rec.autos.simulators

Although far from perfect, I've updated Racer on my webpage.
http://www.racesimcentral.net/

Here are some of the new features:
- Added info on track timelines used in Racer.
- Timelines also really work, so you can now drive an actual lap time!
(best lap time is shown under the current lap time)
- Physics engine reworked to use quaternions (the car's body now has
gyroscopic precession as well).
- Gravity now always point down (not that it won't flip! ;-)
- views.ini file; it's now possible to create dials, timers etc. in a
very flexible way.
- Roll centers for the suspensions
- Actual camera's from the car.ini file are used (keys 0-9)
- Grid/pit positions (easily created from TrackEd too)

Known problems:
- I've just started using the obese InstallShield. In the quest for
speed, no icons will be created yet for the installed programs, so
you'll have to add them yourself (to the Start menu), or just run them
directly from the installation directory.
- Now & then the car starts out shooting for the sky. Pressing Shift-R
may help in that case (or restarting).
- There are funny things in the track; dips that make it harder to
drive. I'll have to test for the closest triangles soon to avoid the
'sinking' of the car.

The current future:
- Replays; I've got some code in, but it's mostly non-functional yet.
- Ghost car; based on the replays ofcourse.
- Mouse or keyboard control; to avoid the need for a joystick or
wheel. (but it will be crude)

As said, there are some problems, so I hope it works ok-ish on most
machines. Enjoy the preview! v0.4 will address some of the more
buggier points.

Ruud van Gaal

Ruud van Gaal, GPL Rank +53.25
Pencil art    : http://www.racesimcentral.net/
Car simulation: http://www.racesimcentral.net/

Shane

  I use InnoSetup for my install programs.  It's free, really easy to use, and
looks professional.
http://www.jordanr.cjb.net

  Thanks for the keyboard input :0)  I'll download and try it on my laptop.
Hopefully I'll find a way to plug my wheel into the thing ;-)

  Just curious, how many triangles are in the track you're using?  How many in
the car?

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

Ruud van Gaal, GPL Rank +53.25
Pencil art    : http://www.marketgraph.nl/gallery/
Car simulation: http://www.marketgraph.nl/gallery/racer/

Thanks! I'll check it out.

You're welcome. :) But the keyboard input isn't yet in this version
(although on the SGI I *only* use the mouse & keyboard, so it
shouldn't be that hard).

Good question. Hm, the car has got perhaps 300 triangles or so. The
track is far more, but is being culled away (details on my site in the
programmer's section) quite effectively. I have a Monza track lying
around, which looks a lot better (it's from F12K), but took up a lot
of texture memory. Haven't played with that since optimizing texture
use though.
I think there are about 100 triangles or so per track object, and
about 90 or so of those. So only 9000, and then some trucks,
grandstands and such. I really can't tell for sure. Might be a nice
statistical thing.
It runs at 40 to 60fps on my PII-400/TNT2Ultra (sadly at only 13fps on
the SGI).
BTW I can recommend the G. Genta book very much; a lot of direct
formulas and numbers. And 5 complete Pacejka samples for different
vehicles.

Ruud van Gaal, GPL Rank +53.25
Pencil art    : http://www.marketgraph.nl/gallery/
Car simulation: http://www.marketgraph.nl/gallery/racer/

rgds

Shane

Hi!

Same here. I edited the section in the "car.ini" that says:

  ; Location; give it a bit of height for the wheels*** on
  ; it will fall down to the ground by itself
  x=0
  y=0.35
  z=0

to

  ; Location; give it a bit of height for the wheels*** on
  ; it will fall down to the ground by itself
  x=0
  y=0
  z=0

but to no avail.
Something f***s up there ;)).

Regards
Crash

  Cool.  I downloaded Racer, but haven't tried it yet.  Got sidetracked working
last night, you know how it goes :-)

  That's a lot of triangles (to me, at least!)  I was curious how many ground
triangle intersection checks you needed to make each loop and how you were
handling that.  Without a really good search pattern, that many triangles would
slow my engine to a crawl.

  >BTW I can recommend the G. Genta book very much; a lot of direct

  Great, thanks!  I've gotta move in about three weeks, so it'll have to wait
yet.  I think I'm dumping Pacejka's methods for now and trying a brush based
model instead.  An experiment based on this type of model worked well, but I've
only handled the lateral direction so far.  It did a good job of predicting how
peak slip angle at max force changes with tire load, and the curve shape varied
pretty well with variations in static and dynamic friction settings.  Anyway,
that's where I'm at.  BTW, I figured out how to model several different
differentials, but I ain't tellin' yet!  Muuuhahaha!!  <Evil grin>  

  I'll still have to check out Genta's book.  Thanks for the tip!

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

Ruud van Gaal, GPL Rank +53.25
Pencil art    : http://www.marketgraph.nl/gallery/
Car simulation: http://www.marketgraph.nl/gallery/racer/

Note that I sometimes see the first frame with a car point the wrong
direction. The next frame, I see the car in the warped direction at
grid position 0, as it should be. Perhaps this first weird frame is
the cause (perhaps draw_every_step=1 will reveal the bug every time).
The suspensions are sensitive to wheel motion and might get kicked
because of this bad frame (at which perhaps the car is *IN* the
tarmac, producing a huge force).

Remind me to build a moon so you can try to target it. ;-) (and
ofcourse, drive on there!)

It's done from TrackEd; the first grid position. As it raytraces into
the track, you can't really set a height. Ehm, ofcourse you can! In
data/tracks/special.ini you'll find the grid positions, including Y
values. You *might* want to try heightening them, but a falling car
quickly goes into a bounce.
Also, try to do a Shift-R real fast to get a softer landing. However,
sometimes after some racing the suspension motion keeps going, even
when not driving.

This isn't used anymore; I'll have to throw it out. It was there to
keep the body from starting in the road. Now, the grid position is
used, and a *magic* value is used to lift the car somewhat to allow
for the wheels to just not touch the ground. This magic value is set
at 0.5 currently and not editable through an ini file. Because it
should be calculated by looking at the suspension rest length and the
radius of the wheel.

Ruud van Gaal, GPL Rank +53.25
Pencil art    : http://www.racesimcentral.net/
Car simulation: http://www.racesimcentral.net/

...

Yes, I'm at a deadline for Saturday for a new project too. It's hard
to get some REAL work done! ;-)

The track is divided into a number of geometrical objects. The DOF
files that are in data/tracks/oval1. For each of those, I keep a
bounding sphere. Then when looking for triangles, I walk the list of
all DOFs and see if the car's inside the bounding sphere. If so, I
walk every triangle and see if it hits.
So I test perhaps an average of 30 triangles PER wheel PER iteration.
So that's still a lot of work, and can use some optimization.
Especially if the track is not divided a lot, or the bounding spheres
do not represent the object very good, the miss-hits can become quite
large and press down on the fps.

Where did you get info on the brush model?

Sounds like it's going nicely. :)

Damn! Lol, I'm not into differentials yet, but well, I'll ask you one
day when you had too much to drink then. ;-)

Ruud van Gaal, GPL Rank +53.25
Pencil art    : http://www.marketgraph.nl/gallery/
Car simulation: http://www.marketgraph.nl/gallery/racer/

Dont forget to cache the last poly hit for ech wheel and make that the first
poly you check next time. Tell me to shut up if you already do that :O)

Chris

Thanks for the re-enlightenment though!

Ruud van Gaal, GPL Rank +53.25
Pencil art    : http://www.marketgraph.nl/gallery/
Car simulation: http://www.marketgraph.nl/gallery/racer/

  Oh, ok.  I'm not that far yet.  Right now, I limit my tracks to about 4-600
ground triangles.  Haven't implemented Chris' suggestion yet as far as caching
triangles.  You're right, this would dramatically increase speed.  I'll
eventually (by the year 3000 or so) check the first wheel's intersection, then
start the other three wheels intersection checks with that triangle, since they
usually hit the same one anyway.  Then, on the next loop, start with the same
triangle they hit last step.  This would boost speed probably 10-50+ times :-)
Hopefully my brain won't catch on fire this time....

  The idea came from looking at the diagrams shown all over the web and in
texts like Doug Milliken's awesome book, "Race Car Vehicle Dynamics."  Look on
page 23.  I took the contact patch and split it into 100 or so sections that
stretch from left to right across the tread, then let all the sections behave
like springs.  It works pretty well, but it's probably too intensive to do it
real time, although that would probably be the ultimate tire model :-)  To find
pneumatic trail and tire aligning torque, all I need to do is find the force
centroid and take the distance from there to the patch center.  (Haven't done
it yet)  I'll get to the longitudinal aspect before too long.  Combining them
shouldn't be too tough (yeah, right!)  

  Either way, I think I'll end up with a massive array of data that's pretty
good.  The cool thing is, by changing the lateral, longitudinal, and vertical
spring rates of the sections, the data could change with air pressure
variations and give realistic results.  

  I don't think I can predict load sensitivity, since I don't know what causes
it (I have an idea, but that would be really complicated to model) , but it can
easily be input directly into the code anyway.  I also believe that the amount
of sliding in each section could give basic tire heating info, but I'm not that
far yet :-)  I'm gonna stay home tonight and see if I can get the new driveline
model with an open differential running.  It works on paper so far, but....who
knows??  lol  You know how it goes :0)

  If this works, I think I could split each section (strips that go from the
left to right side of the contact patch) into several sections themselves (in
the lateral direction).  This way, camber angle would effect localized load
across the tread width and effect lateral and longitudinal force and tire
aligning torque.  

  What I will really need then is a way to reverse engineer Pacejka's formula
to get coefficients to match the "test" data, and run that through the sim the
same way you and almost everybody else probably do it.  I have no clue how tire
engineers derive the coefficients from real test data, but that's what I'd need
to do then.  Maybe an iterative approach could work.  This could be kind of a
"poor man's" numerical analysis tool for tires that could be designed by a
player.  Who knows?  It just might work :0)

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