CATE needs your help ...

[Skyfire76]

Well, in fact, I need your help ... ), and still no solution.

So, guys, and especially if you're used to terrain tiling, please help me.

Regards.

[Snake Man]

quote:Originally posted by Skyfire76:
forest type, with farm to the south, and forest around otherwise, with a road to the north and the east, then I should put the tile #1506, i.e. HFORR2F7.pcx

Now I must say that I have never used that tile type, those are very special tiles. But yes, in your above example it would work, but what about the next tile then due east... theres not seem to be road like that.

I suggest in this situation we extend the forest coverage one tile further.

PMC TFW
Snake Man

[Skyfire76]

I thought this was a simple example ...

Well, the tile to the east could be 1504, or 1553 or 1554 ... depends on the said tile ...

OK, next time, I'll choose a more basic example

Regards.

[Snake Man]

Well what if we just extend the forest coverage if we run into messy parts?

PMC TFW
Snake Man

[Skyfire76]

We'll end up with the biggest forest on earth, probably. Maybe it will even beat some Middle Earth forests, like Mirkwood, or at least Fangorn

Anyway, this more or less illustrates my point : I need rules to know how to tile terrain. As I said, I have all the informations I can think about, it's just a matter of translating them.

Regards.

[Snake Man]

quote:Originally posted by Skyfire76:
I need rules to know how to tile terrain.

So what if I tile some messy segments (after the current CATE results) and I try to explain along with images of what I actually did?

PMC TFW
Snake Man

[Skyfire76]

Let's try that, even if that's not what I'm looking for here, but it could give us some ideas.

Regards.

[Snake Man]

Now we would need input from other people too. So anyone, suggestions?

PMC TFW
Snake Man

[87th_striker]

When I tile, the slope of the place to place the tile is most important, along with the usual terrain features. The slope is defined by the adjacent tiles' height.

A stream should run in the lowest parts of the terrain, not on the sides, but sometimes the adjacent tiles makes it visually wrong, som I either move the tile, or adjust the altitude. Altitude adjustment really is a necessity in alpine terrain, to get the right "looks". For CATE I guess we don't adjust the tile height/altitude, so I guess we use move the tile. So then you would check the adjacent tiles to check wich way the slope would be. Also altitude will have some rules, but I think they are integrated ?

So, sorrounding terrain forms the rules, as long there is no special feature. If there is a basin, you would make some kind of plain or swamp, or a forest. It all depends on altitude. And sometimes you invert some tile usage, to create a visual "break" in the usual terrain. BUT, to get things good tiled, you "randomly" have to apply different set of rules to the same conditions.

Did this make things clearer ?

[Widowmaker]

hmm

tiling per slope (diffence in steepnes) is quite a good suggestion.... however transition/feature tiling mess first..

gonna think aloud
I still am not too aquainted with CATE
I use 1.21 still alot and 2.xx version
I have tested once.
So I got a clear view on things..


first some Knoledge Base stuff

Facts / Assumptions.. :

-tiling per altitude is 100%
-tiling per region is 100%
-transitiontiling is 100 % ok

-roadtiling on its own is rather ok
-rivertiling on its own is rather fucked..
-river road transition is a mess..

Concepts..:

-City= terraintype
after the first cityrun I see cities as regions
not

-3 way tiles..
Korea tileset has them only for Coast
for instance HCOST7A0 offset 704 in Korea texture.bin
it is a road city water transition

for other terrain types the Korea bin/zip does not have them
we need to make them to make CATE life's easier

-multiple(configurable) runs
so lets setup up an order of working
I think the order in wich to do stuff is very important
and CATE needs to make separate runs wich would be great for flexibility
the current way is very flexible
it would be nice to run 2 or more batches of scripts same scripts as now we do in 1 pass but difernet ones

-TerraintypeA or B type transition
I'd like to be able to set wich terrain is to get the transitiontiles
in the case of a transition between regionA(offset16) and regionB(offset32)
I wanna be able to have the transition tiles tiled on top off Region A or on top off Region B
this is a crucial part also


later the runs themselves
I thought I posted this already to get familair..

Widmak

[Skyfire76]

quote:Originally posted by 87th_striker:
When I tile, the slope of the place to place the tile is most important, along with the usual terrain features. The slope is defined by the adjacent tiles' height.

A stream should run in the lowest parts of the terrain, not on the sides, but sometimes the adjacent tiles makes it visually wrong, som I either move the tile, or adjust the altitude. Altitude adjustment really is a necessity in alpine terrain, to get the right "looks". For CATE I guess we don't adjust the tile height/altitude, so I guess we use move the tile. So then you would check the adjacent tiles to check wich way the slope would be. Also altitude will have some rules, but I think they are integrated ?

So, sorrounding terrain forms the rules, as long there is no special feature. If there is a basin, you would make some kind of plain or swamp, or a forest. It all depends on altitude. And sometimes you invert some tile usage, to create a visual "break" in the usual terrain. BUT, to get things good tiled, you "randomly" have to apply different set of rules to the same conditions.

Did this make things clearer ?

Some good ideas here, as far as I can tell. I have already thought about the randomness of some rules, and this could be done, I think, but only when we have found some way of writing simple rules first ...

Same thing goes for slopes : let's try to have a version correctly doing transitions/features/terrain, and then we can enhance that.

Regards.

[Skyfire76]

quote:Originally posted by Widowmaker:
Facts / Assumptions.. :

-tiling per altitude is 100%
-tiling per region is 100%
-transitiontiling is 100 % ok

About the last one : transition tiling is OK on its own, and as long as you don't have too many different terrain types close together.

quote:-roadtiling on its own is rather ok
-rivertiling on its own is rather fucked..

Really ? That's strange because roads and rivers are sharing the same algorithm, so I can't see a reason why one would work and not the other ...

quote:-river road transition is a mess..

Agreed on this one

quote:
-City= terraintype
after the first cityrun I see cities as regions
not

Not sure of what you refer to here, but what I can say is that the way tiles are named/ordered, cities, for CATE, are a terrain type and not a feature (like rivers).

quote:-3 way tiles..
Korea tileset has them only for Coast
for instance HCOST7A0 offset 704 in Korea texture.bin
it is a road city water transition

for other terrain types the Korea bin/zip does not have them
we need to make them to make CATE life's easier

Hum probably, if I understand you well ... but the problem for now is to find a way to describe all tiles/terrain/rules in CATE so that I can find the correct tile number ...

quote:
-multiple(configurable) runs
so lets setup up an order of working
I think the order in wich to do stuff is very important
and CATE needs to make separate runs wich would be great for flexibility
the current way is very flexible
it would be nice to run 2 or more batches of scripts same scripts as now we do in 1 pass but difernet ones

I'm not sure I understand you here. What I can say is that sometimes, I won't have the choice : some things need to be done before others...

quote:-TerraintypeA or B type transition
I'd like to be able to set wich terrain is to get the transitiontiles
in the case of a transition between regionA(offset16) and regionB(offset32)
I wanna be able to have the transition tiles tiled on top off Region A or on top off Region B
this is a crucial part also

Again, this is related to the way tiles are named/ordered. But there's a workaround for this, which I did not really test. Ask me by mail (and see last part of this post before ).


At last, guys, my mail server is FUBAR since yesterday afternoon. You can still send me mails, but I can't read them, and I can't send any mail either. I'll tell you here when it's working again.

Regards.

[Zaggy]

One thing id like to point out BTW, is that the e00's (and hence TDF's) seem to place cities based on the density of population... ie, only the heavily populated areas of cities ever seem to be tiled as cities... 'Suburbia' and such seem to be ignored... Perhaps CATE could be modified to tile cities, and then also tile 'Suburbia' for another, say 5-10 tiles, around the DEFINED city areas...

Just an idea...

[Skyfire76]

This could be done, I think ...

BTW : I made a little update, trying to clean rivers a little more.

As I said before, there's some randomness involved when calculating paths for rivers and roads.

What I've done here is that, when calculating rivers and that I come to a point where I should choose randomly, I check before if there is a tile with less "roads features" than another. If so, I choose this one. If not, I do the usual random trick.

This way, we will probably have a little less messy overlapping and such. Nothing exceptional, though.

Anyway, I can't send it yet since my mail server is still FUBAR ...

Regards.

[87th_striker]

Why not post some examples here, and we all could comment on how to improve each example. Then we let that form some of the rules.

And while I'm at it, the lake (river) problem has to be fixed.

[Snake Man]

quote:Originally posted by Skyfire76:
Anyway, I can't send it yet since my mail server is still FUBAR ...

Hmm damn shame, we should try to get ftp server for you to upload. Do you have full ftp access or is that limited also?

PMC TFW
Snake Man

[Skyfire76]

Not completly sure : they're using a proxy, and I guess it's a sort of simulated FTP via HTTP. Shortly said, I can't use a "real" FTP client, but must use I.E. to do it.

We could try, though ...

Regards.

[Skyfire76]

Back to the original topic /question, and I'll try to be a little more specific

What I have, now, for each tile, is a structure with the following info :

- Terrain : an integer telling me which type of terrain (as defined in conf file) the tile is.

- Airbase = 0 (no airbase), 1 (tile around the AB, used when leveling terrain), 2 (there is an AB here)

- River : 4 bytes array telling me if there is (=1) or not (=0) a river in each side of the tile (the 1st byte is for N side, 2nd for E side, 3rd for S side and 4th for W side)

- SideTerrain : 4 bytes terrain telling me what is the terrain type on tiles around (1st byte for N side, 2nd byte for E side, 3rd byte for S side, 4th byte for W side)

- CornerTerrain : same thing, but for the tiles in the corners (1st byte for NW corner, 2nd byte for NE corner, 3rd byte for SW corner, 4th byte for SE corner)

Example :

- terrain=1
- airbase=0
- River(1)=1 / River(2)=0 / River(3)=1 / River(4)=0
- Road(1)=0 / Road(2)=1 / Road(3)=0 / Road(4)=1
- SideTerrain(1)=SideTerrain(2)=SideTerrain(3)=SideTerrain(4)=1
- Cornerterrain(1)=CornerTerrain(2)=CornerTerrain(3)=Cornerterrain(4)=1

It's an easy example, with a plain terrain of type 1 (let's say it's farm), with a bridge for a E-W road / N-S river, surrounded by type 1 terrain tiles only (so no transitions).

So, what I need is a way to write the conf file so that I can easily match all those informations with a tile number ...

I hope this clears things up ...

Regards.

[Skyfire76]

I see you've been busy submitting ideas ...

Anyway, one quick question : how many different terrain types (forest, plain, ocean, cities, swamp, desert and so on) can I expect to find in any theater ? Give me the upper bound ...

Regards.

[Snake Man]

quote:Originally posted by Skyfire76:
I see you've been busy submitting ideas ...

quote:how many different terrain types can I expect to find in any theater ? Give me the upper bound

Why is there going to be some limit in CATE?

I'd say we should reserve space for lets say... hmm 20 different types. If thats not doable, the minimum would be 10. If we drop much below 10 then we are most likely not be even able to tile default korea.

PMC TFW
Snake Man

[Skyfire76]

I don't know yet (about a limit).

For now, the only solution I see (still waiting for better ones, guys ) would be to take all these informations above and transform them into an unique number (somewhat what was said in the autotiling thoughts thread). The conf file would then just be some translation between these unique numbers and real tile numbers.

It's rather easy to do that as long as we just look at plain terrain and features. But when transitions come into play, its another story ... That's why I may have to put a limit, but nothing sure yet.

Regards.

[Skyfire76]

Some update about my ideas ...

Let's forget transitions for now, and let's say we define an unique number for each tile with this formula :

TerrainType*256 + RiverN*128 + RiverE*64 + RiverS*32 + RiverW*16 + RoadN*8 + RoadE*4 + RoadS*2 + RoadW

With this we can define all plain terrain tiles with all features, in a conf file where you associate such calculated numbers to real tile numbers. And as I considerer cities as a terrain type, you can define cities with roads easily.

Another advantage is that if you don't have, for example, a tile with river + road crossing diagonally, just define with which tile number you would like to replace it (this will probably be in this case a tile with a road only).

Now, the problem is transitions

One idea would be to tile transitions in a first pass with CATE (like it is done now), and then define specific terrain types for each transition (ie. one type for forest to plain, one type for plain to ocean, and so on).

This would work, I think, though it's not really what I was trying to do.

And we still have the problem of transitions between more than 2 terrain types (e.g. forest in the middle, plain to the N, ocean to the E), as well as rivers which are still messy.

Any comment on all this ?

Regards.

[Snake Man]

quote:Originally posted by Skyfire76:
transitions between more than 2 terrain types (e.g. forest in the middle, plain to the N, ocean to the E)

The areas which have three different terrain types are more or less rare. Of course it happens.

Also we should make rule if we want to tile features into ocean tiles - 0 elevation and 0 tilenum. I'd say in my theaters it would not be allowed (but Widowmaker may want something like that to Iceland). That would cure one messy part in the coastline right away.

PMC TFW
Snake Man

[Skyfire76]

I do not see your point about tiling in ocean : last version sent already has an option to apply BMP rules even if on ocean.

Is it what you're talking about ?

Regards.

[Snake Man]

I mean NOT tiling features into ocean tiles. At least in the european L2 file lot of features get tiled into ocean tiles because they are so near (and over) the coastline.

PMC TFW
Snake Man

[Skyfire76]

Hum, I see ...

Note that the above method should allow you to correct this at will : you don't want features on coast tile ? Then, define a tile number pointing to a tile without feature for all CATE tile numbers with features.

Hum, this is not probably very clear.

Let's take the example of forests and coasts (with "random" numbers) :

- Plain Forest= 256 for CATE = 68 for tile number
- Forest + roadW = 257 for CATE = 70 for tile number
and so on

- Coast = 512 for CATE = 125 for tile number
- Coast + roadW = 513 for CATE = 125 again for tile number
and so on

Thus, each time CATE thinks it is tiling roads in coasts, it is simply tiling plain coasts.

Regards.

[T_Rex]

Sorry I was very busy earlier in the week, although I'm not sure what else I could add.

T

[crossup]

Skyfire,
please contact me via email or icq. I have something that may solve most of our problems(and introduce new ones of course). Julian has suggested that CATE could solve my problem in implementing this.One of my projects involves using hires terrain, and with 16x as many tiles, hand tiling for even the smallest island is insane.I have most of the puzzle pieces, I just need a simple algorythm. If we can figure a solution then tiling becomes merely a matter of computer time.
crossup
ICQ 9165219
cross.up@verizon.net

[crossup]

Skyfire,
I forgot: work email- stevew@marylandmicro.com
best way to contact me during US daytime
but Im on ICQ when ever Im online
crossup

[Skyfire76]

T-Rex : you're right, in a sense. Terrain type is also a unique number. But if you want to take into account other features, this above way ensures that you get unique numbers, unless I screw up something, of course

Moreover, due to some personal problems, I'll have to leave sooner than before during some days/weeks, so for example I'll be leaving in less than half an hour today. I'll email you right now (on your both accounts to make sure you receive at least one), but I'm afraid I'll read your answers only tomorrow. And we're lucky that my mail server just got back online after nearly two days out

Regards.

[Skyfire76]

So, SnakeMan, WidowMaker, or other CATE users, what do you say about the idea above ?

As far as I can see, it will work, and has the advantage that you will more or less design how CATE will tile. For example, you could replace some river tiles with lake ones and so on.

The main problem will be to write down the terrain definition file. For each terrain, you will have something like 255 tile numbers to put in the right order, and don't forget that each transition should be considered as a terrain also.

Regards.

[Widowmaker]

I am ok with it.
whatever the size of the definition files need to be it needs to be..
100 times better then handtiling.. lol

I'm all for it

Widmak

[Skyfire76]

Good

Two questions :

1 - There are two ways to do this. The first is to force the user to write down all the numbers CATE needs (i.e. 255 numbers per terrain type). The second will be to let the user omit the ones he does not need/have. In this last case, we should decide if CATE should do nothing for such omitted tiles, or use a default tile (and in this case how do we define it ?).

2 - Do you see one format for that file that would be easier for you to fill ?

Regards.

[Snake Man]

I have full confidence in your Conf file making abilities. You have done outstandig job so far.

PMC TFW
Snake Man

[Skyfire76]

Hi, asking for help again

I've been doing some experiments with CATE lately, still trying to enhance rivers and roads tiling.

What I did is try to include in CATE some A* pathfindind algorithm.

My problem is that such an algorith (for those who know it) work very well on a fixed map (i.e. where we know in advance which tile is easy/hard/impossible to cross).

But with CATE, what I'm trying to do is choose the best path, i.e. the one where the tiles on which we will pass are all defined in the conf file. And the problem is that, for example, if you make a river enter a tile from the south, it may result in a non-defined tile, while if you make it enter from the east, it may result in a correct tile.

So, which tiles are good or not depend on the path we choose, which become at the same time variable as we advance in our search. Needless to say, this is almost undoable.

I managed however to update the A* algorithm to come up with something that is a little better than what CATE was doing before (to be honest, it's better in some cases, transitions for examples, but worse in others, bridges for example).

If you have any idea to enhance this ...

But IMHO, there is no hope with this (a shame, really ...). So, what I would want to try now is to calculate ALL paths between two tiles, and choose the one which has the less undefined tiles.

Problem is I can't manage to come up with a working algorithm for this. So, if anyone has such an alogorithm, I would be most grateful to see it ...

Regards.

[Skyfire76]

I'll give an example of what I'm looking after.

Let's say we want to go from tile A to H in the following "map" :

A B C D
E F G H

I need to check all these paths :

A-B-C-D-H
A-B-C-G-H
A-B-F-G-H
A-E-F-G-H

(Note that I exclude "coming-back"s, like A-E-F-B-C-G-H).

Other example :

A B C
D E F
G H I

Paths to check :

A-B-C-F-I
A-B-E-F-I
A-B-E-H-I
A-D-E-F-I
A-D-E-H-I
A-D-G-H-I

(Note : in CATE, those tile have the following coordinates : A is 0,0 / B is 1,0 / D is 0,1 / ... / I is 2,2)

This should be not so hard, but @#$*ù#!^, I can't find it

Regards.

[Skyfire76]

BUMP !

No one ?

Regards.

[Skyfire76]

BUMP BUMP

Regards.

[codec]

I think what you are looking for is a minimum spanning tree algorithm. This is a way of finding the shortest path between two nodes going via other nodes, where nodes are connected together with a cost.
See Graph Algorithms for one example.

In your case, I would pick a cost of, say, 1 for going from one square to the next if there is a suitable tile, and maybe 100 if not.

The route cost will be like 3 or 4 for completely connected, 102 maybe for one with a missing segment and 201 for one with two missing.

There are other examples of the algorithm on the net, you may find a better example.

Julian

[codec]

Some other useful pages...
minimum spanning tree algorithms
http://www.ics.uci.edu/~eppstein/161/960206.html
http://www.cs.man.ac.uk/~graham/cs2011/greedy/
has an applet showing the graph in action.

Julian.