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

Hmmh, it seems that it works. It was quite easy, but the result is also, 
not absolutely terrific. :-(
If I understand it right, it should work as attached.  I tried it not 
only with my own driver, but also with the berniw tutorial robot. bt1 on 
etrack2 is with the ackermann radius in getAccel() 430 ms slower 
(1:22:85 instead of 1:22:42) then with the original implementation using 
getAllowedSpeed(...). I use in the attached code the v2d class of the 
berniw tutorial robot, createLinearFormula only calculates from two 
points m and b for f(x)=mx+b.

Bye,

Timo

Timo Steuerwald schrieb:

> Hi Bernhard,
> Hi Christos,
>
> I will try my best and post results as soon as possible. Thanks a lot 
> up to now!  :-)
>
> Bye,
>
> Timo
>
>
> -------------------------------------------------------
> SF email is sponsored by - The IT Product Guide
> Read honest & candid reviews on hundreds of IT Products from real users.
> Discover which products truly live up to the hype. Start reading now.
> http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click
> _______________________________________________
> Torcs-users mailing list
> Tor...@li...
> https://lists.sourceforge.net/lists/listinfo/torcs-users
>
>

> > Thanks, that's very useful info, maybe we can start making a
> > repository for
> > little bits of info like that?
>
> Sort of a wiki?
>

Could be a wiki, could be somewhere in the developer docs, I don't
know what's more suitable.

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

Hi All

> Thanks, that's very useful info, maybe we can start making a
> repository for
> little bits of info like that?

Sort of a wiki?

Bye, Bernhard.


-- 

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

>
> So a short summary for track textures:
> - "*_n.rgb" disables mipmapping
> - "tree", "arbor", "trans-" somewhere in the name enable alpha testing.
> - if you have blue or black borders or wrong colors caused by the
> rendering order you can do threshold alpha in gimp on the texture if it
> is acceptable and disabling mipmapping.
> - or you can enable the alpha test to make just almost opaque fragments
> pass, so the wrong background is not that obvious.
> - or combinations of them.
>
>

Thanks, that's very useful info, maybe we can start making a
repository for
little bits of info like that?

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

El ds 16 de 04 del 2005 a les 01:55 +0200, en/na Bernhard Wymann va
escriure:
> Hi Albert
> 
> > I found out that the all-in-one
> > torcs-1.2.3-linux-i686-experimental.bz2.run installer package for torcs
> > won't run by default on Ubuntu Hoary, as it seems there is a missing
> 
> What is the error? Do you have more information about that? It would be 
> very important to know what the problem was, otherwise we cannot fix it.

I cannot reproduce the bug in my Hoary laptop as I installed some of the
libraries for other unrelated stuff.

But I tested it in a Warty desktop where no gtk libs are installed and I
get:

avb@dotstation:~instalats $ sh
torcs-1.2.3-linux-i686-experimental.bz2.run
Verifying archive integrity... All good.
Uncompressing
torcs-1.2.3..................................................................................................
/home/avb/.setup3265: error while loading shared libraries:
libgtk-1.2.so.0: cannot open shared object file: No such file or
directory


> > libgtk-1.x dependency (trying to test the installation in Hoary with the
> > CD of games in mind).

Hopefully next week I will have a fresh Hoary in hands to try again and
confirm the exact error prompt I get.

> 
> That's weird, because it should start a static linked command line 
> installer if gtk is not around... did you try from a command 
> line/xterm/konsole "sh torcs-1.2.3-linux-i686-experimental.bz2.run", 
> perhaps you just clicked it in konqueror (perhaps I did run makeself 
> with the nox11 option, so no xterm opens up, but I cannot remember)?

> > After messing a little bit with those, I managed to get the menu running
> > in what it seems like a wxwindow environment (is that right?). The
> 
> No, GTK 1.2.
> 
> > Now a couple of questions:
> > 
> > (a) Which version of the libpng.so in Hoary is better to symlink?
> 
> I do not know, I would take the most recent one which works.

ok, noted.

> 
> > (b) Is there a way to launch the installer without the graphical
> > interface so that no extra packages are needed? Or is it an option to
> 
> Yes, should happen automatically...
> 
> > distribute the image installed by the experimental installer?
> 
> Why not compile from source then, so you can be sure that all the ABI's 
> match (its compiled with gcc 3.3.4) and choose the compiler switches 
> according to your requirements?

My feeling is that lack of experience in this field would cause more
troubles than anything... :-)

That's why the installer is a great add-on for easing the deployment.

Bests,

    Albert.

Thank !  :-)
It works as I want. :-)

Bye
Speedy



Bernhard Wymann a =E9crit :

> Hi SpeedyChonChon
>
> Yes, very easy. If you are interested what happens here, google about
> "mipmapping", its a method to remove aliasing artifacts. You can switch
> it off simply with renaming your texture, simply add an "_n" to the mai=
n
> name, example:
> before: "speedyasphalt.rgb"
> after : "speedyasphalt_n.rgb"
>
> That should do the trick. Of course you have to rename it as well in th=
e
> ac file, that works best in a text editor or with sed.
>
> Btw. if you want partially transparent objects like the new fence in
> aaborg or g-track-2:
> - name the file arbor*_n.rgb or tree*_n.rgb, there is IIRC a third one,
> but I cannot remember it currently (look it up in the loader source
> file, ac3dload or acload.cpp, something like that).
> - non transparent areas have to be more that 70% opaque (alpha).
>
> Bye, Bernhard.
>
>


--=20
SpeedyChonChon

----------------------------------------------------------
AIM : /SpeedyChonChon/
ICQ : /302814595/
MSN : /ab...@ho.../
http://speedy.chonchon.free.fr
----------------------------------------------------------

Hi Albert

> I found out that the all-in-one
> torcs-1.2.3-linux-i686-experimental.bz2.run installer package for torcs
> won't run by default on Ubuntu Hoary, as it seems there is a missing

What is the error? Do you have more information about that? It would be 
very important to know what the problem was, otherwise we cannot fix it.

> libgtk-1.x dependency (trying to test the installation in Hoary with the
> CD of games in mind).

That's weird, because it should start a static linked command line 
installer if gtk is not around... did you try from a command 
line/xterm/konsole "sh torcs-1.2.3-linux-i686-experimental.bz2.run", 
perhaps you just clicked it in konqueror (perhaps I did run makeself 
with the nox11 option, so no xterm opens up, but I cannot remember)?

> After messing a little bit with those, I managed to get the menu running
> in what it seems like a wxwindow environment (is that right?). The

No, GTK 1.2.

> Now a couple of questions:
> 
> (a) Which version of the libpng.so in Hoary is better to symlink?

I do not know, I would take the most recent one which works.

> (b) Is there a way to launch the installer without the graphical
> interface so that no extra packages are needed? Or is it an option to

Yes, should happen automatically...

> distribute the image installed by the experimental installer?

Why not compile from source then, so you can be sure that all the ABI's 
match (its compiled with gcc 3.3.4) and choose the compiler switches 
according to your requirements?

Thank you very much for your report, we will just be able to improve the 
binary with feedback from different distros. Could you please try to run 
the installer like I suggested above and report the result?

Bye, Bernhard.


-- 

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

Hi all,

I found out that the all-in-one
torcs-1.2.3-linux-i686-experimental.bz2.run installer package for torcs
won't run by default on Ubuntu Hoary, as it seems there is a missing
libgtk-1.x dependency (trying to test the installation in Hoary with the
CD of games in mind).

After messing a little bit with those, I managed to get the menu running
in what it seems like a wxwindow environment (is that right?). The
problem is that the wxwindows aren't included by default in Hoary, so
that should be manually added beforehand. This wouldn't be a problem if
the "image" directory is what goes distributed in the CDs.

The problem right now is that I have the installed "image" looking for a
libpng.so.3 dependency:

avb@magneto:~/instalats/torcs/torcs-1.2.3$ ./torcs
/home/avb/instalats/torcs/torcs-1.2.3/lib/torcs-bin: error while loading
shared libraries: libpng.so.3: cannot open shared object file: No such
file or directory

So I added a symlink to one of the libpngs present in Hoary:

avb@magneto:~/instalats/torcs/torcs-1.2.3$ ll /usr/lib/libpng*so*
lrwxrwxrwx  1 root root     20 2005-03-29 15:13 /usr/lib/libpng10.so.0
-> libpng10.so.0.1.0.18
-rw-r--r--  1 root root 138440 2004-12-05
04:39 /usr/lib/libpng10.so.0.1.0.18
lrwxrwxrwx  1 root root     19 2005-03-29 14:59 /usr/lib/libpng12.so.0
-> libpng12.so.0.1.2.8
-rw-r--r--  1 root root 142612 2004-12-05
04:39 /usr/lib/libpng12.so.0.1.2.8

avb@magneto:~/instalats/torcs/torcs-1.2.3$ sudo ln
-s /usr/lib/libpng12.so.0 /usr/lib/libpng.so.3

And then it all goes very well.

Now a couple of questions:

(a) Which version of the libpng.so in Hoary is better to symlink?
(b) Is there a way to launch the installer without the graphical
interface so that no extra packages are needed? Or is it an option to
distribute the image installed by the experimental installer?

Thanks in advance,

    Albert.

Hi Christos

>>- non transparent areas have to be more that 70% opaque (alpha).
> 
> 
> Meaning that when alpha > 70%, the object is completely opaque?

No, look up glAlphaTest, basically it discards all fragments which are 
below 0.7. The fragments above pass without modification. Be aware that 
mipmaps as well filter the alpha value, so even if your pixels are above 
70% they can drop below in the mipmaps if they have transparent neighbours).

Because track side objects are not depth sorted (IIRC, might be wrong) 
you can apply a function (I think it was layer, threshold alpha in gimp 
2.0) to make alpha either 0.0 or 1.0, so the objects are in the 
transparent areas full transparent and in the opaque parts full opaque.

That helps if there is no depth sorting in place to avoid color 
combination artifacts like the slightly blue borders of some trees. They 
are caused by the rendering order: sky box (e.g. blue), tree (tree 
combined with blue) and finally the landscape (green). Because the 
z-Buffer for the tree is already set the tree combined with blue remains 
in the "wrong" (unintended) combined color.

This ispires me to an idea to try later, perhaps I can scrap the alpha 
test on the trees which I have created (simply rename them from tree to 
something else), because they have IIRC (hey, its already some weeks 
ago...) just 0.0/1.0 alpha values.

Perhaps this will speed up things a bit (I have the feeling that at 
least some graphics boards do not like very much alpha testing, but that 
might be wrong, I have to measure it first, it could as good increase 
performance because the fragments are discarded earlier).

So a short summary for track textures:
- "*_n.rgb" disables mipmapping
- "tree", "arbor", "trans-" somewhere in the name enable alpha testing.
- if you have blue or black borders or wrong colors caused by the 
rendering order you can do threshold alpha in gimp on the texture if it 
is acceptable and disabling mipmapping.
- or you can enable the alpha test to make just almost opaque fragments 
pass, so the wrong background is not that obvious.
- or combinations of them.


Bye, Bernhard


-- 

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

Hi Christos

>>Thats good. IMHO most papers which do not try to be user friendly are
>>junk, that's perhaps because these things are called papers, to name the
>>valuable part;-)
>>
> 
> 
> hee

;-)

> dv/dt v(s) = a + b v(s)^2,

No, I stated it wrong in the previous post, its:

dv/ds*v(s) = a + b*v(s)²

There is no time explicitely involved at all, I like more the "energy" 
than the "acceleration" approach for that problem, its more intuitive 
for me, but that is a matter of preference.

> dv/dt is basically the deceleration. Numerically integrating this and
> comparing with your analytic solution agrees perfectly... oh perhaps
> it is just a notation thingy. You by dv/dt v(s) you just mean the
> derivative of v with respect to t, with v having the value v(s),
> right?

No, dv/ds is the change (derivative) of the speed over the braking 
distance and v(s) is the speed at distance (position) s. Btw. dv/ds 
quite a funny property because if you think about its unit it must be 
[1/s]. I usually do just a dimension check on intermediate results and 
do not think too much about its meaning;-)

Bye, Bernhard.


-- 

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

>
> Thats good. IMHO most papers which do not try to be user friendly are
> junk, that's perhaps because these things are called papers, to name the
> valuable part;-)
>

hee

> Hmm, I got this in the robot tuorial and I'm quite sure that it is correc=
t:
> dv/dt*v =3D a + b*v=B2.
>
> http://www.berniw.org/torcs/robot/ch3/braking3.html
>

Hm, your function is

dv/dt v(s) =3D a + b v(s)^2,

while mine is

dv/dt =3D a + b v(t)^2

dv/dt is basically the deceleration. Numerically integrating this and
comparing with your analytic solution agrees perfectly... oh perhaps
it is just a notation thingy. You by dv/dt v(s) you just mean the
derivative of v with respect to t, with v having the value v(s),
right?


> > is some kind of second order differential equation. What can one do
> > with such a thing?
>
> Lazy as I am I would feed it to maple or lookup in my diffeq collection
> if somebody clever already solved it (my math education is some years
> ago and I would need to refresh quite some stuff):-)
>
> Bye, Bernhard.
>
>

--=20
Christos Dimitrakakis
IDIAP (http://www.idiap.ch/~dimitrak/main.html)

Hi all

>> PS. I was looking at the 'brake distance' function and tried to get it
>> from first principles, so I arrived at a form dv/dt = a + bv^2, which
> 
> 
> Hmm, I got this in the robot tuorial and I'm quite sure that it is correct:
> dv/dt*v = a + b*v².
> 
> http://www.berniw.org/torcs/robot/ch3/braking3.html

Sorry I was confused, you have v(t), mine has v(s), so its not the same 
animal, sorry.

Bye, Bernhard.


-- 

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

Hi Christos

> BTW, I was working on a tutorial on estimation & control, with

I skimmed through it, great, that looks very promising:-) I will have a 
deeper look at it next week.

> application on vehicle control (i.e. TORCS). Maybe you can contribute
> a few things? I aim it to be user-friendly, but sufficiently in-depth
> mathematically
> to wet the appetite of people comfortable with some math that have
> little background in statistics.

Thats good. IMHO most papers which do not try to be user friendly are 
junk, that's perhaps because these things are called papers, to name the 
valuable part;-)

> PS. I was looking at the 'brake distance' function and tried to get it
> from first principles, so I arrived at a form dv/dt = a + bv^2, which

Hmm, I got this in the robot tuorial and I'm quite sure that it is correct:
dv/dt*v = a + b*v².

http://www.berniw.org/torcs/robot/ch3/braking3.html

> is some kind of second order differential equation. What can one do
> with such a thing?

Lazy as I am I would feed it to maple or lookup in my diffeq collection 
if somebody clever already solved it (my math education is some years 
ago and I would need to refresh quite some stuff):-)

Bye, Bernhard.

-- 

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

> > Hello peeps,
>
> Short for people? (like prolly -> probably)
>

Yes, though I can't remember anyone using it recently apart from me.

> > To be more precise, an approximation of the radius, based on the
> > assumption that there is sufficient friction. Probably on more
> > assumptions, I can try and detail it if you want.
>
> Hm. The problem statement was "maximum speed depending on the actual
> steer rad value". The function we are looking for is maxspeed(steer,
> carmodel), isn't it?
>
> If you do not assume these things, the problem makes no sense anymore,
> no? (because you can steer as much as you like for any speed, and up to
> a certain degree the radius will shrink with increasing steering, so the
> function would be a one liner: "return infinity;" )...
>

Yes, you are correct. I was confused.

> > I don't know if it is a sufficiently good calculation for when you're
> > driving at the limit.
>
> I expect it to work quite well for big radiuses with low centrifugal
> force, but I do not bet any money;-)
>

Hm, given that at the limit, the centrifugal force in the rotating
frame of reference equals the centripedal force in the inertial frame,
I expect the calculation to be exact. The only modelling problem is
the relationship between steering and radius, which could be
troublesome if the car is not set up to be neutral. But as
you say, the error cannot be very big for properly set up cars.

It would be interesting to see what the modelling error is when you
try to calculate the centripetal force from the steering angle
instead.

> > very good, maybe a parametric estimation method would be
> > more appropriate. Estimation methods have their own problems, but in
> > *theory* they allow you to build as an accurate model of the
> > system as you want.
>
> This would be useful for other problem statements if you could keep the
> model simple and perhaps find a fitting function or a precomputed table
> to interpolate the needed values on the fly for e.g radius(speed, carmodel).
> To use your suggested method with the problem one would need to state
> the problem differently I think (e.g. introduce a maximum allowed angle
> between the speed vector and the steering wheels).
>

In theory it is possible to estimate any function you want. In
practice you frequently have problems calculating inverse solutions
from estimates, so it might be better to estimate invserse models,
though that has its own robustness problems.

BTW, I was working on a tutorial on estimation & control, with
application on vehicle control (i.e. TORCS). Maybe you can contribute
a few things? I aim it to be user-friendly, but sufficiently in-depth
mathematically
to wet the appetite of people comfortable with some math that have
little background in statistics.

http://www.idiap.ch/~dimitrak/simu/Estimates.ps.gz
http://www.idiap.ch/~dimitrak/simu/Estimates.pdf

PS. I was looking at the 'brake distance' function and tried to get it
from first principles, so I arrived at a form dv/dt = a + bv^2, which
is some kind of second order differential equation. What can one do
with such a thing?

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

> - non transparent areas have to be more that 70% opaque (alpha).

Meaning that when alpha > 70%, the object is completely opaque?
-- 
Christos Dimitrakakis
IDIAP (http://www.idiap.ch/~dimitrak/main.html)

Hi SpeedyChonChon

> I wonder if it's possible to delete this blur 
> <http://speedy.chonchon.free.fr/images/torcs/flou.jpg> (the start line 
> for an example).

Yes, very easy. If you are interested what happens here, google about
"mipmapping", its a method to remove aliasing artifacts. You can switch
it off simply with renaming your texture, simply add an "_n" to the main
name, example:
before: "speedyasphalt.rgb"
after : "speedyasphalt_n.rgb"

That should do the trick. Of course you have to rename it as well in the
ac file, that works best in a text editor or with sed.

Btw. if you want partially transparent objects like the new fence in
aaborg or g-track-2:
- name the file arbor*_n.rgb or tree*_n.rgb, there is IIRC a third one,
but I cannot remember it currently (look it up in the loader source
file, ac3dload or acload.cpp, something like that).
- non transparent areas have to be more that 70% opaque (alpha).

Bye, Bernhard.


-- 

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

Bernhard Wymann schrieb:

> Hi Timo
>
> If that does not help let me know, I can derive that quite quickly (I 
> have a very similar function ("radius") in berniw.h (it takes three 
> points instead of the wheelbase and an angle, but its basically the 
> same). If you figured it out, please post it:-)
>
Hi Bernhard,
Hi Christos,

I will try my best and post results as soon as possible. Thanks a lot up 
to now!  :-)

Bye,

Timo

Hi

I wonder if it's possible to delete this blur 
<http://speedy.chonchon.free.fr/images/torcs/flou.jpg> (the start line 
for an example).

Thank
-- 
SpeedyChonChon

----------------------------------------------------------
AIM : /SpeedyChonChon/
ICQ : /302814595/
MSN : /ab...@ho.../
http://speedy.chonchon.free.fr
----------------------------------------------------------

Hi all

> Hello peeps,

Short for people? (like prolly -> probably)

> To be more precise, an approximation of the radius, based on the
> assumption that there is sufficient friction. Probably on more
> assumptions, I can try and detail it if you want.

Hm. The problem statement was "maximum speed depending on the actual 
steer rad value". The function we are looking for is maxspeed(steer, 
carmodel), isn't it?

If you do not assume these things, the problem makes no sense anymore, 
no? (because you can steer as much as you like for any speed, and up to 
a certain degree the radius will shrink with increasing steering, so the 
function would be a one liner: "return infinity;" )...

> I don't know if it is a sufficiently good calculation for when you're
> driving at the limit.

I expect it to work quite well for big radiuses with low centrifugal 
force, but I do not bet any money;-)

>>If that does not help let me know, I can derive that quite quickly (I
>>have a very similar function ("radius") in berniw.h (it takes three
>>points instead of the wheelbase and an angle, but its basically the
>>same). If you figured it out, please post it:-)
>>
> 
> 
> The berniw calculation is 'inscribe a circle on 3 points', right?

Yes.

> Does the wheel base calculation boil down to 'intersection of two
> lines'? If so, you still have a problem defining a circle. Will the
> center be at the intersection and the radius from the intersection to
> the middle of the fixed wheel axis? I think that creates some
> technical problems, but maybe you can ignore them.

Exactly:-) You talk about an error of perhaps 1 meter, so if the radius 
is 20 meters (what is quite tight) it is already negligible. I really 
think that a 5-10% approximation is good enough, because there are a lot 
of unknowns anyway (e.g. you do not know easily the normal forces of 
each wheel, evaluating the tire model is not too cheap as well).

> I suggest you try these methods first. If they turn out to not be

Mee too, of course. Somtimes I had the "sad" experience that the simple 
stuff worked as good or even better than the full scientific pull, that 
can be quite frustrating if the 30 minutes approch is almost as good or 
better than the 20 hours full pull...
When you keep it simple at the start you gain as well quickly more 
experience about the problem, so you can fix things before you wasted 
too much time (e.g. "Do I try to do the right thing?").

> very good, maybe a parametric estimation method would be
> more appropriate. Estimation methods have their own problems, but in
> *theory* they allow you to build as an accurate model of the
> system as you want.

This would be useful for other problem statements if you could keep the 
model simple and perhaps find a fitting function or a precomputed table 
to interpolate the needed values on the fly for e.g radius(speed, carmodel).
To use your suggested method with the problem one would need to state 
the problem differently I think (e.g. introduce a maximum allowed angle 
between the speed vector and the steering wheels).

Bye, Bernhard.


-- 

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

Hello peeps,

>
> > I want to compute in getAccel() (I use as basis the tutorial robot from
> > berniw) the maximum speed depending on the actual steer rad value.
> > I'm sure it is possible to compute a radius out of the steer value, but
> > I am really bad in maths, so can please anybody help?
>

> Ackermanns law:
> http://en.wikipedia.org/wiki/Ackermann_steering_geometry
>
> To compute it you just need the wheelbase (distance between rear and
> front "axis", I guess I compute that in berniw if you need to look up)
> and the steering angle, they define two tangents. Now you can draw the
> two perpendicular straights, and you get the center. From that you can
> compute the radius.
>

To be more precise, an approximation of the radius, based on the
assumption that there is sufficient friction. Probably on more
assumptions, I can try and detail it if you want.

I don't know if it is a sufficiently good calculation for when you're
driving at the limit.

> If that does not help let me know, I can derive that quite quickly (I
> have a very similar function ("radius") in berniw.h (it takes three
> points instead of the wheelbase and an angle, but its basically the
> same). If you figured it out, please post it:-)
>

The berniw calculation is 'inscribe a circle on 3 points', right?

Does the wheel base calculation boil down to 'intersection of two
lines'? If so, you still have a problem defining a circle. Will the
center be at the intersection and the radius from the intersection to
the middle of the fixed wheel axis? I think that creates some
technical problems, but maybe you can ignore them.

I suggest you try these methods first. If they turn out to not be
very good, maybe a parametric estimation method would be
more appropriate. Estimation methods have their own problems, but in
*theory* they allow you to build as an accurate model of the
system as you want.


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

Hi Timo

> I want to compute in getAccel() (I use as basis the tutorial robot from 
> berniw) the maximum speed depending on the actual steer rad value.
> I'm sure it is possible to compute a radius out of the steer value, but 
> I am really bad in maths, so can please anybody help?

You can do an approximation of the radius if you assume that there is no 
centrifugal force (Ackermann's law, see link below). Given the possible 
maximum friction force and the radius you can compute an approximation 
of the possible speed (btw. I do not now without more thinking if it is 
possible to get an explicit formula -> so perhaps numeric solving or a 
speed independent maximum friction force approximation is necessary).

To compute it exactly is much more difficult because the whole dynamic 
of the car, track and the tires matter.

Ackermanns law:
http://en.wikipedia.org/wiki/Ackermann_steering_geometry

To compute it you just need the wheelbase (distance between rear and 
front "axis", I guess I compute that in berniw if you need to look up) 
and the steering angle, they define two tangents. Now you can draw the 
two perpendicular straights, and you get the center. From that you can 
compute the radius.

If that does not help let me know, I can derive that quite quickly (I 
have a very similar function ("radius") in berniw.h (it takes three 
points instead of the wheelbase and an angle, but its basically the 
same). If you figured it out, please post it:-)

Have fun, bye,

Bernhard.


-- 

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

Hello All!

I want to compute in getAccel() (I use as basis the tutorial robot from 
berniw) the maximum speed depending on the actual steer rad value.
I'm sure it is possible to compute a radius out of the steer value, but 
I am really bad in maths, so can please anybody help?

Thanks in advance!

Bye,

Timo

Hi Ethan

> I ran the windows .bat file and then I opened the workspace, set to
> release and compiled as per the readme. Are you using the CVS sources?

No, just releases are compiling usually on Windows, Linux is the main 
platform, the porting starts usually short before release time. I found 
that I have perhaps forgotten to update one of the docs, please follow 
the instructions of the README and use the all-in-one source package. 
Everything else is unsupported, like it is mentioned in the release 
notes on sf.net in the file section.

Did you run setup_win32-data-from-CVS.bat?

Bye, Bernhard.

 From README:

Windows Installation from Source
--------------------------------
- requires VC++ 6.0
- untar the archive into a path without whitespaces and special characters.
- cd into the torcs-1.2.3 directory
- run setup_win32.bat
- run setup_win32-data-from-CVS.bat
- open "TORCS.dsw" with VC++ 6.0, hit cancel for the missing projects.
- do not try to build a debug version, the projects are not set up!
- select the TORCS project and the w32-Release version
- compile project
- cd into the "runtime" directory.
- run "wtorcs.exe"



-- 

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

Hi Ethan

> I get this error when compiling torcs-1.2.3 on Visual C++ 6.0 SP5
> 
>  
> 
> It looks like there is an xml.h file though in the src directory. Do you
> know what I could do to fix this? 

I guess you did not read the readme(s), follow the instructions (it 
works with vc++ 6.0 sp5, exactly my setup).

Bye, Bernhard.


-- 

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

> Greetings all,
>
> I'm very new to the TORCS project. I've downloaded the source and been
> through a bit of it, but I think I'm missing something. All the code I have
> seems to be just for an application shell. Is there a file or group of files
> that specficially deal with car physics? That's really all I'm interested in
> and I can't seem to find it.
>
> Even more specifically the angular acceleration stuff...  I've been through
> a bunch of online tutorials (car physics for gamers especially) but there's
> a lot of information lacking and the sample code is laden with errors.
>

The source simulation is in src/modules/simu/
Simuv2 is simplified physics.
Simuv3 should have proper angular momentum and stuff, but only the CVS
version, not the source tarball.

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