[ODE] Re: Simulating Wheels

[Serge Kourdakov]

Hi

everyone   who discuss wheels simulation

> > About the realistic car sims - if it's not an original idea, how come
> > nobody's done it?

>The fact that a single bad piece of data or an
> inappropriately simulated element can significantly screw up the resulting
> handling doesn't help either.

actually as Doug Milliken ( author of several  vehicle dynamics books) many
times pointed at rec.autos.simulators newsgroup
that
no one knows what is the 'real' physics of a car. all the models available
are just some crude appoximations... so even being implemented all the best
assumptions in the end could come to be looking as a pure  'fake' ....

the same concerns Pacejka magic formulaes. While they are very much  OK
still this is not the most correct description of tires.
sometimes lacking important details.
just search discussions on Pacejka magick formula  in mentioned newsgroup
then take Pacejka book on tyre modelling itself... etc ...and the things
wills tart to look less easy to implement.

more .. for ex on Pacejka  any availbale commercial  soft to get Pacejka
coeefs costs above USD 15 000 - it is not difficult to implement to anyone
the algo for fit data from tires to Pacejka coeffs...still this and many
other tasks require time...  and the more such tasks it is more difficult to
end...

so my point  such subtle points goes more and more while you move along the
development.

BTW anyone who is intrested in auto sims could visit the page
http://www.inrets.fr/ur/sara/Pg_simus_e.html there are some vehicle sims
which costs even as much as USD 50 millions ....
http://www-nrd.nhtsa.dot.gov/departments/nrd-12/NationalAdvancedDriverSimula
tor.html

also personally  I already spend two years developing driving sim code (
though have to manage a group of developers in fields of graphics etc.. so I
spent on developing vehicle dynamics only fraction of my time .. but anyhow
it took me two years to come to some reliable results  - actually it
included as a first step a clean design of classes and init system.. as it
comes a real mess for any end user if there is no clean sturcture for the
code) ... yes my efforts not ended with nothing ...the code now is  used in
several research projects.. still I cannot call my code to be perfect... it
is not based on ODE  for now  ( I use computation of motion and rotation in
body frame  and also use penalty methods for ground forces... so this is a
sort differnt from ODE approach... )( though I had and have a strong idea to
have at least human - driven (not automatically guided  'robot cars')  cars
in my project to be ported to use physics from ODE

but
actually implementation of physics includes not only tires and suspensions
( though they are important) this requres implementing differntials,
clutches  or torque converters, modelling engine ( which could be modelled
at different level of details..) brakes ( and abs for them)  traction
control

so just seems that developing good vehicle physics based on ode will succed
but this will require quite a bit of  efforts ( and knowledge) .

Regards
Serge