IWTE - World editing - (General Discussion)

[leo.civil.uefs]

[QUOTE=makanyane;13518504]How vanilla? absolutely all the settlement files inc terrain? And are you sure it ever worked for that location? infantry and rams and things all have different degrees of slope they can cope with, so it might be just that they can't use the direct route up the slope (too steep) and are therefore getting lost trying to plot a zig-zag route??? The actual vanilla terrain has that flattening affect from the merge plane thingy ( the one I never know what to call )

Absolutelly vanilla except for textures, wich are now the gondorian ones.

I It never worked well in that location, but as I said, works well in custom battle in a diffrent terrain.

Santa says, that he said "by the end of the year..." - he's still trying - and also Please answer his question about which version of milkshape you use 1.8.2 or 1.8.5 etc?

IWTE_v13_11_C is what Im using right now.

I always use the newest version.

[makanyane]

Thanks Tellos,
also judging by the way the terrain textures get influenced by what would have been on the battlemap (shore/hill type stuff) then maybe that can have an effect too in terms of what it thinks the terrain movement type is (?) - Leo's issues seem to be about bits that wouldn't have been accessible anyway, that still aren't accessible but by different methods, I'm wondering if there's a difference in calculation cost (nodes/whatever) according to whether something is blocked by pathfinding, or by slope, or by water or combinations. Giving the ai possible routes from A-B doesn't always mean they actually find them...

Santa says, that he said "by the end of the year..." - he's still trying - and also Please answer his question about which version of milkshape you use 1.8.2 or 1.8.5 etc?

IWTE_v13_11_C is what Im using right now.

I always use the newest version.

please note, I'm not shouting....

[leo.civil.uefs]

please note, I'm not shouting....

You should... that time, it was my fault.

1.8.5 BETA 1

[Tellos Athenaios]

terrain steepness might be telling the ai something that the pathfinding on it's own isn't - if just removing the reflective plane changes it that'd be very odd

It does. If you take a look at data/descr_pathfinding.txt you will see that per unit type class (infantry, cavalry, etc.) there is a maximum slope which they can handle, including penalty costs for the path that takes them over the slope. Also this cost may vary according to terrain type and terrain type seems to be according to some classification by the engine. At any rate terrain type does not correspond to the normal ones we know from e.g. sound effects. It seems as if the engine segments a map based on certain criteria into areas which are, for example, "obstructed", "steep terrain" or fit any of the other categories or none ("free") which allows it to prefer marching over open terrain to, say, drowning in water.

Presumably the engine will prefer the route with lowest cost to its target, and depending on the slope of the paths towards it, a target may even be impossible to reach for certain classes of troops.

[leo.civil.uefs]

Here is one of the issues caused.

Its is absolutelly the same settlement from vanilla med2, except for textures.

It works absolutelly well for AI in custom battles, but not in the strat map, as you can see on the pictre, wich is self explanatory.



Ive tried enlarging the cair andros isle, giving it more terrain arround the settlement, Ive made it really large, actually turnng it on part of the continent and not a isle anymore, and yet the problem stills.

I can make them make more openend curves and then avoid the problem by editing the pathfinding by blocking it in the area where they cure, but I see now pathfinding is so complex and sentitive that I dont would risk this.

It seems it is the same problem as esgaroth.

[makanyane]

Its is absolutelly the same settlement from vanilla med2, except for textures.

Leo - did that go through IWTE to change the textures? There are some slightly unexplained numbers at the very end of the file that might be doing something if we changed them... can you try it with the absolute vanilla version?

but otherwise it could just be a difference between exactly how it treats custom battles versus strat stuff, which would be very interesting to understand!

[leo.civil.uefs]

Leo - did that go through IWTE to change the textures? There are some slightly unexplained numbers at the very end of the file that might be doing something if we changed them... can you try it with the absolute vanilla version?

but otherwise it could just be a difference between exactly how it treats custom battles versus strat stuff, which would be very interesting to understand!

Tried the settlement absolutelly vanilla including textures... no difference. Problem stills.


Also, it was actually a citadel (3 line sof gates) but set as a large_castle. So I tried it as a citadel, so matching the type with the model, and all I got is a different bug, but the same problem as before, they attack with ladders but still get stuck, running arround the ladders and hitting each other like idiots.

[leo.civil.uefs]

sorry for the double post but.

I've made osgiliath to be set as the gondor fortress (like cair andros), and the problem still happens there.

One interesting thing I noticed, in this test, both the reflective plane from the settlement and the water from the river were loaded! I though it was not possible to have 2 water surfaces separated and in different heighs. Reflective planes in the settlements had always abrogated the water comming from the strat map showing just the river/sea/lake bottom. Wird, the more I mess with this the more thing appears. lol

[makanyane]

I've made osgiliath to be set as the gondor fortress (like cair andros), and the problem still happens there.

there'd be another interesting test that could be done, which is basically temporarily wipe most of the map info, and use 5,5,5, as map heights everywhere, temperate_deciduous_forest as climate, and one of the fertility_medium type things as ground type, which is about as bland as you can get... regenerate map.rwm and then retest and see if the problem persists in the campaign battles.

[leo.civil.uefs]

there'd be another interesting test that could be done, which is basically temporarily wipe most of the map info, and use 5,5,5, as map heights everywhere, temperate_deciduous_forest as climate, and one of the fertility_medium type things as ground type, which is about as bland as you can get... regenerate map.rwm and then retest and see if the problem persists in the campaign battles.

tried it mak, the cam,paign does not even starts...lol
To be honest, I dont see how it could help, and judgind by the tone of your post, I see you dont know much more than me about this, although Im happy you still try to help me even so.

Cannot find a fix for this BIG problem, its really a issue for my patch.

Pray for me.

---------------------------------------


EDIT: Ok, a new test with a interesting different result. I tried placing the settlement in minas tirith, far from tge river so no water involved. The result is, the AI can now use the ram properly, I mean, it happens the same problem as described in the picture I posted but in a far smaller scale, and at all, a single confused man runs around the ram but soon get in his position and they ram the gate.

come on mack you're far more inteligent than me, tell me what that means.

I also did several tests using different locations in custom battles, all locations works well.

Even with this test in Minas Tirith, I can say what is happening. When trying on a region of rohan, they came with a ram and ladders, and at all got stuck in the first perimeter, doing nothing.

Its quite dificult to test with all siege engines setupd in campaign battles as you never know what they will choose.



-------------------​-------------------​-------------------​-------------------​-------------------​-------------------​-------------------​-------------------

Ok. Swagger found out what was causing the issues. It was the ReallyBadAI used in TATW 3.2, causing ladders and rams crews to malfunctioning. Both med2 vanilla and newest version of RBAI can fix this problem.

However... despite of these problems beeing fixed, the AI still dont know how to take the teutonic citadel, maybe I will need to remove it. Anyway they never worked well even in vanilla. Do you agreed that this issue is more related to our old blue/green pathfinding editing via IWTE than to texts like the ones from RBAI?

Germanicu5 included a descr_geography_new.db and text in his RBAI latest version... why? Has it any relation with AI pathfinding or something?

[makanyane]

I always thought the Citadel problem was it didn't really handle multiple perimeters very well anyway... best I've seen it work is when it can blow up walls with artillery and then take sets of unused ladders through the gap, if it has to remember to go back and pick up used ladders/rams after it's captured a gate it seems to struggle to get organised enough to get to 3rd perimeter even without opposition

how has it been working on your 3 perimeter jobs? Does having less potential things to attack on those make it work better?


descr_geography_new.db can change the terrain of the battlemap - generally not the stuff under the settlements as that should get over-ruled by the .worldterrain/veg
but the db file can alter the amount of hills/slopes etc for a climate/ground_type combination (you can actually make maps where it's impossible to move at all, if you put in silly values!)
it also alters the amount of trees & large rocks that are distributed in a climate, and those have a sort of exclusion zone thing to stop troops walking through them (set in the vegetation.db)
so those things can effect movement on normal battlemaps - you'd need to do a file comparison to see what he actually changed. And by the way if you're including it, it's better to only include the .db file, and don't use rebuild_geography_db = true in the .cfg file

[makanyane]

I think you can get river and reflective plane water close to each other, and appearing in the same settlement, but not touching, I remember a previous experiment and conversation about this with KK. The bits you can't do are use a reflective plane with any sea inlet water from coastal areas (and probably depth generated water which is likely the same thing in M2's mind), or use two reflective planes at different levels.

However that's all rather a long time ago so I might have some of it mixed up... old age is not helping my memory

[wilddog]

There will be a major bug fix in the next IWTE version. We noticed it in recent testing though it has been around for several years now (though not sure why not reported as its a pretty serious bug). The problem seems to be tied to a process I was trying to fix due to the different handling of 'unique' vertexes. A routine was written to expand the number of unique combinations of vertexes, UVs and Normals but NE-castle walls (and those used in the NE stone forts actually contracted by 2 (which I hadn't thought would happen) causing the triangle indexes to be completely messed up. I've now corrected but took quite a while to find. I need to ask Mak to try and check if the uniqueness has to be within group versus structure as we both remember some issue with another city that didn't seem to follow the same rules.
I am hoping it doesn't but it may still have some repercussions mostly tied to processes to split or recombine structures.

We still need to complete the complex lighting as we spotted something strange on stone fort C for one of its walls when damaged so need to retest that and I'm still to complete the light shading bit fully.

[leo.civil.uefs]

As I said all the bugs rerarding to rams/ladders crews are now fixed but, the curious of all this is, the citadel works absolutelly well in custom battle and still not in the campaign map in themr of pathfinding, the AI get totally lost.

I dont know the files included in RBAI, but there is for sure something different from custom battles to campaign battles.

And that brings me to... oh my... all my settlements were tested in custom battles and using a earlier version of RBAI... wich means, there is a change a cataclism ca occur in my mod now.

Does anyone knows if there is really parameters that makes the AI act different in custom and campaign battles?

By the way I've done some tests in campaign for southern european citadels and it seems that they are bugged too. They work well for when they come from the front of the settlement and assault the gates in parallel one by one. But I have disabled the front gate and made them comne from back, supposing they would then enter the settlement and circle the second line of walls to attack the second gate and then do the same for the third one, they just broke the second gate and then got stuck. Im goig to totally disable vanilla citadels with 3 gates from my sub-mod. Dain's Halls has 3 gates and work pretty well... at least in custom battles, and I belive this is because the 3 gates are lined and parallel.

[wilddog]

OK after several days trying to track down some strange lighting tied to the effects (which led to the discovery of the problem posted above the last post), we have now concluded our strange lighting effect has actually always existed. The issue is tied to the compressed embedded references within the animinstance file. These do have references to the same complex related effect and shading information. In the copy animation process you can choose a similar animation which we thought was all working but it picks up the references to the complexes and effects of the .world object it was picked from. This is incorrect and leads to the situation where the animation is looking for a complex effect which is either wrong or simply doesn't exist. This in turn causes the weird lighting issue we have been struggling with to resolve. Its mostly an issue tied around ending up with less lighting effects than that particular animation contains although obviously its also affected by distance etc relative to that particular animations lighting effects.

For the EB guys, the DBM guys never got anywhere with effects other than copying our default which was wrong. I don't think they ever assigned any effects although the default used expected one. If they didn't use any animations anyway then it won't be a problem as the issue we are trying to resolve is purely tied to using animations.

Leo, sorry we don't know (Mak neither) so she suggests you ask on a more general forum or on a thread associated with Battle AI questions.

[leo.civil.uefs]

hum... Ive tested a vanilla teutonic citadel in TEUTONIC CAMPAIGN expansion, and it works prety fine fr the AI... the poinst is, why then it does not work well in TATW?

Might be the AI configuration files...

[wilddog]

Final version for this year

with version 13_12_B
1. Adjusted effect position slightly as it was using just the centre of the collision to position rather than the collision's transformation matrix. The resetting of the complex lights button will reset this for all existing effects to ensure correct alignment for previously loaded effects.
2. Complex effect information now added with vertex assignment. Please use button under collisions tab to set it. This replaces previous default information which was actually incorrect. Currently the distance is set to 18 metres and takes into account culling by intervening effect facing triangles. (I may tweak further but it takes approx 2.5 minutes on my machine to run and twice as long on older machines. It also hogs memory which I will try and address in future versions.
3) Shading assignment has been added to replace the flat default setting of 235. Please use button under specials tab to set. This routine needs to be smoothed off as it currently doesn't allocate to triangles which meet at a point (rather than a line).
4) The 3d lighting viewer has been left as is and shows the shading if no textures are found (so you can see what I'm talking about).
5) Shading and vertex info is now written into the milkshape model. The default is version 1.8.5 (uses vertex subversion 3). If this is switched off it assumes vertex subversion 2 and only stores the information about 2 effect/vertex assignments. The effect assignments cannot be edited and should be reset currently using the reset button as mentioned in item 2. The shading can be set and is based on assignment to one of two bone ids. This is only in case the setting process hasn't set correctly.
6) A bug tied to the incorrect assignment of effect information to the vertexes has a temporary fix (which is likely to be permanent unless it causes other issues). The correct fix requires rebuilding of the correct vertex effect settings and adding them to the compressed component of the .anim file which is likely going to be a lot of work with very little eventual usage.
7) Fixed bug that caused wall models to get scrambled when reading from ms3d in very rare circumstances.


The milkshape version is only significant to the ms3d files from the .world files. I'll probably improve the other versions also at a later stage.
I still want to do more work on the complex effects and the shading to see if I can improve timings and smooth the shading.
The tool does put up a warning message on the need to reset the complex effects. This is really a reminder that in the end you need to do that. Its not necessary to do it each time but they will be technically out of line until you do reset them.

There are still further improvements tied to these processes but each one causes an increase in time to do due to the number of vertices. Main one I want to try and introduce is better culling when the triangles in the way don't face the effect. These are mostly associated with the damage ones where fires are placed inside damaged hollows and no facing triangles were introduced as you don't really see them (much). Looks like CA didn't cull those either though.

[leo.civil.uefs]

thanks for this. and count with me for testing the light effects.

[wilddog]

Leo - I believe Mak sent you a PM as your set up may have some difficulty as it doesn't really fir to the 'vanilla' sort of setup. I'll see what I can do to help on this. Main thing let me know any crashes. Mak had a couple but they were memory ones when she was trying to expand the light distances and was consequently picking up exponential numbers of vertices and triangles to test against.. whilst running several other things!

... have a great new Year's eve.

[AnthoniusII]

@Wildog
It comes to mind a vanilla feature that was strange.
Agrantonio mentioned in the early days of his attempts to this tool that vanilla game HAD orthodox church models (like the Catholic ones or the Mosques) but for some reason they did not apear in game or simply did not work. Is there any kind of hope to solve this matter?