torcs-devel Mailing List for TORCS - The Open Racing Car Simulator

Bernhard Wymann wrote:
> Hi
> 
>> One of them is the porsche-gt1, which has a much higher max torque than
>> is warranted, though its power is more or less OK.. it is possible to
>> just fix that on CVS without breaking anything.
> 
> Please make sure that all robots and the human player setups are still
> reasonable afterwards, test/fix them all (seriously!).
> 

Yes - the only things that I am considering submitting on CVS these days
without a lot of testing are simuv3 things, which is 'testing' anyway.
Have you tried the modified susp code by the way?  I think now the simu
is almost as fast as simuv2 and has a number of fixes.  The
wheel/suspension code has been simplified to the point where it's more
or less just 'simuv2 in full 3d'.  I'm asking in order to decide whether
I should finish it up or just forget about it completely and concentrate
on the robots or help you with the networking.

There is one thing that does not agree with the simuv2, however, and
that's the drivetrain inertia which makes necessary to add more braking
power on the driven wheels. I am still maintaining my separate car-set
for that purpose (otherwise robots can't stop in the pits, for example),
so I could quite easily pass on the fix to simuv2 and update all the
vehicles in CVS.   Sorry, but I really like much more how cars drive
with that fix.  Oh, and I have also updated the limited slip
differential which was actually completely broken too. (Surprisingly the
viscous coupler seems to be OK).

So, to recap differences:

1) Rotational inertia of car and quaternions instead of angular
velocities and angles.

2) Calculation of forces relative to road plane, i.e. reaction force is
always perpendicular to the road plane.  This does not change anything
in flat tracks, but the difference is huge in banked curves - and is the
basic reason why simuv3 was made in the first place :)

3) Some bugfixes on the calculation of relative velocities between
wheels and ground

4) Rolling/pitching calculated properly according to position of forces
rather than using the fake 'roll center'

5) Bugfix in drivetrain inertia calculation

6) Now maintaining momentum when engaging gears so you can use the
clutch to spin wheels even with low-powered car

7) Bugfixed open differential, limited slip differential

8) Added a SimulationOptions class

9) Calculating the maximum theoretical downforce given an air resistance
(follows from conservation of energy) - useful when designing new cars,
maybe - found out that current cars have all their downforces around 4x
the normal, so added an option to change that (see 8)

10) Now taking into account the mass of the wheel, which makes for a
much more realistic (and smoother) ride on rough ground.


So the only patches here that actually add any CPU time are 1,2 and 10.
I'd like to keep all of them, but you are the Man.


Thrown away patches:

1) Calculation of exact point where wheel should touch the road (too slow)

2) Attempted to calculate forces properly when car not parallel with the
track (firstly, it is impossible to do it properly and secondly even a
half-arsed attempt is slow)

3) Collision - didn't work - this one doesn't either.  Must re-do.



>> The mclaren-f1 has a slightly different problem.  Its Aero section is
>> such that you can never reach its reported max speed of 380km/h +
> 
> Ah, a word about that. I do not care much about such data, because these cars
> exist(ed) in very different configurations. My goals are the following over the
> long term:
> - Anonymizing the car names/textures (no logos/names anymore, just "car7-trb1").
> - Make them fun to drive, so every car should feel a bit different and have its
> strength and weaknesses).
> 


OK

> I plan to put out a release soon, is it ok if I update the olethros textures or
> would you like to do it? The bot still needs to be ported to Windows, otherwise
> I will take the 1.2.4 version for the release.
> 

I can do it - when? I won't be around for 2-3 weeks.

> If I find time I will fix as well the ole-road start, the beginning/end mismatch
> looks bad.
> 

I'll put that in my TO-DO list.


> Bye, Bernhard.
> 

Bye


-- 
Christos Dimitrakakis
Homepage: http://www.idiap.ch/~dimitrak/main.html
Music: http://olethros.dmusic.com

Hi

> One of them is the porsche-gt1, which has a much higher max torque than
> is warranted, though its power is more or less OK.. it is possible to
> just fix that on CVS without breaking anything.

Please make sure that all robots and the human player setups are still
reasonable afterwards, test/fix them all (seriously!).

> The mclaren-f1 has a slightly different problem.  Its Aero section is
> such that you can never reach its reported max speed of 380km/h +

Ah, a word about that. I do not care much about such data, because these cars
exist(ed) in very different configurations. My goals are the following over the
long term:
- Anonymizing the car names/textures (no logos/names anymore, just "car7-trb1").
- Make them fun to drive, so every car should feel a bit different and have its
strength and weaknesses).

I plan to put out a release soon, is it ok if I update the olethros textures or
would you like to do it? The bot still needs to be ported to Windows, otherwise
I will take the 1.2.4 version for the release.

If I find time I will fix as well the ole-road start, the beginning/end mismatch
looks bad.

Bye, Bernhard.

-- 

Visit my homepage http://www.berniw.org
Official TORCS racing: The TORCS Racing Board, http://www.berniw.org/trb

I think there are a couple of cars which might need some updating, but I
am not updating data that can cause havoc on the CVS before I'm pretty
sure they won't cause trouble with stuff... so

One of them is the porsche-gt1, which has a much higher max torque than
is warranted, though its power is more or less OK.. it is possible to
just fix that on CVS without breaking anything.

So I think I'll do that if there are no objections.

The mclaren-f1 has a slightly different problem.  Its Aero section is
such that you can never reach its reported max speed of 380km/h +
firstly because there is too much air resistance and secondly because in
the Aero section all the balance is shifted to the front, making the car
oversteer a lot unless you add some rear wing.   My solution to fixing
the max speed without breaking all the other cars is to

a) Change its aero section such that when there are no wings it reaches
the max speed and is still handles quite well:

b) Add a small Front Wing, which has a default adjustment such that the
total force is approximately the same as that with the old Aero section.
This seems to work well enough not to make a difference with the robots,
since these don't set the "Front Wing" to anything in this car.

These are the actual changed sections:

  <section name="Aerodynamics">
    <attnum name="Cx" min="0.20" max="2.0" val="0.32"/>
    <attnum name="front area" unit="m2" min="1.0" max="3.0" val="1.8"/>
    <attnum name="front Clift" val="0.675"/>
    <attnum name="rear Clift" val="0.825"/>
  </section>

  <section name="Front Wing">
    <attnum name="area" unit="m2" val="0.5"/>
    <attnum name="angle" unit="deg" min="0" max="30" val="4"/>
    <attnum name="xpos" unit="m" min="1.0" max="2.5" val="2.3"/>
    <attnum name="zpos" unit="m" min="0.0" max="0.2" val="0.00"/>
  </section>



-- 
Christos Dimitrakakis
Homepage: http://www.idiap.ch/~dimitrak/main.html
Music: http://olethros.dmusic.com

Hello everyone,

My thesis is almost finished so I found some time to start ironing out
the bugs in simuv3.  Current update:

- A small bugfix in the differential.cpp.
- Changed default AERO factor to be 4.0 again.
- Changed wheel code so that it takes into account the mass of the wheel
when updating the suspension.  This should fix a strange behaviour that
                       occurs when a car enters a very rough patch of
ground.  If the suspension is long enough, then what you see is the
wheels being always stuck to the ground, while the car is flying high.
So now the wheels move according to their own mass.  This is a candidate
fix for going into simuv2.

Ciao,
Christos

-- 
Christos Dimitrakakis
Homepage: http://www.idiap.ch/~dimitrak/main.html
Music: http://olethros.dmusic.com