[ODE] Vehicle Gears/DriveDrain/Engine techniques.

[amundbørsand]

"Jeff Weber" <jeffreyweber@hotmail.com> skreiv:

> I have a general idea how i'm going to implement the gears/drive train, and 
> yes there will be an analog gas pedal involved.

I'm getting there as well.. soon. But I don't have access to any analog
input devices, unfortunately..

> From this paper http://home.planet.nl/~monstrous/tutcar.html I got the 
> following that I plan to use:

> maxEngineTorque = LookUpTorqueCurve(rpm)
> engineTorque = throttlePosition*maxEngineTorque

> driveTorque = engineTorque * gearRatio * differentialRatio * 
> transmissionEfficiency.

I guess your gear ratios are in the form 1/3 and so on? Because your
wheel torque gets smaller as the gearing gets higher, but you probably
know that already.

> I then plan to apply the drive torque by setting the hinge2 joint's MaxForce 
> = driveTorque.
 
> What I'm currently debating, is how "realistic" do I try to be.  According 
> to Matt Barnett's post, it may not be easy to find all the precise torque 
> data I need so right now i'm leaning toward "faking" the torque curves and 
> just tuning it till it feels right.  My simulation needs only to feel right 
> to the driver, it doesn't necassarily need to be based on precise vehicle 
> characteristics.

I guess the problem is knowing how much internal friction there is in a
drive train (differential, gearbox and stuff), but you could probably
find some numbers from a.. what is it called.. you know, when people
test their car's performance on those roller things, when you spin the
wheels on the rollers and they measure the torque and hp curves and
stuff. Well, if you could get your hands on that raw data (I know they
do various approximations to subtract internal friction and drivetrain
rotational inertia to calculate the engine power, but if you could just
get the torque curves at the wheels and use that for your calculations,
that would be totally realistic).

> Any other input is more than welcome.

This is a very interesting discussion (at least for me :), I'm in the
middle of this too. Or rather, in the middle of thinking about it.
Haven't gotten too far with the programming yet, but I'm getting there.. 

Something my car sim scream for is a better "tire model", ie. friction
circles/curves and stuff. The friction approximation ODE uses just don't
seem to cut it in my case, the car behaves very oddly when wheel
slippage is occuring. I don't really know what's wrong, but I'm guessing
it has something to do with the use of a friction pyramid instead of a
friction cone, could I be right? Also, rubber tires have a maximum
friction force when they are slipping at a sligt angle. There are some
formulas for this, is it Pajejka it is called or something?

I haven't looked into it yet, but how hard would it be to implement a
different friction approximation? That is, some kind of gradual blend
from static friction to sliding friction?

For great info on car simulation programming, check http://www.racer.nl/

-- 
amundbørsand <amund@c2i.net>