rec.autos.simulators

The problem is that if you lose the back end (even a little) it is
almost impossible to correct because you cannot turn the steering
which the required amount of opposite lock.

Does this mean that in a real car the steering wheel rotation and the
angle of the wheels are not linearly proportional? Or do I have some
other problem in my simulator like maybe the body inertia is too low?

Colin

Just kidding, the problem is, in a real car you have a possibility to
turn the wheel more than you usually can do with the regular ones. Or,
in a real racing car, the steering may be more twitchy, but that is
offset by better feedback through both the wheel and your seat so you
tend to drive smoother (that's why I believe a proper karting simulation
would be very, very tough to drive). You can also turn the speed
sensitivity in GPL by a switch in the .ini file. The speed sensitivity
is the first thing I look for in any racing sim, and it's also the first
thing I turn off :).

I noticed that FF helps immensely in this situation as well. In a non-FF
wheel, or a poor FF implementation, the resistance is given just by the
position of the wheel from the center, and you may easily overcontrol
the car. A FF wheel, on the other hand, tells you when any side of the
car weighs up, and you tend to let this info drive the car as well.

-Gregor

Yes, and it's possible in GPL's core.ini I think. But to answer
Colin's question; steering doesn't have to be linear, and probably
often isn't. So put in a non-linear function, like x^3 and interpolate
between that and linear (depending on how linear/non-linear you want
it). That really makes things a lot more controllable, as I learned
from Racer v0.4.6 to v0.4.7 somewhere.

Ah yes, I should read the entire post before responding! ;-)

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

But, as said by the other posters, what you are experiencing in GPL is
due to the speed sensitive steering feature, which btw has nothing to do
with the steering linearity.  This feature can be turned off by putting
these two line in core.ini:

[ Hack ]
steer_ratio = 0

It would seem as I had guessed that the inertia of the car body was
too low and hence made the car feel very twitchy. I tried plugging in
a few numbers from some of the cars from Ruud's sim and the ride it
much smoother. I was originally calculating the inertia tensor from
the dimens of the body and the mass :-

invBIT(0,0) = 1.0f / ((Mass / 3.0f) * ((dims.y*dims.y) +
(dims.z*dims.z)));
invBIT(1,1) = 1.0f / ((Mass / 3.0f) * ((dims.x*dims.x) +
(dims.z*dims.z)));
invBIT(2,2) = 1.0f / ((Mass / 3.0f) * ((dims.x*dims.x) +
(dims.y*dims.y)));

But if I use values like x = 1 / 2460, y = 1 / 3900, z = 1 / 1180 then
things are a whole lot more sensible :-)

You boys are definately right about FF though, I'm going to have a go
at that next week and see if I can't come up with a good solution.
I'll have to implement Mz first though to get some alignment torque
going.

Colin

Not a real answer to your problem, just a note that you are in good
company:

There is a story in one of the books about Lotus--Colin Chapman designed
one of his early cars with just enough steering lock for the tightest
corner the car would encounter...  then one of his drivers didn't have
enough steer angle available to catch a spin.  Chapman learned that this
wasn't an area to scrimp on, and subsequent Lotus cars had lots of lock
(forget the amount, but perhaps 45 degrees at the front wheels?)  And, fast
enough steering ratio so that the driver didn't have to move their hands to
make the correction.

A trivial question, is the angle of steering measured as the amount a
wheel can turn in either direction from straight OR the angle from
full left to full right lock?

Colin

Well, I should have said "...45 deg each way from center (90 deg total)".
I don't think there is a standard way to report it.  Good point!