NTAI

AF · Post by AF » 12 Oct 2005, 21:06

The only code that would be afefcted by he map size thats in the AI would be cains metal extractor placement. So I am inclined to agree with you nwo you've said that about maps.

Perhaps zaphod should give the code the ocne over, afterrall zaphod has said in PM's to me it sues the same method as JCAI but JCAI does it better. I'm sure cain would like the extra time from pointing out any errors.

Post by **jcnossen** » 12 Oct 2005, 21:13

My code, including the metal map class is on spring CVS. I don't really know if it does it better, but I think it's bugfree. There is some disabled and unfinished caching stuff in the code though, because the metal map processing takes very little time (So it wasn't really necessary to cache on disk)

AF · Post by AF » 12 Oct 2005, 21:14

Isnt your code bundled with springs code in the source code downlaod?

Post by **jcnossen** » 12 Oct 2005, 21:15

AF · Post by AF » 14 Oct 2005, 21:31

A new version for 0.65B2 with file logging and a few tweaks, aswell as a switch to using the code alik has for mex finding integrated into cains class.

The links at the top =D

note: I havent been able to fix that error, but it should last a lot longer before ti happens now, and having the logging, I need as many examples of testAI.log to verify what i told triaxx in the other thread.

Post by **jcnossen** » 14 Oct 2005, 23:15

Seeing that you continue to work on ntai.. where is the source?

AF · Post by AF » 15 Oct 2005, 17:33

It's in a place somewhere between my computer and sourceforge. But I wasnt going to release untill i had fixed the bug, which I believe I have, I'm just playtesting again to verify it.

AF · Post by AF » 15 Oct 2005, 18:09

I fixed the elusive bug and now it works fine. It has my new html logging clas itnegrated and cains latest mex algorithm.

Download here

And as Zaphod noted, I'll upload the source shortly.

HiEnergy · Post by **HiEnergy** » 16 Oct 2005, 00:34

For me the NTAI (NTAI V0.2B2) does not do anything. Only says "Build location collision found" then the commander just sits there.
I put the contents of NTAIV0.2B2.rar (a single dll named ntai.dll) in the folder AIDLL\globalai. Then I tried to play against NTAI on Flooded Desert. What's wrong with that?

AF · Post by AF » 16 Oct 2005, 17:07

Well, I think you may have a problem that's mroe of a result fo the map you're playing on. The logic of the AI says that shouldnthappen. If ti still doesnt do ought, select commander and press stop and it should move onto its next task. Otherwise try a different map. That shouldnt really happen but if I removed the check and allowed it to build it could result in instability.

Other than that amsse had nice results doing JCAI vs NTAI playtests on brazillian battlefields with that version.

Masse · Post by **Masse** » 16 Oct 2005, 17:14

Other than that amsse had nice results doing JCAI vs NTAI playtests on brazillian battlefields with that version.

its Masse

yeah NTAI "wins" it just controls forces stupidly and the metal mex algorithm still sucks

but "beats" JCAI

AF · Post by AF » 16 Oct 2005, 17:15

Suffice to say the insanely rubbish target routine has been replaced this morning, though not released.

Triaxx2 · Post by **Triaxx2** » 16 Oct 2005, 17:34

That's good to hear. I assume the new one is more streamlined, and faster?

AF · Post by AF » 16 Oct 2005, 17:48

I suggest you look at how JCAI does it =D, NTAI uses a similair approach now, exept I've added a few more parameters that to some may seem obvious. Such as sending scouts to the mex spots given from cains class that are nearest to the enemy position.

GrOuNd_ZeRo · Post by **GrOuNd_ZeRo** » 16 Oct 2005, 17:52

Will NTAI ever support mods and if so how?

AF · Post by AF » 16 Oct 2005, 18:16

If you gave me a list of what you wanted the commander to build, then what each factory and conbot it builds should build, and the unit health of each attack unit or builder then I'll add it in. Otherwise I'll cotninue to move NTAI to dynamic building so eventually NTAI will decide for itself what to build, without any config files at all. TAI had this but it hadnt been finished to a compilable state before the main file cotnianing it was corrupted.

Stanky105 · Post by **Stanky105** » 18 Oct 2005, 22:58

I also get the "build location collision found," with the commander sitting there not doing anything. I also have version .2B2

It does this on every single map I have, including Brazilian Battlefield.

submarine · Post by **submarine** » 19 Oct 2005, 00:06

i tried it out, on some maps it crashes right at the beginning (hill heaven for example) on other maps (hills and plains) it builds metal makers like hell but nearly no other buildings (especially it doesn├é┬┤t build more tahn one or two mexes)

AF · Post by AF » 19 Oct 2005, 20:15

And yet nobody has posted a copy of AIlog.htm

If the commander isnt building anything, take cotnrol fot eh AI and press stop, ti should continue to the next build task, otherwise keep pressing.

Cain has tested and shown that the interface function NTAI uses suddenly fails to work as well in multiplayer situations, try running it manually through spring.exe or just by yourself....

And if you think it builds metal makers like hell you should see what ti used to build. Perhaps the construction collision errors are due to cains metal class which ahs a habit of returning 0,0,0 every now and again, for some people ti always returns 0,0,0 hence the collision message.

At least tell me what the last 3 lines in AILog.htm are, I need to know how far the code got so i can fidn any bugs. Or even if the log file exists!

CaptainMaim · Post by **CaptainMaim** » 19 Oct 2005, 22:22

I've got one, it's 11 megs... The AI went pretty far for me. Inspite of everything... I noticed though that my computer got a lot more sluggish after I came back and unpaused it from eating dinner. And eventually the game crashed nicely with a little AI dll error mentioning some unhandled exception or something.

The only extra unit I was using was the stargate... I can post the AILOG.html for you, or the replay. I was doing a JCAI + NTAI vs me game.. And I was going easy on them (cause I wanted to have some fun with them.) I wanna know why both the AI's congregated in the player 2 starting spot? NTAI TOTALLY overgrew the JCAI base (JCAI was like a little splash in the middle of an NTAI pool.) I also noticed that after the NTAI's Can rush (which I was a little surprised to see.) After the cans had forced their way on what I can assume was a direct line between start points 1 and 2, it reached startpoint 1 and stopped. The reenforcements dribbled in and NTAI never attacked me again. Maybe next time I'll just setup a happy little killing zone since the AI only mangaged to get 3 units at startpoint 1 and then never moved them again.
<Imagines devious forms of self amusement centered around AI's flaws>

Here's the first 3 lines:

HamburgerHill.smf
xta_se_v065.sdz
map size: 80 x 80

And here's the last several lines from the log file:
[84:27] Chaser::UnitDestroyed
[84:27] Chaser::UnitDamaged
[84:27] factor::UnitIdle() new task
[84:27] CBuild()
[84:27] CORMEX
[84:27] CBuild()3
[84:27] CBuild()3m
[84:27] Build location collision found!
[84:27] Chaser::UnitDamaged
[84:28] Chaser::UnitDamaged
[84:28] Chaser::UnitDamaged
[84:28] Chaser::UnitDamaged
[84:28] Chaser::UnitDamaged
[84:28] Chaser::UnitDamaged
[84:28] Chaser::UnitDamaged
[84:28] Chaser::UnitDamaged
[84:28] factor::UnitIdle() new task
[84:28] tasknum at one?
[84:28] CBuild()
[84:28] CORARAD
[84:28] CBuild()3

I hope that all this helps.

Spring RTS Engine

NTAI