rec.autos.simulators

by **Haqsa** » Sun, 27 Apr 2003 10:29:37

Anybody know how to fix the NR 2003 AI so they don't rear end you in the
braking zones on short tracks and roadies? This is absurd. There isn't a
single track that requires braking where they are not hitting me in the ass
if they follow me into the braking zone. And it doesn't occur at the
beginning of the braking zone it is in the middle or even towards the end.
They do not estimate their distance properly when following somebody into
the braking zone. I thought this was pretty basic - if you follow somebody
into a braking zone you have to brake WHEN they brake, not WHERE they brake.
The AI does not seem to know that, they run the lap the same way regardless
of whether someone is in front of them or not. The really annoying thing is
that this worked in GPL (other than you-know-who), but it has not worked
right in any of the NASCAR games they have made since then. Does anyone
know a file I can edit to fix this? Anyone know if it is being worked on?
If not I think I am done with this game.

by **Marc Collin** » Sun, 27 Apr 2003 13:31:22

Agreed this is a serious problem. Look back a bit at my posts about
Marty--it is basically undriveable because of the "blind" AI. The behaviour
of the AI during practise mode at the larger ovals also suggests they are
blind. Look at how they pile into an incident as though it wasn't there.
Unless a yellow flag comes on, which it doesn't during practise mode, the AI
range of view out the front seems to be about 2 or 3 virtual feet in front
of the car--far too little to react to anything be that braking points,
incidents, whatever.

You are also correct that the NASCAR series has been far, far, far behind
GPL in this regard. As the most extreme example, which I try in all sims as
a test, is drive backwards on a track in GPL and watch the AI react to you.
Then try it in N3, N4, NR2002 or NR2003. Case rested.

Marc

by **Haqsa** » Mon, 28 Apr 2003 03:38:59

Well I found one possible source of help here:
http://www.tptcc.com/board/viewtopic.php?t=23. Appparently there is a tool
built into NR2003 that allows you to build your own AI corridors. Not sure
if that will address this particular problem, but my thinking is that if one
deliberately drove laps in which you braked early at all the critical
braking points, and then built new .lp files from this, that this might
force the AI to back off a bit. I really don't have time to play with this
stuff though, I really would like to see Papy address this themselves.

by **William Bradsha** » Mon, 28 Apr 2003 04:12:43

> Well I found one possible source of help here:
> http://www.tptcc.com/board/viewtopic.php?t=23. Appparently there is a tool
> built into NR2003 that allows you to build your own AI corridors. Not sure
> if that will address this particular problem, but my thinking is that if one
> deliberately drove laps in which you braked early at all the critical
> braking points, and then built new .lp files from this, that this might
> force the AI to back off a bit. I really don't have time to play with this
> stuff though, I really would like to see Papy address this themselves.

For what it's worth, I do think Papy should fix the problem in a patch,
too. From the sounds of it, it hampers the otherwise good AI.

-Will

by ZP » Mon, 28 Apr 2003 07:47:35

The AutoAIStrength works well for me. It's a feature (you can find in the
playrecs.ini file, in the player folder, a demonstration of what i'm
speaking) that once you have run a race in a particular track the AI is
automatically adjusted to fit your speed for future races on that track.

It's not the adaptive speed however.

by **Marc Collin** » Mon, 28 Apr 2003 11:48:39

Won't help a lot--you'll spend hours tweaking and then you'll be nailing
them in the rear because they brake too soon. They largely brake OK now,
except that they ignore the presence of the player. I assume they ignore
the presence of the other AI, too, but because they all act so similarly, it
doesn't present problems.

GPL was actually quite forgiving to newbies in this regard. Once you get
within 5% of the speed of the AI, there are very, very few problems (we all
know the one notorious one). The AI seem to be especially wary of a car
driving erratically (the player!!).

In NR2003, you have to be within 0.5% of the speed of the AI to avoid being
treated like a pinball. I can only assume that increasing the AI's
awareness distance kills the frame rates and/or causes mayhem on pit lane.
Otherwise, Papy would have fixed it long ago.

Marc

> Well I found one possible source of help here:
> http://www.tptcc.com/board/viewtopic.php?t=23. Appparently there is a
tool
> built into NR2003 that allows you to build your own AI corridors. Not
sure
> if that will address this particular problem, but my thinking is that if
one
> deliberately drove laps in which you braked early at all the critical
> braking points, and then built new .lp files from this, that this might
> force the AI to back off a bit. I really don't have time to play with
this
> stuff though, I really would like to see Papy address this themselves.

> > Agreed this is a serious problem. Look back a bit at my posts about
> > Marty--it is basically undriveable because of the "blind" AI. The
> behaviour
> > of the AI during practise mode at the larger ovals also suggests they
are
> > blind. Look at how they pile into an incident as though it wasn't
there.
> > Unless a yellow flag comes on, which it doesn't during practise mode,
the
> AI
> > range of view out the front seems to be about 2 or 3 virtual feet in
front
> > of the car--far too little to react to anything be that braking points,
> > incidents, whatever.

> > You are also correct that the NASCAR series has been far, far, far
behind
> > GPL in this regard. As the most extreme example, which I try in all
sims
> as
> > a test, is drive backwards on a track in GPL and watch the AI react to
> you.
> > Then try it in N3, N4, NR2002 or NR2003. Case rested.

> > Marc

by **Marc Collin** » Mon, 28 Apr 2003 11:49:26

Not relevant to what we are discussing here. But I'm glad you like using
it.

Marc

by **Haqsa** » Mon, 28 Apr 2003 23:48:37

Marc have you tried playing with the track.ini at all? I found a few
parameters in there which might help, but not sure when I will have time to
test them.

From the Watkins Glen track.ini:

ai_decel_modifier = 0.77 ; braking grip efficiency
ai_drafting_distance = 1.15 ; car lengths behind while
racing
ai_dlongpad_scale = 1.8 ; scales padding in braking
zones (smaller = more scaling)
ai_dlat_pad = 0.8

Now the same values from Infineon:

ai_decel_modifier = 0.78 ; braking grip efficiency
ai_drafting_distance = 1.15 ; car lengths behind while
racing
ai_dlongpad_scale = 1.5 ; scales padding in braking
zones (smaller = more scaling)
ai_dlat_pad = 0.8

From Martinsville:

ai_decel_modifier = 0.94 ; braking grip efficiency
ai_drafting_distance = 1.07
ai_dlongpad_scale = 5.0
ai_dlat_pad = 0.9

And from Atlanta, to show how different the speedway AI is:

ai_decel_modifier = 0.70 ; braking grip efficiency
ai_drafting_distance = 1.25 ; car lengths behind while
racing
ai_dlongpad_scale = 25.0 ; scales padding in braking
zones (smaller = more scaling)
ai_dlat_pad = 0.85

First it seems like the AI braking grip at Martinsville is way off compared
to other tracks, so it might be helpful to reduce the ai_decel_modifier
there. Drafting distance might not be a major effect, but again if they
can't anticipate properly then increasing the drafting distance a tad might
help. The ai_dlongpad_scale is the one that I think will really make a
difference. Have you tried it? I found that at Infineon I wasn't getting
rear ended much, and notice that it has the smallest value for
ai_dlongpad_scale. Perhaps reducing it to 1.4 at all the problem tracks
would help. Also ai_dlat_pad is not documented in the files I looked at but
would seem to have something to do with the lateral separation the AI uses
in the braking zone. I think changing this might help another problem I
had, which was that AI would go for the dive bomb pass without enough room,
tag me as they tried to go around, and spin me in the process. It is not
clear whether it should be reduced or increased though, it doesn't appear to
work the same way as ai_dlongpad_scale.

Unfortunately I have guests here this weekend and I don't know when I will
have time to test any of this stuff. If anyone has tried or will try any of
these parameters I would really like to hear what kind of results you get.

Thanks,
Hal

by **Jay Taylo** » Tue, 29 Apr 2003 05:10:35

The speed and braking you use during the creation of the AI's lines has
little effect on their speeds or braking. The smoothness of the lines is
how the AI calculate their speeds, not the actual speeds values recorded in
the lp's. Your best bet is to try to go into your track.ini and look in the
ai_track section and adjust this value.....

ai_dlongpad_scale = X.X ; scales padding in braking zones (smaller = more
scaling)

Jay Taylor

> Well I found one possible source of help here:
> http://www.tptcc.com/board/viewtopic.php?t=23. Appparently there is a
tool
> built into NR2003 that allows you to build your own AI corridors. Not
sure
> if that will address this particular problem, but my thinking is that if
one
> deliberately drove laps in which you braked early at all the critical
> braking points, and then built new .lp files from this, that this might
> force the AI to back off a bit. I really don't have time to play with
this
> stuff though, I really would like to see Papy address this themselves.

> > Agreed this is a serious problem. Look back a bit at my posts about
> > Marty--it is basically undriveable because of the "blind" AI. The
> behaviour
> > of the AI during practise mode at the larger ovals also suggests they
are
> > blind. Look at how they pile into an incident as though it wasn't
there.
> > Unless a yellow flag comes on, which it doesn't during practise mode,
the
> AI
> > range of view out the front seems to be about 2 or 3 virtual feet in
front
> > of the car--far too little to react to anything be that braking points,
> > incidents, whatever.

> > You are also correct that the NASCAR series has been far, far, far
behind
> > GPL in this regard. As the most extreme example, which I try in all
sims
> as
> > a test, is drive backwards on a track in GPL and watch the AI react to
> you.
> > Then try it in N3, N4, NR2002 or NR2003. Case rested.

> > Marc

by **Haqsa** » Tue, 29 Apr 2003 11:21:08

I just had enough time to try only a few variations on that tonight, but
what I tried didn't seem to fix the problem. I read that comment as meaning
that I want smaller values for it. As I posted in another note in this
thread, the speedways use very large values for this number, short tracks
less, road courses least, so I assumed that I would want to make it smaller
still. At the Glen I cut it from 1.8 to 0.8 and I'm still getting punched
in the rear at the entry to the inner loop. However while I was doing this
I took a look at some of the replays, and I think I'm starting to see a
pattern. Decreasing that number did cause them to give me more room
initially, so that they hit me later in the braking zone, but the reason
they were hitting me turned out to be due to the fact that they were trying
to go around me and not giving me enough room. So now I am thinking that
ai_dlatpad is the one that needs changing, or maybe both. Hopefully I will
have time to try this tomorrow.

by **Haqsa** » Tue, 29 Apr 2003 11:33:19

That's really what started all this trouble. I ran a race at the Glen with
Auto AI Strength on and afterwards it set the strength at 102%. That
sounded great and I felt real good until I tried racing at that speed and
got run over going up the hill on the first lap of practice. I'm running
1:15's, why it decided that it should set the AI at 102% is beyond me, it
should have been more like 90%. So I started manually setting the AI
strength but I'm finding that even with them cut down to my lap speeds I am
still getting dive bombed in the slow corners, particularly at the entrance
to the inner loop. I think Marc is right, the root problem is the AI just
don't know what's in front of them. However since we have had that problem
for three years now I am not holding my breath that Papy is going to fix it.
So I am trying to find AI file tweaks to work around the problem, even if it
means forcing the AI to stay farther away from me. Thanks for the
suggestion though.

by **Haqsa** » Wed, 30 Apr 2003 11:48:33

Well Marc I am making very definite progress. Three variables in the
track.ini are having a very strong effect on the braking zone problem. I
have been running practice sessions at the Glen and today for the first time
I had a fairly good one. I changed these variables to the values shown:

ai_drafting_distance = 1.2 ; (was 1.15) car lengths
behind while racing
ai_dlongpad_scale = 0.8 ; (was 1.8) scales padding
in braking zones (smaller = more scaling)
ai_dlat_pad = 1.2 ; (was 0.8)

With these values I was still getting light taps occasionally, but they were
well within control and most of the time if I made a deliberate mistake the
AI simply went around me without contact. It's not perfect yet, I still
need to do some fine tuning, but I think the cure to the braking zone
problem lies in those three variables. Won't fix the problems with accident
avoidance, but it should make the road courses and maybe even the short
tracks much more playable.

by **R Knap** » Fri, 02 May 2003 05:50:50

It's kinda sad that we still have these kinds of problems at this stage of
the game. There is so much to like about Papy games but this lousy AI is
aggravating!

> I just had enough time to try only a few variations on that tonight, but
> what I tried didn't seem to fix the problem. I read that comment as
meaning
> that I want smaller values for it. As I posted in another note in this
> thread, the speedways use very large values for this number, short tracks
> less, road courses least, so I assumed that I would want to make it
smaller
> still. At the Glen I cut it from 1.8 to 0.8 and I'm still getting punched
> in the rear at the entry to the inner loop. However while I was doing
this
> I took a look at some of the replays, and I think I'm starting to see a
> pattern. Decreasing that number did cause them to give me more room
> initially, so that they hit me later in the braking zone, but the reason
> they were hitting me turned out to be due to the fact that they were
trying
> to go around me and not giving me enough room. So now I am thinking that
> ai_dlatpad is the one that needs changing, or maybe both. Hopefully I
will
> have time to try this tomorrow.

> > The speed and braking you use during the creation of the AI's lines has
> > little effect on their speeds or braking. The smoothness of the lines
is
> > how the AI calculate their speeds, not the actual speeds values recorded
> in
> > the lp's. Your best bet is to try to go into your track.ini and look in
> the
> > ai_track section and adjust this value.....

> > ai_dlongpad_scale = X.X ; scales padding in braking zones (smaller =
> more
> > scaling)

> > Jay Taylor

rec.autos.simulators

NR 2003 Short Track / Roadie AI

NR 2003 Short Track / Roadie AI

NR 2003 Short Track / Roadie AI

NR 2003 Short Track / Roadie AI

NR 2003 Short Track / Roadie AI

NR 2003 Short Track / Roadie AI

NR 2003 Short Track / Roadie AI

NR 2003 Short Track / Roadie AI

NR 2003 Short Track / Roadie AI

NR 2003 Short Track / Roadie AI

NR 2003 Short Track / Roadie AI

NR 2003 Short Track / Roadie AI

NR 2003 Short Track / Roadie AI

NR 2003 Short Track / Roadie AI