Strange AI behavior, not working.

[leo.civil.uefs]

Mak... would be humanly possible to understand all these parameter that can be acesses by IWTE in ndescr_geography_new.db? All these numbers, values...?

just did some tests with esgaroth.

added a small reflective plane and then the water generated by the campaign map was gone. No difference at AI behavior at all, though it could be something with the water running under the ground where the troops moves.

As I said before no climate makes it work, its not the climate, it is something else.

[makanyane]

Mak... would be humanly possible to understand all these parameter that can be acesses by IWTE in ndescr_geography_new.db? All these numbers, values...?

It's pretty much the same as the geography.db from RTW - so Muizer's guide is still the best information
http://www.twcenter.net/forums/showt...p-modification

you might need to look at the .txt file version for reference if the abreviations in IWTE are different.

btw, there's a setting called 'water_height_modifier' in the .txt file - it's not included in IWTE as we tested it and changing the value in the .txt file doesn't result in any changes in the .db file when it regenerates (e.g. it's not actually included in the .db)

just did some tests with esgaroth.

added a small reflective plane and then the water generated by the campaign map was gone. No difference at AI behavior at all, though it could be something with the water running under the ground where the troops moves.

As I said before no climate makes it work, its not the climate, it is something else.

did you try what I did in my post here
http://www.twcenter.net/forums/showt...1#post13556273
which was basically either raise the map heights a lot or do a vector change of +60 on both the .world and the .worldterrain - that made the settlement height a lot higher than the water level and the ai worked... (interesting but not very useful as it looks silly)

added a small reflective plane and then the water generated by the campaign map was gone

I'm not sure if that actually makes the 'water' disappear (like with troops able to walk there), or if it just can't display both types of graphics on same map.

[leo.civil.uefs]

did you try what I did in my post here
http://www.twcenter.net/forums/showt...1#post13556273
which was basically either raise the map heights a lot or do a vector change of +60 on both the .world and the .worldterrain - that made the settlement height a lot higher than the water level and the ai worked... (interesting but not very useful as it looks silly)

ok. Tried several heighs, 60, 80... 500! 60 was enough to make things work and beyond 60 no real change was noticed.

in custo battles, they arrange the army before entering the settlement in a single line giving me the signal that all would work, and in fact, it does. The same does not happen in the campaign map, but still the AI works well at all, reaching the plaza with all units. I got an dwarven amy with 966 men attacking me, would be better if I could get a full stack for a proper test but you know getting some like this in a campaign battle is really difficult.

Anyway, it seems that it is really the heigh causing the problem.

btw, there's a setting called 'water_height_modifier' in the .txt file - it's not included in IWTE as we tested it and changing the value in the .txt file doesn't result in any changes in the .db file when it regenerates (e.g. it's not actually included in the .db)

Thats interesting... any more info about this?

[makanyane]

ok. Tried several heighs, 60, 80... 500! 60 was enough to make things work and beyond 60 no real change was noticed.

in custo battles, they arrange the army before entering the settlement in a single line giving me the signal that all would work, and in fact, it does. The same does not happen in the campaign map, but still the AI works well at all, reaching the plaza with all units. I got an dwarven amy with 966 men attacking me, would be better if I could get a full stack for a proper test but you know getting some like this in a campaign battle is really difficult.

Anyway, it seems that it is really the heigh causing the problem.

did have one other possibly useful thought - I think it must be height versus the fact that there is water there, and not just height (?) so you could try raising the height of the .world/terrain slightly like about 20, but also raise the height of the bottom of the 'lake' area of the terrain, so that water does not get dragged in from the river. If that works without water at all you could then add back in a reflective plane.

I did start having a go on a without water version, but got into other problems, that without real water you have to block the area around the settlement in pathfinding, and I ended up with some ai units doing that thing where they attempt to circle the settlement and get stalled. (the settlement plan does break the rules you came up with about the ai wanting to be able to circle the plaza from different directions!)

Thats interesting... any more info about this?

It doesn't get included in the .db - that means it's not used by the game, which means it's interesting in terms of "I wonder what CA were going to do with that" - but not interesting as in actually any use...

[leo.civil.uefs]

did have one other possibly useful thought - I think it must be height versus the fact that there is water there, and not just height (?) so you could try raising the height of the .world/terrain slightly like about 20, but also raise the height of the bottom of the 'lake' area of the terrain, so that water does not get dragged in from the river. If that works without water at all you could then add back in a reflective plane.

I tried this a looooooooooooong time ago, Ive raised the terrain under the lake until some centimeters under the water, nothing changed on AI behavior.

did new tests with esgaroth, got almost a full stack attacking me on campaign map, it does not work anymore. however, Im not sure if it is due to them getting messed with a siege tower or the ai behavior itself.

[leo.civil.uefs]

ok mak, confirmed, the AI is working well in esgaroth, it was just a mess they did while trying to use the siege tower due to the small space for allocating it in front of the wall. so I just removed the docking points, so no more ladders or siege towers on the walls, it is ok to me to have esgaroth with no dockable walls but you can still place your men above them while defending or destroy the walls with siege engine while ttacking.. So esgaroth is done. Thanks mak!

I still have 2 issues:


Minas tirith.


The second gate works well and after breaking it the AI attacks the plaza with all its units.
Its working well for catapults, trebuchets and ballistas hitting it but the problem is the ram. The ram crew gets lost in the curves of the city and the men manning the ram itself will not start hitting the gate until all crew is near it.

Point is, works in all maps on custom battles but not in the campaign map. Tried the height thing like we did with esgaroth ( +200, +60, -100) and no deal. Tried changing the Minas Tirith region to another climates but still no deal.

EDIT: FORGET EVETYTHING I SAID ABOUT MINAS TIRITH. THE BUG IS SPORADICAL BOTH FOR CAMPAIGN AND CUSTOM BATTLES.

Cair Andros and other settlements:

The idiotic ram/ladder crews running arround it and hitting each other, like I described before. It does not happen in Dol ALmorth, wich share the same settlement.
As I said before, its not a climate bug, I tried using the climate used in dol amroth on cair andros and no deal.

Actually, I tried chaging the cair andro's region on stratmap to be the same from dol amroth in all tga's on bap/base, no results.

I also noticed the same problem happening in barad dur but again, only in camapign battles, custom battles are not affected by this bug.

Its a mistery, though we know that it does not happen in dol amroth so this is the way I need to follow...






Any guess?

[leo.civil.uefs]

more tests.

for minas tirith, I tested lots of different custom location and they all worked well except RHUNIC STEPPE, HILLS OF RIDDERMARK and HARADWAITH. They were th only ones to give me the same bug as the campaign battle, though it seems sporadic, sometimes RHUNIC STEPPE works well.

I noticed the unit moving the ram could be getting confused by the amount of other units using the same path so then I got anothjer test in campaign battle with less attackers, still bugged. Did it then to a working custom location, with tons of units, they were all messed with the ram unit but still worked well, so it is not the cause of the problem.

The strange is that the custom location set for minas tirith in custom battles, by default in TATW has exactly the same coordinates of minas tirith in the campaign map so they should not be different.

Tried changing the map_heights on minas tirith location on campaign map, darker and brighter, no results, still the same.

Any ideas?

EDIT: FORGET EVETYTHING I SAID ABOUT MINAS TIRITH. THE BUG IS SPORADICAL BOTH FOR CAMPAIGN AND CUSTOM BATTLES.

[makanyane]

If the custom battle site is identical to the campaign map location then I can't see any building/climate related reason why it'd be different.

Your later tests where the results were 'sporadic' makes it sound like the whole thing might just be sporadic... That or back to looking at the battle ai which I don't understand.


Only suggestion is have you tried altering the descr_pathfinding values for the rams/ladders?

[leo.civil.uefs]

Only suggestion is have you tried altering the descr_pathfinding values for the rams/ladders?

ok, what is this about the .txt one or the .db one? I dont even know if it is possible to edit the .db one, guess is the text one so?



So, I did some tests just by editing the .txt file.

min_slope    10        ; minimum slope before penalty applies
max_slope    50        ; maximum slope the unit can handle
penalty        1.03    ; penalty per degree

these lines seems auto explanatory. I've tried changing each one individualy, and all times I got similar results.

Yes, it seems like the problem is related with the slopes/angles.
I got it to work well while making such tests.

What is still causing mental suffering is why the hell it works so fine in some custom locations for custom battles even with no changes in this .txt file.
More than this, custom battles still seem to work better than the campaign one even during the tests.

EDIT: FORGET EVETYTHING I SAID ABOUT MINAS TIRITH. THE BUG IS SPORADICAL BOTH FOR CAMPAIGN AND CUSTOM BATTLES.

[leo.civil.uefs]

forget what I said about it beeing sporadical.

Take it as permanent, as I did many many tests.

EDIT: FORGET EVETYTHING I SAID ABOUT MINAS TIRITH. THE BUG IS SPORADICAL BOTH FOR CAMPAIGN AND CUSTOM BATTLES.


I dont belive the descr_pathfinding is the way, plus it would affect all the game so I dont want to risk it.

How could I know the "height" of the custom maps I tested? This way I could set the same for the campaign map one and/or lowering/elevating the entire settlement in IWTE. I belive it is related with the heights, as esgaroth.

[makanyane]

I dont belive the descr_pathfinding is the way, plus it would affect all the game so I dont want to risk it.

I really meant temporarily - there's movement cost per degree slope, absolute max slope, and a min width thing in there... if it turns out changing them fixes things, it'd give a clue if slope or width of access is part of the problem.

How could I know the "height" of the custom maps I tested? This way I could set the same for the campaign map one and/or lowering/elevating the entire settlement in IWTE. I belive it is related with the heights, as esgaroth.

only thing I can think of is to make a stepped pit somewhere in the terrain, I think if you go below sea level and don't have a reflective plane, you'd get water in it... so you'd be able to see how close you are to sea level. Otherwise just try and work out where the location is relatively on your map_heights and see visually if it looks like light or dark bit.

[Germanicu5]

I've just accidentally found this thread, I doubt I can do additional research myself.

Is there some progress on bugtracking?

As for pathfinding - it was altered for TATW to enable attacking new settlements (more slope tolerance) and on some mountain maps that were previously unplayable, as far as I remember geography files were also changed to flatten terrain a bit.

Now for trouble with attacking some settlements - the most popular issue can be seen with Imladris - it is defined as a settlement with a wall (in world files), so AI can't attack it. I scripted manual attack in this case, but it is only available with the latest RBAI.

If we refer to AI attacking citadels etc. - all settings were optimised to work with variety of settlements, but AIs stage progression can be easily interrupted, although I made sure it's as close to optimal as it gets.

I don't recommend abandoning scripts for TATW as they only help with sieges.

Regards

[leo.civil.uefs]

I've just accidentally found this thread, I doubt I can do additional research myself.

Is there some progress on bugtracking?

As for pathfinding - it was altered for TATW to enable attacking new settlements (more slope tolerance) and on some mountain maps that were previously unplayable, as far as I remember geography files were also changed to flatten terrain a bit.

Now for trouble with attacking some settlements - the most popular issue can be seen with Imladris - it is defined as a settlement with a wall (in world files), so AI can't attack it. I scripted manual attack in this case, but it is only available with the latest RBAI.

If we refer to AI attacking citadels etc. - all settings were optimised to work with variety of settlements, but AIs stage progression can be easily interrupted, although I made sure it's as close to optimal as it gets.

I don't recommend abandoning scripts for TATW as they only help with sieges.

Regards

I've trying to contact you FROM YEARS now and you never answered me.

My submod managed to fix everything and make everything to work "perfectly" as possible in the engine. I dont need anything more.
I would say TATW has now 85% or so of AI performance in siege battles in the custom settlements. Wich is pretty enough for me.

[Germanicu5]

@leo.civil.uefs
I apologise, but I've been absent for the past 1.5 years and I haven't even managed to look through all private messages as I'm trying to actually mod, I'll gladly look into your findings, I bet KK would also appreciate your ongoing work.

To answer your PM in public - no, my latest AI didn't make it to TATW 3.2, which is a shame, since I finally got rid of a rare CTD with reinforcing armies when it was released.

I also give my permission to include my AI in any submods if it's included in the main mod.

As for the AI itself - if it can be improved, it's always w.i.p. for me.

Regards

[tomdoof2236]

Ok are you guys talking about the ram/tower bug? they try to say dock the ram and then the AI pulls it back and fourth constantly? I ran into this myself and found a few interesting results. Seems RIVERS on the battle tile cause this bug for sure, units with warcry cause this bug, water does not seem to cause it. It will throw a code SEE CHARLIE or something like that if the tile has a river on it. But settlements on battle tiles with rivers seem to break the AI.

[leo.civil.uefs]

Ok are you guys talking about the ram/tower bug? they try to say dock the ram and then the AI pulls it back and fourth constantly? I ran into this myself and found a few interesting results. Seems RIVERS on the battle tile cause this bug for sure, units with warcry cause this bug, water does not seem to cause it. It will throw a code SEE CHARLIE or something like that if the tile has a river on it. But settlements on battle tiles with rivers seem to break the AI.

Sorry I never had problems with this (ram/tower)

Everything I know is in this thread.