber is now better suits
for real time  sims ( as now it is included in form of sin .... though
not sure ( as currently I omit camber) but seems there suold be less
probs....)

maybe the problem was my coding :)  ( the way I coded approach ) but I
did not like the result....
( just personal :) )

also see some benefits of using newer MF for me this is more important
argument - to have "all in one" rather than have something which is
fixed here and there to implement the same.....

 I have seen papers on genetical algorithms  to find fit curves for
data.
actually  for my eyes not too difficult to implement....
ex
http://archimedes.ing.unibs.it/papers/documenti/ISATA-genet-02.pdf

it also states the approach is OK for newer MF....( also some hints
how to get coeffs from experimental data is included at the end of
Pacejka book)

some day when I'm OK with free time I will code a genetic algo
approach.. but so far I use example data .... currently it is OK for
me... and as new Pacejak is scalable  ( using scale coeffs) I just do
that if I need some different properties of tire....

Regards
Serge