Skirmish AI: E323AI 3.22.4 - Page 29

Skirmish AI: E323AI 3.22.4

Here is where ideas can be collected for the skirmish AI in development

Moderators: hoijui, Moderators, hoijui, Moderators

User avatar
Error323
AI Developer
Posts: 237
Joined: 28 Nov 2006, 16:46

Re: Skirmish AI: E323AI 3.19.0

Post by Error323 »

An assert is still raised within CGroup::reset(), just didn't show up during testing, bah. Back to the drawing boards.
0 x

User avatar
FaerieWithBoots
Posts: 149
Joined: 17 Jun 2009, 13:21

Re: Skirmish AI: E323AI 3.19.0

Post by FaerieWithBoots »

but is it more stable then the previous versien, or in other words, should i dl this version?
0 x

User avatar
Error323
AI Developer
Posts: 237
Joined: 28 Nov 2006, 16:46

Re: Skirmish AI: E323AI 3.19.0

Post by Error323 »

It is definitely not less stable... I can't say much more :/ working on it.
0 x

User avatar
Error323
AI Developer
Posts: 237
Joined: 28 Nov 2006, 16:46

Re: Skirmish AI: E323AI 3.19.1

Post by Error323 »

E323AI v3.19.1 - High Templar
Changelog:
  • Fixed 2 crashes
E323AI SO/DLL: HighTemplar
E323AI BUGS: Issues
E323AI SRC: Github

- Error323 & Slogic
0 x

User avatar
KaiserJ
Community Representative
Posts: 3113
Joined: 08 Sep 2008, 22:59

Re: Skirmish AI: E323AI 3.19.1

Post by KaiserJ »

thx for the continued support of this AI

i've been working on the customizable part of shard and pitting it against various e323AI opponents... had a decent config where shard smashed e323 with raiders around then 10 minute mark, but we're still early days yet

anyways cool :) thx for the new opponent... i really love how this AI works, the closest to a "real player" as i've seen IMO
0 x

User avatar
Jazcash
Posts: 5302
Joined: 08 Dec 2007, 17:39

Re: Skirmish AI: E323AI 3.19.1

Post by Jazcash »

KaiserJ wrote:works, the closest to a "real player" as i've seen IMO
Yeh, a real rank 3 player :D But seriously, nice AI ^_^
0 x

User avatar
BrainDamage
Lobby Developer
Posts: 1164
Joined: 25 Sep 2006, 13:56

Re: Skirmish AI: E323AI 3.19.1

Post by BrainDamage »

0 x

SiriusDragon
Posts: 14
Joined: 21 Mar 2010, 07:13

Re: Skirmish AI: E323AI 3.19.1

Post by SiriusDragon »

Odd bug to report... I was attempting to spectate 2 E323AI players on FinnsRevengeRV4.smf, the game was XTA 9.6

While one player built up their base fine, the other AI did nothing. I checked the logs and found heaps of lines like this for one of the ais, presumably the one that wasn't working:

[00:00] (EE): CPathfinder::getClosestNode failed to lock node(45,15) for pos(5760,1920)

Up until those errors the logs are basically identical.

Any thoughts on why this would be happening?

Nice work on the AI btw :) It killed a RAI I set it against pretty dang quickly...

Edit: Addition to my bug report. I spotted this line in both ai logs too:

[00:00] (II): Number of enemies: 2

Howcome it thinks there are 2 enemies? I suppose it must be counting spectators as enemies instead of just ignoring them.
0 x

slogic
AI Developer
Posts: 626
Joined: 17 Mar 2008, 19:03

Re: Skirmish AI: E323AI 3.19.1

Post by slogic »

I bet this is because that map has metal spots on islands. Current AI state has an ugly behavior when there are unreachable land resources or any other unreachable entities which AI treats as targets.
0 x

SiriusDragon
Posts: 14
Joined: 21 Mar 2010, 07:13

Re: Skirmish AI: E323AI 3.19.1

Post by SiriusDragon »

That must be it then :-\ It's pretty much an exact copy of SupCom's Finns, so there's 2 metal patches on 2 islands either side of the main land mass.
0 x

slogic
AI Developer
Posts: 626
Joined: 17 Mar 2008, 19:03

Re: Skirmish AI: E323AI 3.19.1

Post by slogic »

SiriusDragon wrote:[00:00] (II): Number of enemies: 2
In fact this is number of enemy units, not number of enemy players.
0 x

User avatar
Error323
AI Developer
Posts: 237
Joined: 28 Nov 2006, 16:46

Re: Skirmish AI: E323AI 3.19.1

Post by Error323 »

Hey everyone,

I'm currently working on a buildmap that allows the AI to plan where to place buildings in order to e.g. keep the entrance of factories clear.

As I'm working on that I thought about how it would be nice to have realtime debugging info on maps. So with help from Kloot (he implemented the interfaces within spring :D), here are some results of realtime threatmap rendering as a texture and a realtime profiler.

Its currently only available in the master of both E323AI and spring itself as they are new features. You can look at it yourself by fetching both latest E323AI and spring and when ingame execute: /debugdrawai while spectating E323AI.

Just thought I'd let you guys know!

-- Error323
Attachments
realtimedebug.png
Realtime updates on threatmaps and profiler.
(2.73 MiB) Downloaded 2 times
0 x

CerealKiller159
Posts: 6
Joined: 05 Feb 2010, 21:53

Re: Skirmish AI: E323AI 3.19.1

Post by CerealKiller159 »

Very cool, air support will certainly help in med-late game.

I have a question, can the build map be used to reserve space for buildings to be built later (like nanos by factories)?

Also, recently, I have been noticing that sometimes the AI hardly builds any t2 stuff. Sometimes it will only have one advanced builder and all its doing is helping other builders. (I guess because it maxes out on t1 builders?)

So, put simply, a system where the ai reclaims t1 economic units and buildings then replaces them with high tech ones would prevent swarms of low tech builders, and prevent super huge bases of windmills and mmakers. (if it wants to build a fusion, it picks a spot away from t2 buildings, then reclaims all t1 buildings in the way before building)

And my last comment: I noticed that KAI can have way more builders around factories and still not clog the entrance. It seems to keep the builders as far away from the factory as possible by moving them back if they get pushed too close. Also, not assembling groups in front of the factory helps too.
0 x

SirMaverick
Posts: 834
Joined: 19 May 2009, 21:10

Re: Skirmish AI: E323AI 3.19.1

Post by SirMaverick »

[00:02] (EE): CPathfinder::getClosestNode failed to lock node(41,5) for pos(5184,608)
[00:02] (EE): CPathfinder::getClosestNode failed to lock node(40,4) for pos(5064,479)
[00:02] (EE): CPathfinder::getClosestNode failed to lock node(40,4) for pos(5064,479)
[00:03] (EE): CPathfinder::getClosestNode failed to lock node(41,5) for pos(5184,608)
[00:03] (EE): CPathfinder::getClosestNode failed to lock node(40,4) for pos(5064,479)
[00:03] (EE): CPathfinder::getClosestNode failed to lock node(40,4) for pos(5064,479)
[00:03] (EE): CPathfinder::getClosestNode failed to lock node(41,5) for pos(5184,608)
Sometimes the com get's stuck, doesn't do anything. Happens rarely.
Attachments
team-0.log
(216.8 KiB) Downloaded 14 times
0 x

User avatar
Zydox
Lobby Developer
Posts: 453
Joined: 23 May 2006, 13:54

Re: Skirmish AI: E323AI 3.19.1

Post by Zydox »

Isn't it time for a new version with air support soon? :)
0 x

User avatar
MisterBenn
Posts: 21
Joined: 19 Apr 2010, 15:50

Re: Skirmish AI: E323AI 3.19.1

Post by MisterBenn »

Hello there, just to let you know I made some config files to make E323AI 3.19.1 run with Supreme Annihilation U41. I made this by:

* Taking the defaults generated when the mod runs the first time
* Where the BA config file provides an entry for a SA unit or building, use the tags that BA has.
* Removing the BUILDER tags from minelayers (they were being stockpiled in the early stages like Constructors but were remaining idle or sent to battle!).
* Removed SCOUTER and ATTACKER from the Flea. Arm were obsessed with Fleas for scouting, but since E323AI does both scouting and harassment in the early game, Arm steadily loses ground to Core in a 1 on 1 as the fleas lose all the mini combats. Now in the early minutes, Peewees are mostly used.

Set up like this, E323AI bots seem to play SA well, certainly the early game has no major problems. I attached the files in case anyone cares to use them, I tested them a few times in 1 on 1 games.

I like that in 3.19.1 the bases are more defended now! That's a great improvement. The assisting / repairing / reclaiming of workers works really well too.

One thing I noticed on small maps though is that the bots struggle to keep mexes alive outside the bases. On 8x8 maps both Arm and Core end up with just the metal income from their immediate bases (everything else gets sniped, including wandering constructors), and as a result they struggle to push their metal income beyond 10 and thus don't reach T2 easily. I wonder if either they should consider putting up defences or guards near some resource sites...

Edit: Config files were updated, please see next page.
Last edited by MisterBenn on 23 Apr 2010, 05:29, edited 1 time in total.
0 x

SirMaverick
Posts: 834
Joined: 19 May 2009, 21:10

Re: Skirmish AI: E323AI 3.19.1

Post by SirMaverick »

MisterBenn wrote:I like that in 3.19.1 the bases are more defended now! That's a great improvement. The assisting / repairing / reclaiming of workers works really well too.
They could reclaim a bit more. I often see AIs stalling with 1k or more metal in or around their base.
One thing I noticed on small maps though is that the bots struggle to keep mexes alive outside the bases. On 8x8 maps both Arm and Core end up with just the metal income from their immediate bases (everything else gets sniped, including wandering constructors), and as a result they struggle to push their metal income beyond 10 and thus don't reach T2 easily. I wonder if either they should consider putting up defences or guards near some resource sites...
Not just defense. The constructors itself need protection. They often get killed when trying to build mex or defense.
0 x

User avatar
MisterBenn
Posts: 21
Joined: 19 Apr 2010, 15:50

Re: Skirmish AI: E323AI 3.19.1

Post by MisterBenn »

You're right about the wreackage sometimes lying around for a while, I suspect that's down to not interrupting orders while they're in progress. Overall I reckon they do a good job though.

That reminded me of something else that occurred to me... when many constructors are assisting a factory which is stalling metal, could a constructor or two repair hurt units nearby instead? That only costs energy and crops up a lot due to all the raids.
0 x

User avatar
hoijui
Former Engine Dev
Posts: 4342
Joined: 22 Sep 2007, 09:51

Re: Skirmish AI: E323AI 3.19.1

Post by hoijui »

nice job MistBenn!
i have not tried it yet, but sure is a good thing to have this AI work with SA too :-)
thanks, in behalf of the users!
0 x

bbe
Posts: 10
Joined: 15 Nov 2008, 17:24

Re: Skirmish AI: E323AI 3.19.1

Post by bbe »

I've been trying 3.19.1 (with spring 0.81.2.1) on BA 7.12 and am having some mixed results. On Parched Mesa a lot of units seem to get stuck around the base with loads of adv. solar. All of a sudden I get tremendous lag (about 15minutes in). Turns out after the game I've got a 300MB log file full of pathing errors (I'm on another machine atm, so can't post a snippet). My online game seems to start desyncing when I experience this lag. Could these be related? Can a lagging machine cause a desync? I'm a linux (Gentoo) user and have manually compiled spring to get the latest 3.19.1.

I'm going to try a few more things to pin this down. One would be to use the latest master of E323 and the other is to disable the logging to see if it stops desyncing.

I'll keep you posted.
0 x

Post Reply

Return to “AI”

cron