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

Eric Espie wrote:

> Christos Dimitrakakis wrote:
>
>> Since I was working a bit on the skids, I also put my hand a bit on
>> the smoke creation. Now the smoke is launched with some speed from the
>> tyres depending on the relative speed of the patch to the road
>> surface.
>>
>> Unfortunately all surfaces are treated equally, while obviously the
>> cloud initial speed and its damping and size should depend on the
>> surface the car is on. A quick hack would have me do something
>> similar to what I do in grskidmarks and grsound, where I look for a
>> substring like 'sand' 'gravel' etc. However, especially for the
>> skidmarks, things like cloud/skid colour and 'looseness' of the track
>> would help to get more consistent graphical effects.
>

The adaptation of the smoke/skidmark in function of the surface, was 
never done
because we didn't add the necessary information in the surface definitions.
According to Eric, I think that the best way is to add the necessary 
informations
inside the surface definition, and to make smoke/skidmark modulation 
depending
on the surface the wheel is on.
If I remember well, currently, there is only the information about the 
segment the car
is on, but there is no individual surface information for each wheel. (I 
think it was the
main reason the skids/smoke was not improved.)


>>
>> I started adding some different types of damage, like your wheels
>> becoming misaligned or your axle becoming bent. Currently damage is
>> managed by a single variable. I propose to keep the 'damage' variable
>> as a summary of how damaged the car is and to split the other damages
>> into manageable bits for the simulation. Maybe, as Bernhard suggested
>> those other types of damages should be limited to the 'pro' drivers.
>>

It could be possible to made damages more realistics in applying damage
forces on the body of the car. But actually, the damage forces are only
known in the simulation part and not transmitted to ssgraph. The 
deformations
could be a bit precomputed using ACCC to generate a deformation skeleton
of the car, in creating  ponderated relationships between the points of 
the 3D geometry.
For example, to define which points must be moved when there is a front 
damage, and
how much depending of the force.
A lot of work .

>
> I was thinking about reworking on the car definitions using
> materials for the different parts.
> For example say that a wheel is steel made and compute
> the weight and inertia from its dimensions.
> Also tire dimensions could be used to compute the stiffness
> and adherence...
>
> For the tracks, the surface definitions can contain the
> associated textures for skid and smokes.
>
> For the dammages, It would add a lot of interest if they were
> localized.
>
> may be pieces can be defined by their angle of exposition,
> for example the engine could be dammaged if a collision
> occurs within the 30 deg from the rear...
>
> May be a custom driver category could be defined
> if people want race rules and not heavy collision damages,
> or low traction...
>
> Bye,
> Eric.


cgu

cgu

Christos Dimitrakakis wrote:
> Since I was working a bit on the skids, I also put my hand a bit on
> the smoke creation. Now the smoke is launched with some speed from the
> tyres depending on the relative speed of the patch to the road
> surface.
> 
> Unfortunately all surfaces are treated equally, while obviously the
> cloud initial speed and its damping and size should depend on the
> surface the car is on. A quick hack would have me do something
> similar to what I do in grskidmarks and grsound, where I look for a
> substring like 'sand' 'gravel' etc. However, especially for the
> skidmarks, things like cloud/skid colour and 'looseness' of the track
> would help to get more consistent graphical effects.
> 
> I started adding some different types of damage, like your wheels
> becoming misaligned or your axle becoming bent. Currently damage is
> managed by a single variable. I propose to keep the 'damage' variable
> as a summary of how damaged the car is and to split the other damages
> into manageable bits for the simulation. Maybe, as Bernhard suggested
> those other types of damages should be limited to the 'pro' drivers.
> 

I was thinking about reworking on the car definitions using
materials for the different parts.
For example say that a wheel is steel made and compute
the weight and inertia from its dimensions.
Also tire dimensions could be used to compute the stiffness
and adherence...

For the tracks, the surface definitions can contain the
associated textures for skid and smokes.

For the dammages, It would add a lot of interest if they were
localized.

may be pieces can be defined by their angle of exposition,
for example the engine could be dammaged if a collision
occurs within the 30 deg from the rear...

May be a custom driver category could be defined
if people want race rules and not heavy collision damages,
or low traction...

Bye,
Eric.
-- 
=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=
                        TORCS
             The Open Racing Car Simulator
         AKA The Other Release Coming Soon (Skin'r)
                  http://torcs.org
=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=

Since I was working a bit on the skids, I also put my hand a bit on
the smoke creation. Now the smoke is launched with some speed from the
tyres depending on the relative speed of the patch to the road
surface.

Unfortunately all surfaces are treated equally, while obviously the
cloud initial speed and its damping and size should depend on the
surface the car is on. A quick hack would have me do something
similar to what I do in grskidmarks and grsound, where I look for a
substring like 'sand' 'gravel' etc. However, especially for the
skidmarks, things like cloud/skid colour and 'looseness' of the track
would help to get more consistent graphical effects.

I started adding some different types of damage, like your wheels
becoming misaligned or your axle becoming bent. Currently damage is
managed by a single variable. I propose to keep the 'damage' variable
as a summary of how damaged the car is and to split the other damages
into manageable bits for the simulation. Maybe, as Bernhard suggested
those other types of damages should be limited to the 'pro' drivers.

-- 
Christos Dimitrakakis
IDIAP (http://www.idiap.ch/~dimitrak/main.html)

Christophe Guionneau wrote:
> Christos Dimitrakakis wrote:
> 
>>>here is a sample of accc command line with a base track and shadows.
>>>
>>>accc.exe -g alpine-1.acc -l0  alpine-t.ac -l1 alpine-1-shad.ac -d3 1000
>>>-d2 500 -d1 300 -S 300 -es
>>>
>>>    
>>>
>>
>>What are the other two options, -l2 and -l3?
>>
> -l2 and l3 are for the other textures , for example in g-track3 there is 
> a permanent skidmark
> on the whole track (-l2)
> and -l3 for another layer , but the acc loader (and the grvtxtable 
> class) in torcs is not ready for 4 layers,
> -l3 is for futur uses 
> 
>>
>>Also, I have tried generating the shading for the whole track (i.e.
>>both track and terrain) in the following manner:
>>
>>1. I use trackgen -a to create track + terrain .ac file and the
>>elevation map.
>>2. I generate shading from the elevation map.
>>3. I generate the shade .ac file, (which covers both track and
>>terrain)
>>4. I combine with the above command, but the result is that only the
>>track's shading is visible. Any idea why?

Did you update to the latest trackgen, because older versions
had this known bug.

>>
> I don't know but may be here is some clues:
> Only object in the layer 0 are kept in the result, that means that the 
> -l0 file must contains all the objects,
> and the objects MUST have the same name in all the layers..
> the best way is to use your <track>.ac to generate <shad_track>.ac :
>     accc +shad <track>.ac <track_shad>.ac
> accc will generate a new ac file with a basic vertical mapping on all 
> object.
> and then use accc -g to combine all the layers. (don't forget to create 
> the shadow2 texture).
> 
> cgu
> 

Eric.
-- 
=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=
                        TORCS
             The Open Racing Car Simulator
         AKA The Other Release Coming Soon (Skin'r)
                  http://torcs.org
=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=

Christos Dimitrakakis wrote:

>>here is a sample of accc command line with a base track and shadows.
>>
>>accc.exe -g alpine-1.acc -l0  alpine-t.ac -l1 alpine-1-shad.ac -d3 1000
>>-d2 500 -d1 300 -S 300 -es
>>
>>    
>>
>
>What are the other two options, -l2 and -l3?
>
-l2 and l3 are for the other textures , for example in g-track3 there is 
a permanent skidmark
on the whole track (-l2)
and -l3 for another layer , but the acc loader (and the grvtxtable 
class) in torcs is not ready for 4 layers,
-l3 is for futur uses 

>
>Also, I have tried generating the shading for the whole track (i.e.
>both track and terrain) in the following manner:
>
>1. I use trackgen -a to create track + terrain .ac file and the
>elevation map.
>2. I generate shading from the elevation map.
>3. I generate the shade .ac file, (which covers both track and
>terrain)
>4. I combine with the above command, but the result is that only the
>track's shading is visible. Any idea why?
>
I don't know but may be here is some clues:
Only object in the layer 0 are kept in the result, that means that the 
-l0 file must contains all the objects,
and the objects MUST have the same name in all the layers..
the best way is to use your <track>.ac to generate <shad_track>.ac :
    accc +shad <track>.ac <track_shad>.ac
accc will generate a new ac file with a basic vertical mapping on all 
object.
and then use accc -g to combine all the layers. (don't forget to create 
the shadow2 texture).

cgu


>
>  
>

Christos Dimitrakakis wrote:

>To complete my track how-to I need to know:
>
>about the Surfaces:
>
>What is "color R/G/B 1/2"?
>What does the "texture mipmap" size do? (generates a mipmap that is a
>particular size??)
>
>about accc:
>
>When using the -g option, what are all the various components exactly?
>So far I only used the base track and the base mapped (for the
>shading).
>
>  
>
here is a sample of accc command line with a base track and shadows.

accc.exe -g alpine-1.acc -l0  alpine-t.ac -l1 alpine-1-shad.ac -d3 1000 
-d2 500 -d1 300 -S 300 -es

d3 , d2,d1 are used when the objects are grouped:
    -the track is splitted in segment, based on the TKMN segment 
(generated by trackgen)
    -all the 3D objects are attached to a TKMN group depending on its 
distance .
    -inside a TKMN group, 3 branches (sub-groups) are created, using the 
distances d1,d2,d3 and the distance
       of each object.
This 3D structure, may be used some days to improved rendering 
performances. I have experimented
some improvement, by eliminated some sub-groups depending on the 
distance of  view, but I was not
really satisfied. Part of the code is under  #ifdef GUIONS in grscene.cpp

-S  is used to split big object in smaller one

-es is for "extended Strip" , and it is really mandatory !!!!!

hope its help.
cgu

To complete my track how-to I need to know:

about the Surfaces:

What is "color R/G/B 1/2"?
What does the "texture mipmap" size do? (generates a mipmap that is a
particular size??)

about accc:

When using the -g option, what are all the various components exactly?
So far I only used the base track and the base mapped (for the
shading).


-- 
Christos Dimitrakakis
IDIAP (http://www.idiap.ch/~dimitrak/main.html)

> >
> > I do this, basically:
> > skidState->setTexture ("data/textures/grey-tracks.rgb", FALSE, TRUE,
> > TRUE);
> >
> > which should do the same, through plib, I hope.
>
> setTexture(filename, wrap, wrap, mipmap);
>
> So what about trying TRUE, TRUE, TRUE? Perhaps you have the texture
> coordinates not laid out the way you expect? Do the texture coordinates
> advance on the skids (I'm too lazy at the moment to look into the code)?
>

Aie, I hadn't realised there were two wraps :P.

Now the problem is.. how to make the skids more realistic on
terrains other than asphalt.

1. Different textures.
2. Keeping the same texture but changing it's colour depending on the
terrain.
3. Using the texture as an overlay for the terrain, i.e. like a
bump-map a light-dark map.. but I don't think that this is actually
possible.

For mud/dirt terrains, when skidding you should create 'wider'
skids... in any case I have had problems in getting the texture to
integrate nicely with the terrain texture. Firstly, the colours must
match somewhat, so I will be needing some information about the
terrain texture. Secondly, the edges of the texture are a bit sharp,
so I will probably need to soften them for some terrains.


Lastly, the direction of the car's drift is important in the formation
of both skids and dust particles.
-- 
Christos Dimitrakakis
IDIAP (http://www.idiap.ch/~dimitrak/main.html)

Hi Christos

> So... did someone check out the new skids code? When the texcoords go
> over 1.0 it always displays the same texture line.. :/

Did you commit to the main branch? I recognised no changes at the 
checkout... Did you configure the texture with GL_REPEAT (I don't know 
which coordinate you use along the skids, I assume "s")?

glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_REPEAT);

bye, Bernhard.

-- 

visit my homepage http://www.berniw.org

So... did someone check out the new skids code? When the texcoords go
over 1.0 it always displays the same texture line.. :/


-- 
Christos Dimitrakakis
IDIAP (http://www.idiap.ch/~dimitrak/main.html)

Hi all,

I have done a tutorial to import the NFS (3 and 4) cars into TORCS.

<http://torcs.sourceforge.net/sections.php?op=viewarticle&artid=31>

Feedback is welcome,
Eric.
-- 
=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=
                        TORCS
             The Open Racing Car Simulator
         AKA The Other Release Coming Soon (Skin'r)
                  http://torcs.org
=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=

OK. I have updated simuv3, though the collision code is still not
working very well.

I have also updated skidmarks, plus a sample skidmark texture in
data/textures.

What I would like to do for skidmarks would be to have a
darken-lighten texture rather than a colour texture and to specify
the colour using the colour array if necessary. So, I would like to
set colour to 0,0,0,skid and the darken/lighten effect would be set to
transparent for skids on the track. For mud/dirt skids, the colour map
would be transparent and the darken/lighten effect would be
transparent inversely proportional to the amount of skidding..

How would that be
possible? Can I have two textures, one colour and one lighten/darken?
Or do we need to write a new engine bit for darken/lighten?

-- 
Christos Dimitrakakis
IDIAP (http://www.idiap.ch/~dimitrak/main.html)

Eric Espie wrote:
> Note that the "module" and "idx" columns are for internal use,
> the race engine uses this to continue a race set with the same
> robots.

and the "Current" section is also for internal use...

Bye,
Eric.
-- 
=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=
                        TORCS
             The Open Racing Car Simulator
         AKA The Other Release Coming Soon (Skin'r)
                  http://torcs.org
=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=

Eugen Treise wrote:
> Hi all!
> 
> I have updated the stylesheet.
>  - Column "time" shows now the difference to the 1st placed driver in 
> seconds or in laps.
>  - Time values are better formated.
>  - Some small changes
> 
> You can download style.zip from http://eugentreise.de.tt/torcs.html

Thanks, I have commited it in CVS.

> 
> Things which could be changed in the results-*.xml files:
>  - "pits stops" -> "pit stops" ?

done ;-)

>  - Drivers with 0 points are reverse ordered in the Standings section 
> (see sample result files on the above page, i.e. Quick Race: i arrived 
> 2nd, in Standings i am 19th)

good catch, I have fixed that too, it's in CVS now.

> 
> 
> bye, Eugen
> 
> PS: I had to clear the browser cache to see the new formatings.
> 
> 

Note that the "module" and "idx" columns are for internal use,
the race engine uses this to continue a race set with the same
robots.

Bye,
Eric.
-- 
=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=
                        TORCS
             The Open Racing Car Simulator
         AKA The Other Release Coming Soon (Skin'r)
                  http://torcs.org
=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=

Hi all!

I have updated the stylesheet.
  - Column "time" shows now the difference to the 1st placed driver in 
seconds or in laps.
  - Time values are better formated.
  - Some small changes

You can download style.zip from http://eugentreise.de.tt/torcs.html

Things which could be changed in the results-*.xml files:
  - "pits stops" -> "pit stops" ?
  - Drivers with 0 points are reverse ordered in the Standings section 
(see sample result files on the above page, i.e. Quick Race: i arrived 
2nd, in Standings i am 19th)


bye, Eugen

PS: I had to clear the browser cache to see the new formatings.

> >
> > Really? actually it works this way.. one segment goes from 0 to 1 and
> > the other for 1 to 0. I'll try it out by incrementing too..
> >
>
> It's because your texture is symetric, by doing
> that you could not apply a text to your skidmaks for example,
> you can try, you'll see what I mean.
>

Hm, maybe. I modified it so that it goes beyond 1, but what happens is
that it works for the first 2 iterations of the texture in every new
skidmark, but then it just displays the same texture row. I set the
WARP to true when I load the texture, but it does not seem to help.

Anyway, I wrote the first version of the track howto. It can be
found at http://www.idiap.ch/~dimitrak/torcs/track-howto.html
The only current
problem is that shading seem so to only be applied to the track. That
happens even though the track-shade.ac generated has shadings for the
terrain too. (Verifiable by copying it to track.ac and loading it with
torcs).

Also, does the lightsource do anything apart from creating specular
highlights on the cars?



-- 
Christos Dimitrakakis
IDIAP (http://www.idiap.ch/~dimitrak/main.html)

Great I'll add the file to CVS and modify the output xml file too.

Note that races and championships are described in xml files
in the "config/raceman" directory, so it should be pretty easy
to configure your own race or championship.

Thanks,
Eric.

Eugen Treise wrote:
> Hi all!
> 
> I have send this email with an attachment yesterday, but it didn't 
> arrive at the list. So i try this time without the attachment.
> 
> I have created an XSL file (updated since last post, download from 
> http://eugentreise.de.tt/torcs.html) to format the results-*.xml files 
> into readable html files. This stylesheet makes it possible to view the
> results files directly withhin your browser. You can also convert 
> results files into html files by using an xslt processor like SAXON
> http://saxon.sourceforge.net/.
> 
> 
> The results-*.xml files have to be modified as folows:
> 
> - change path to params.dtd (for Windows system, not sure about Linux
> directory structure):
> <!DOCTYPE params SYSTEM "../../config/params.dtd">
> 
> - set reference to the xsl file, place the following line under the line
> with <!DOCTYPE... :
> <?xml-stylesheet type="text/xsl" href="../../config/style.xsl"?>
> 
> I placed style.xsl in config directory, where params.dtd is.
> Now open the modified results file with a browser (i tested with Mozilla
> 1.6 and IE 6.0).
> The stylesheet is applicable for all categories of result files.
> 
> Additional things, which could be done in results-*.xml files (code
> changes in torcs xml export functions):
> - Currently there is no car name in practice results file. It could be
> added to the Drivers or Practice section in practice mode, to see what
> car was driving.
> - "dammages" -> "damage"
> - results-*.xml files seem not always to follow the params.dtd rules 
> (params element), though it is not important.
> 
> Things i wasn't able to do:
> - format date (duration in seconds to date-time format)
> - move output of Standings to the end of html
> 
> Torcs developers, you can include the style.xsl file in torcs if you
> think it is usable. Modify it as you wish, the current html output is
> quite simple.
> 
> File is in PC format.
> 
> 
> Eugen
> 
> 
> 
> -------------------------------------------------------
> This SF.Net email is sponsored by: IBM Linux Tutorials
> Free Linux tutorial presented by Daniel Robbins, President and CEO of
> GenToo technologies. Learn everything from fundamentals to system
> administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click
> _______________________________________________
> Torcs-devel mailing list
> Tor...@li...
> https://lists.sourceforge.net/lists/listinfo/torcs-devel
> 
> 


-- 
=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=
                        TORCS
             The Open Racing Car Simulator
         AKA The Other Release Coming Soon (Skin'r)
                  http://torcs.org
=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=

Hi all!

I have send this email with an attachment yesterday, but it didn't 
arrive at the list. So i try this time without the attachment.

I have created an XSL file (updated since last post, download from 
http://eugentreise.de.tt/torcs.html) to format the results-*.xml files 
into readable html files. This stylesheet makes it possible to view the
results files directly withhin your browser. You can also convert 
results files into html files by using an xslt processor like SAXON
http://saxon.sourceforge.net/.


The results-*.xml files have to be modified as folows:

- change path to params.dtd (for Windows system, not sure about Linux
directory structure):
<!DOCTYPE params SYSTEM "../../config/params.dtd">

- set reference to the xsl file, place the following line under the line
with <!DOCTYPE... :
<?xml-stylesheet type="text/xsl" href="../../config/style.xsl"?>

I placed style.xsl in config directory, where params.dtd is.
Now open the modified results file with a browser (i tested with Mozilla
1.6 and IE 6.0).
The stylesheet is applicable for all categories of result files.

Additional things, which could be done in results-*.xml files (code
changes in torcs xml export functions):
- Currently there is no car name in practice results file. It could be
added to the Drivers or Practice section in practice mode, to see what
car was driving.
- "dammages" -> "damage"
- results-*.xml files seem not always to follow the params.dtd rules 
(params element), though it is not important.

Things i wasn't able to do:
- format date (duration in seconds to date-time format)
- move output of Standings to the end of html

Torcs developers, you can include the style.xsl file in torcs if you
think it is usable. Modify it as you wish, the current html output is
quite simple.

File is in PC format.


Eugen

Tommy Persson wrote:
>>From: Eric Espie <eri...@fr...>
>>
>>>- Make the road 1000 m wide and use road information from another
>>>  source to drive in a suitable way.  Is this possible?  Or is there
>>>  some limit on the width of a road?
>>
>>no limit, but the road is flat along the width, see the attached image,
>>it the view of the generated triangles.
> 
> 
> I managed to get this to work.  But I think I got warnings about
> initial fuel parameter to high for a track with width 1500 m and
> increased length.  Maybe I missed some parameter somewhere.

It's only a warning, I should fix it in the robots code...

> 
> This leads to more questions.  Is it possible to specify the x,y start
> coordinate for the first segment?  And can the track have negative x
> and y?

well, the first segment starts at (0,0) but may be the code of the
track loader (src/modules/track/...) can be changed.

> 
> 
> I have asked a lot of questions.  Maybe I should explain what I am
> trying to do.  I work in a research project with an autonomous
> helicopter.  One important part of the project is the autonomy and
> high level reasoning.  We are flying in an area in Sweden and we have
> a 3D model of this area and we have an OpenSceneGraph based visualiser
> for our simulations.  The helicopter has a camera so in the simulation
> we have to control the camera.  We use the camera to try to find cars
> and similar things
> 
> When I saw that you could have robots in torcs as dynamically loadable
> modules I thought it would be fun and maybe useful to use torcs for
> generating data (car positions) to the rest of our system and maybe
> use the torcs tracks to fly the helicopter in.  So the thing I have
> worked with the last week is interfacing to other programs.  So I have
> managed to:
> 
>    - Use torcs tracks in our visualizer and sending car positions from
>      torcs to our visualizer so I can see the cars there.  I can then
>      fly our simulated helicopter in our visualizer and to point the
>      camera at the cars.
> 
>    - I took our 3D model and managed to use it in torcs and using a
>      very wide road I could drive around in our model.
> 
> Things that would be nice to get to work:
> 
>    - Have crossings and drive the cars sensible in crossings.  For my
>      purpose that could be faked in some way since I just need a good
>      way to generate data of cars driving in crossings.

Currently if you generate crossings, you'll have to fix the 3D scenery
to remove the extra barriers.
The collision will work but the robots won't be aware of the crossings.

> 
>    - Translate the track specification to a format that our program
>      unserstand.
> 
>    - Control the camera in torcs from a robot.

calls from a dynamic lib to another dynamic lib will cause problems,
so I recommend that you write an interface in the tgfclient static lib
this will be the link between the src/modules/graphic/... lib and the robot.

> 
>    - Display a helicopter model in torcs at a position given by a robot.
>      (Is this possible now?)

with small modifications in the graphic part it should be the less difficult
part to modify.

> 
> 
> I really liked torcs.  It was easy to use.  It was relatively easy to
> interface to.  The possibility to have robots was very nice.  Nice
> work.

thanks

> 
> /Tommy Perssomn
> Linköping University
> Sweden
> 
> 

Bye,
Eric.
-- 
=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=
                        TORCS
             The Open Racing Car Simulator
         AKA The Other Release Coming Soon (Skin'r)
                  http://torcs.org
=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=

> From: Eric Espie <eri...@fr...>
> > - Make the road 1000 m wide and use road information from another
> >   source to drive in a suitable way.  Is this possible?  Or is there
> >   some limit on the width of a road?
>=20
> no limit, but the road is flat along the width, see the attached image,
> it the view of the generated triangles.

I managed to get this to work.  But I think I got warnings about
initial fuel parameter to high for a track with width 1500 m and
increased length.  Maybe I missed some parameter somewhere.

This leads to more questions.  Is it possible to specify the x,y start
coordinate for the first segment?  And can the track have negative x
and y?


I have asked a lot of questions.  Maybe I should explain what I am
trying to do.  I work in a research project with an autonomous
helicopter.  One important part of the project is the autonomy and
high level reasoning.  We are flying in an area in Sweden and we have
a 3D model of this area and we have an OpenSceneGraph based visualiser
for our simulations.  The helicopter has a camera so in the simulation
we have to control the camera.  We use the camera to try to find cars
and similar things

When I saw that you could have robots in torcs as dynamically loadable
modules I thought it would be fun and maybe useful to use torcs for
generating data (car positions) to the rest of our system and maybe
use the torcs tracks to fly the helicopter in.  So the thing I have
worked with the last week is interfacing to other programs.  So I have
managed to:

   - Use torcs tracks in our visualizer and sending car positions from
     torcs to our visualizer so I can see the cars there.  I can then
     fly our simulated helicopter in our visualizer and to point the
     camera at the cars.

   - I took our 3D model and managed to use it in torcs and using a
     very wide road I could drive around in our model.

Things that would be nice to get to work:

   - Have crossings and drive the cars sensible in crossings.  For my
     purpose that could be faked in some way since I just need a good
     way to generate data of cars driving in crossings.

   - Translate the track specification to a format that our program
     unserstand.

   - Control the camera in torcs from a robot.

   - Display a helicopter model in torcs at a position given by a robot.
     (Is this possible now?)


I really liked torcs.  It was easy to use.  It was relatively easy to
interface to.  The possibility to have robots was very nice.  Nice
work.

/Tommy Perssomn
Link=F6ping University
Sweden

> From: Eric Espie <eri...@fr...>
> 
> With AC3D this is the method I use:
> 
> 1 - CTRL-A to select all in "vertex" mode
> 2 - menu Vertex->"Snap together by distance" value 0.005
> 3 - switch in "surface" mode
> 4 - menu Surface->Triangulate
> 5 - switch in "object" mode
> 6 - menu Object->"Optimize vertices"
> 7 - menu Object->"Optimize surfaces"
> 
> Then your object must be ok to be loaded in TORCS.
> (if you are using AC3D 4.0 remove the "crease" lines from
> the generated .ac file)

Thanks for this information.  I did work to get an ac file I could
load.  I tried accc on it also but accc dumped core with:

cd /usr/local/torcs-1.2.2/share/games/torcs/tracks/revinge/revflat/; accc +etn revflat.ac revflat.acc
starting loading ...
loading  REVPHOTOSV object
Computing normals for REVPHOTOSV
found in REVPHOTOSV , a duplicate coord with not the same u,v, split is required
numtri = 4 on num =4
loading  REVHUS110 object
Computing normals for REVHUS110
found in REVHUS110 , a duplicate coord with not the same u,v, split is required
numtri = 2 on num =36
numtri = 2 on num =36
numtri = 2 on num =36
numtri = 4 on num =36
numtri = 8 on num =36
numtri = 2 on num =36
numtri = 2 on num =36
numtri = 2 on num =36
numtri = 4 on num =36
numtri = 2 on num =36
numtri = 4 on num =36
numtri = 2 on num =36
loading  REVHUS1F1 object
Computing normals for REVHUS1F1
found in REVHUS1F1 , a duplicate coord with not the same u,v, split is required
numtri = 2 on num =10
numtri = 1 on num =10
numtri = 3 on num =10
numtri = 4 on num =10
loading  REVHUS1F2 object
Computing normals for REVHUS1F2
found in REVHUS1F2 , a duplicate coord with not the same u,v, split is required
numtri = 2 on num =4
numtri = 2 on num =4
loading  REVHUS1F3 object
Computing normals for REVHUS1F3
found in REVHUS1F3 , a duplicate coord with not the same u,v, split is required
/usr/local/torcs-1.2.2/bin/accc: line 31: 31926 Segmentation fault      (core dumped) $LIBDIR/accc-bin $*


/Tommy Persson

Tommy Persson wrote:
>>From: Bernhard Wymann <be...@bl...>
>>
>>>I want to make a grid like road (alike city road maps). How can I do that in the
>>>xml file? multiple structures for <segments>?
>>
>>The current simulation does not allow that, the track must be one closed 
>>"string".
> 
> 
> I am also interesting in getting the simulation working in an area
> with branching roads.  I was going to try different things but I can
> ask here so I do not try anything the will not work.  So here are
> possible approaches:
> 
> - Make a closed track but make it overlapping in some ared.  For
>   example an "eight".  Will this work?

Yes, but trackgen won't be able to generate the terrain around
the track, you'll have to add it yourself.

> 
> - Make the road 1000 m wide and use road information from another
>   source to drive in a suitable way.  Is this possible?  Or is there
>   some limit on the width of a road?

no limit, but the road is flat along the width, see the attached image,
it the view of the generated triangles.

> 
> 
> /Tommy Persson
> 

Eric.
-- 
=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=
                        TORCS
             The Open Racing Car Simulator
         AKA The Other Release Coming Soon (Skin'r)
                  http://torcs.org
=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=

Ashwani Kumar wrote:
> Hi,
> 
> Please point me to the code where you build tracks. I have tried to dig 
> but nothing very useful. I have succeeded in removing the barriers.. but 
> the car will still not go outside the track.
> 
> Also, I want to develop multiple tracks like grids.. what could be a 
> good starting point for that?
> 
> Thanks a lot
> Ashwani
> 

The track structure is defined in src/interfaces/track.h
the track loader is in src/modules/track
Some helper functions are in src/libs/robottools/rttrack.cpp

The track structure is used widely in all the code,
(robots, simulation) and a complete rebuild of the structure
won't be easy :-(

Eric.
-- 
=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=
                        TORCS
             The Open Racing Car Simulator
         AKA The Other Release Coming Soon (Skin'r)
                  http://torcs.org
=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=

Tommy Persson wrote:
>>From: Eric Espie <eri...@fr...>
>>
>>>3. I tried loading the trackgen-generated track with ac3d to add some
>>>scenery, but after I save it again torcs cannot load it. Is there some
>>>extra step I need to take?
>>
>>You have to triangulate every surface, torcs only work with triangles.
>>(lines must be removed too).
>>Beware not to move the track itself because the cars follow the track
>>defined in the xml file not the 3D one.
>>
> 
> 
> So how do you triangulate your model?  I have a model I would like to
> use and tried to triangulate it in ac3d but it failed for some reason
> and said it probably was because of overlapping surfaces.  So are
> there any other tools to use that can fix this?
> 
> /Tommy Persson
> 

With AC3D this is the method I use:

1 - CTRL-A to select all in "vertex" mode
2 - menu Vertex->"Snap together by distance" value 0.005
3 - switch in "surface" mode
4 - menu Surface->Triangulate
5 - switch in "object" mode
6 - menu Object->"Optimize vertices"
7 - menu Object->"Optimize surfaces"

Then your object must be ok to be loaded in TORCS.
(if you are using AC3D 4.0 remove the "crease" lines from
the generated .ac file)

Eric.
-- 
=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=
                        TORCS
             The Open Racing Car Simulator
         AKA The Other Release Coming Soon (Skin'r)
                  http://torcs.org
=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=

Hi,

Please point me to the code where you build tracks. I have tried to dig 
but nothing very useful. I have succeeded in removing the barriers.. 
but the car will still not go outside the track.

Also, I want to develop multiple tracks like grids.. what could be a 
good starting point for that?

Thanks a lot
Ashwani