rec.autos.simulators

GPL: AI behavior

Michael E. Carve

GPL: AI behavior

by Michael E. Carve » Wed, 14 Oct 1998 04:00:00

Kudos to Papyrus for making great strides in improving the AI behavior
in GPL.  However, there still seems to be a few issues that need
addressing.

The first one I will have to reserve judgement until my own driving
style improves.  However, there seems to be a huge tendency for the AI
to run up over the human car under braking.  It maybe my own fault, as I
tend to brake a little early and harder than the AI.  But on too many
occasions I have been run over by AI who insist that I am not there.
Instead of moving inside or outside they insist on following their own
braking patterns no matter what.

The second one is very obvious and frustrating.  On a number of
occasions the AI will insist on moving over onto my car even though I am
1/2 -2/3's in front of them.  They bring their front right wheel in
between my left wheels (usually just in front of the rear wheel) and
continue to move right on over to the right.  While this leads to
spectacular wrecks and highlight reels, it is not correct.  I have 3
replays of this behavior (2 on straights and 1 in the Parabolica).  In
all occasions I am holding my line and have clearly gotten past the AI
and the AI does not NEED to move over at the point that it does.  If
they wait a split second (or let off just a hair) they could easily drop
behind me in plenty of time to continue their move safely.  On one
occasion it was a back-marker and I would have clearly been able to
finish the pass before it needed (or wanted) to move back to the normal
racing line.  I offer them the same courtesy and only hope this will be
addressed in the next drivers' meeting.

The third one is still a little too early to judge, but it seems that
there is still the tendency of the AI to be "real fast" while in front
of the human and too laid back while far behind the human.  While it
definately isn't as pronounced as I have seen in early Papyrus efforts,
it still seems to be an issue in GPL.  If I can drive consistently
(which on a rare occasions I do manage), I can hold a comfortable 1 - 2
second lead over an AI car, but once I muss a corner and the AI can get
around me, it seems that it can always increase its lead even when I am
back in my consistent, smooth driving zone.  

Were any of these issues dealt with during the beta trials?  Has
Papyrus created enough modifiable variables in the ai.ini files to help
correct some of these problems?

--
**************************** Michael E. Carver *************************
     Upside out, or inside down...False alarm the only game in town.

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=<[ /./.  [-  < ]>=-=-=-=-=-=-=-=-=-=-=-=-=-=

Berry Kaise

GPL: AI behavior

by Berry Kaise » Wed, 14 Oct 1998 04:00:00


>Kudos to Papyrus for making great strides in improving the AI behavior
>in GPL.  However, there still seems to be a few issues that need
>addressing.

>The first one I will have to reserve judgement until my own driving
>style improves.  However, there seems to be a huge tendency for the AI
>to run up over the human car under braking.  It maybe my own fault, as I
>tend to brake a little early and harder than the AI.  But on too many
>occasions I have been run over by AI who insist that I am not there.
>Instead of moving inside or outside they insist on following their own
>braking patterns no matter what.

I've had this happen as well. Braking into the Parabolica is frustrating.
Many times I've had the AI run up by ***and send us both into the gravel
trap. I've even braked late and still had it happen. However, maybe I'm
delusional but I could swear that some AI are much more agressive than
others. It seems that I'm getting rammed by certain drivers..sorry don't
know all there names yet :))... and not others.

Haven't run across this one yet :)

>The third one is still a little too early to judge, but it seems that
>there is still the tendency of the AI to be "real fast" while in front
>of the human and too laid back while far behind the human.  While it
>definately isn't as pronounced as I have seen in early Papyrus efforts,
>it still seems to be an issue in GPL.  If I can drive consistently
>(which on a rare occasions I do manage), I can hold a comfortable 1 - 2
>second lead over an AI car, but once I muss a corner and the AI can get
>around me, it seems that it can always increase its lead even when I am
>back in my consistent, smooth driving zone.

>Were any of these issues dealt with during the beta trials?  Has
>Papyrus created enough modifiable variables in the ai.ini files to help
>correct some of these problems?

>--
>**************************** Michael E. Carver *************************
>     Upside out, or inside down...False alarm the only game in town.

>=-=-=-=-=-=-=-=-=-=-=-=-=-=-=<[ /./.  [-  < ]>=-=-=-=-=-=-=-=-=-=-=-=-=-=

Berry Kaiser
Jim Bonewa

GPL: AI behavior

by Jim Bonewa » Wed, 14 Oct 1998 04:00:00

On Tue, 13 Oct 1998 16:39:21 GMT, "Michael E. Carver"


>Kudos to Papyrus for making great strides in improving the AI behavior
>in GPL.  However, there still seems to be a few issues that need
>addressing.

>The first one I will have to reserve judgement until my own driving
>style improves.  However, there seems to be a huge tendency for the AI
>to run up over the human car under braking.  It maybe my own fault, as I
>tend to brake a little early and harder than the AI.  But on too many
>occasions I have been run over by AI who insist that I am not there.
>Instead of moving inside or outside they insist on following their own
>braking patterns no matter what.

I have also had this happen to me many times at Monza. When running in
the middle of a group of cars, I have to brake to keep from hitting
the rear of the AI in front and my car gets hit from the rear and
totally ruins a good race.This problem needs to be fixed so that the
AI will either follow your car through the turn or else take a
different line.

>--
>**************************** Michael E. Carver *************************
>     Upside out, or inside down...False alarm the only game in town.

>=-=-=-=-=-=-=-=-=-=-=-=-=-=-=<[ /./.  [-  < ]>=-=-=-=-=-=-=-=-=-=-=-=-=-=

Rick Baumhaue

GPL: AI behavior

by Rick Baumhaue » Wed, 14 Oct 1998 04:00:00

I ran into an interesting AI glitch yesterday.  I was running a Pro-Short
race at Mosport, just to see how the AI cars race against each other, so I
pulled off and retired at the start.

The AI developed a MAJOR problem with the tighter part of the two-corner
hairpin at the far end of the course, and no less than 75% of the field was
eliminated in 3 separate multi-car pileups coming off the turn.  Each
started with one car getting his left front wheel in between the right side
wheels of another car, which led to partial spins and two stationary cars.

What made the incidents much worse was the behavior of the following
drivers, who simply didn't make enough of an adjustment to the partially
blocked track in front of them.  Once there were 4-5 cars involved, the
other cars following looked like they were trying to RAM their way through
the obstruction.  The only thing that kept the incidents from claiming the
whole field was the eventual disappearance of the "retired" cars.

Has anybody else seen this?  Seriously, there were FOUR CARS left at the end
of the race!  Chris Irwin got a podium, behind Clark and Rodriguez!

Rick

Vitzthu

GPL: AI behavior

by Vitzthu » Thu, 15 Oct 1998 04:00:00

Yup - seen this a fair amount at Mosport.  Corner 5b is very sharp and
slow; it seems that this is a great place for passing as well, and the
AI know this.  When holding a normal line the AI tend to take this
corner *very* wide (much more than I ever do) but will tuck under for a
pass.

So, first prang is when one car tries to tuck up under the first and
causes both to spin.  They take up almost all the track... add a couple
of more cars to a blocked path and soon there is a massive traffic jam.
It seems the AI always want to keep on moving, so they will slowly go to
space to try and get past - only... there is no space!

Open wheeled cars love to collect each other, and at the slow speed 5b
of Mosport it can take an eternity for this to clear up.  Then add my
driving into the equation: I come screaming down the gulley (ignore a
yellow flag - it's just an AI up on the bank right? :), wheels squeeling
under braking, up the uphill with a huge dose of right wheel lock and
then... surprise!  Major pile up!  Boom!

Sigh... should have paid attention to the flag!

Cheers,
Gian Vizthum.


> I ran into an interesting AI glitch yesterday.  I was running a Pro-Short
> race at Mosport, just to see how the AI cars race against each other, so I
> pulled off and retired at the start.

> The AI developed a MAJOR problem with the tighter part of the two-corner
> hairpin at the far end of the course, and no less than 75% of the field was
> eliminated in 3 separate multi-car pileups coming off the turn.  Each
> started with one car getting his left front wheel in between the right side
> wheels of another car, which led to partial spins and two stationary cars.

> What made the incidents much worse was the behavior of the following
> drivers, who simply didn't make enough of an adjustment to the partially
> blocked track in front of them.  Once there were 4-5 cars involved, the
> other cars following looked like they were trying to RAM their way through
> the obstruction.  The only thing that kept the incidents from claiming the
> whole field was the eventual disappearance of the "retired" cars.

> Has anybody else seen this?  Seriously, there were FOUR CARS left at the end
> of the race!  Chris Irwin got a podium, behind Clark and Rodriguez!

> Rick


rec.autos.simulators is a usenet newsgroup formed in December, 1993. As this group was always unmoderated there may be some spam or off topic articles included. Some links do point back to racesimcentral.net as we could not validate the original address. Please report any pages that you believe warrant deletion from this archive (include the link in your email). RaceSimCentral.net is in no way responsible and does not endorse any of the content herein.