Page 5 of 5
Posted: 08 Dec 2005, 21:24
by jcnossen
I don't really get it either, you make your first proper C++ class and immediatly think you have the optimal solution and there is no way to improve the algorithm? Like torrasque says, there are always other ways to do it. Don't you think people with more experience would know this better?
On first sight I can think of doing a early out test for sectors that have low metal in them... there you have it, a faster algorithm.
Probably someone else can think of more stuff...
Posted: 08 Dec 2005, 21:55
by krogothe
Sorry guys i had a bad day and took it out on you...
Ill stop the hijack here!
Posted: 08 Dec 2005, 22:31
by AF
tis alright, I'm havign a bad week myself
Posted: 08 Dec 2005, 22:48
by Masse
this was bad day for me too... this must be some global bad day for dudes...
im serious here
Posted: 08 Dec 2005, 23:16
by submarine
lol me too
Posted: 26 Dec 2005, 14:06
by athemis
I've experienced a weird attack behaviour when doing a quick AAI vs NTAI game on SmallDivide:
NTAI starting in the bottom right corner built up its base just fine and then sent several attack waves toward the AAI base. But actually NTAI sent its units just to the top left corner of the map, doing no harm to AAI, because there were no buildings on their way. The units remained there for the rest of the game and every new "attack" wave was exactly the same: they all ended in the top left corner.
It's the same on other maps, River Dale for example.
Posted: 26 Dec 2005, 15:20
by AF
athemis wrote:I've experienced a weird attack behaviour when doing a quick AAI vs NTAI game on SmallDivide:
NTAI starting in the bottom right corner built up its base just fine and then sent several attack waves toward the AAI base. But actually NTAI sent its units just to the top left corner of the map, doing no harm to AAI, because there were no buildings on their way. The units remained there for the rest of the game and every new "attack" wave was exactly the same: they all ended in the top left corner.
It's the same on other maps, River Dale for example.
SMALL DIVIDE?!?!?!?!? NATI WORKS ON SMALL DIVIDE?!?!? AND RIVERDALE?!?!?!?
Forgive em for the reaction but not once in NTAI's history ahve I ever managed to play a game without it instantly crashign as soon as it tries to find a aplce to build something.
Otherwise the attack behaviour you describe was eradicated after 0.28b4, I suggets you upgrade. The latest versions of NTAI have a check so if they do somehow want to attack the 0,0,0 co-ordinates, the co-ordinates are discarded and the whole process is dropped.
Posted: 27 Dec 2005, 03:29
by Targon
well actually with the current old (newest) release with me on Mountain Siege NATAI vs AAI it sent an attack wave at 0,0,0 once. Never did that again though in any other game I observed.
Posted: 27 Dec 2005, 17:30
by Andreask
I tried NTAI 28.10 on CometCatcher with AA 1.3-S, the first game i played vs a core Opponent, the secod game i pitted 2 Core AIs against eachother.
During the first game, the AI sent units to 0,0,0, but not during the second game.
During the first game, the AI never attacked me but only reacted to my attacks in such a manner that it slaughtered its bots on my defences.
I just overran its LLT defence with Hammers around the 10th minute, but when i reached its commander, the game crashed.
During the second game both AIs expanded, goung k-bots.
Neither of them attacked, although they both had a dozen kbots a few minutes in.
Strangely, both AIs built mexes on spotrs where there was no metal, like right next to another mex.
When the con-k-bot from the black ai wanted to eventually build a mex in yellows territory, it got destroyed by a LLT.
After that, both AIs sent all the bots they had to that location, mutualy annihalating them all on one LLT each, scince the bots didnt fire, or fired only at point blank.
A few seconds after the attacks the game crashed once again.
What disturbed me the most was, that during the whole game in both games, i could see colored streaks on the minimap and in the over-head-view, indicating ai-unit positions and behavior (such as yelllow streaks for con-units, white for buildings and green for attackers) although i had no radar up.
Posted: 27 Dec 2005, 18:08
by AF
The streaks are gone in the next version.
You tried it with XTA at all? Since nobody, not one person in this entire community will help me adjust NTAI for AA save the tweak caydr made with the techforge, it seems NTAI pretends that AA is XTA and thus makes mistakes aswell as some things simply not working.
Untill a new version si released the change caydr made in changing the anem of the techforge would mean NTAI never built it anyway, I dont even know what the techforge builds or what its unitname is, as I dont get the chance ot play AA, or online at all these days.
Mex class has been replaced with krogothes class, you should get similair accuracy to KAI and AAI now if not better.
Atatck and scouting code is being totally rewritten too.
The crash would be an engine bug where the engine crashes durign D-Gunning or similair events. D-Gunning is disabled in the next version too unless I fix it.
In the next version atatck units will attack a target if ti is within double its maximum weapons range, I'm going to add more simple behaviours but would this be a good idea to keep or would ti be better if it's taken away? I've noticed when untis attack they now focus their firepwoer on targets and attack much better but for lone units I'm not so sure.
Posted: 28 Dec 2005, 00:52
by Triaxx2
Send me along a copy of the file you need to make it work, and I'll fix it for you. I've got the time to do it.
Posted: 28 Dec 2005, 15:15
by AF
That wont be as simple as it sounds triaxx.
The cosntruction code in the next NTAI version ahs been compeltely erased and rebuilt from the ground up to get rid of a miriade fo errors that afflicted NTAI and other AI's (namely the virtual memory si low crash).
The whole thing is now totally mod independant and works fine now.
Posted: 28 Dec 2005, 23:49
by Triaxx2
Then why does it need tweaking? Shouldn't it tweak itself?
Posted: 29 Dec 2005, 12:48
by AF
Because the code executes and then at the very end it crashes/causes an exception message depending on whose computer ti runs on
Code: Select all
Tunit::UnitIdle(){
/*... con code here....*/
G->L->print("message 1");
}
followed by
Code: Select all
Global::UnitIdle(int unit){
try{
if(units.empty() == false){
for(list<Tunit>::iterator tl = units.begin(); tl != untis.end(); ++tl){
if(tl->uid == unit){
if(tl->bad == false){
tl->UnitIdle();
L->Print("marker 2");
break;
}
}
}
}
}catch{
L->print("exception in unit::idle");
}
}
The result in the logfile is:
Code: Select all
[00:xx] marker 1
[00:xx] exception in unit::idle
which baffles me completely.
Posted: 02 Jan 2006, 18:03
by cain
on my old class:
the hotspots wasn't spot with metal on it, they where place where the medium metal distribution was higher than the norm, not places where to build but places to rush and defend as they where valuable.
my class was a brute force solution: finding all the metal an extractor will extract on a particular map spot for every point of the map, sorting the spot, and returning the nearest spot in a radius with the high value. The removal of near extractors to avoid superposition was done after the search, not during the search.