Random Map Generator - How to hook it into the game?

[Sllypper]

I used to share this dream.

Faf client can generate maps easily enough.

Duck got map gen working, I have played on his maps. He did a good job and broke through lots of barriers.

Twin rivers and Loki are interesting in a way that random won't be. Random isn't good, careful design and craftsmanship is good.

I personally would play random maps, but id have low expectations.

Random maps could fine tuned manually by real mappers

You're right, but I also feel that Duck went to a different direction I would. He generated asymmetrical topological maps and tried to fit them to be playable on FAF. I would go the opposite way: generate a map through its gameplay elements first (spawn area, expansion, reclaim pools, valleys, mesas, rivers, lakes), make the negative area unplayable and make it symmetrical. Not undermining Duck's work, which is fantastic and pioneering, but I'd like to see this approach and I think it would work for better playability.

Edit: What I would look for in a random map generator is for it to be capable to generate from 1v1 to team games and 8P Phantom games, for it to be fair, not mattering how bad the map is for everyone (like 1 mex flat maps, or tiny island maps), and for it to have several different map types (topographical) and skins (texture, color and lighting wise).
I feel that this approach would be able to fulfill this with more ease.

[Katharsas]

Guys, i think your should look at Galatic War, it would be a perfect combination to have the Galactic War maps being randomly generated. Galactic War already generates its universe and links between maps randomly i think.

And since Galactic War has its own backend + API + client, you can use it as a way of testing randomly generated maps without the need to make changes in the "normal" FAF server/client.

See: /viewforum.php?f=50

[nine2]

Well I'd be interested in seeing what you accomplish. The cool galactic war thing would be if you can generate 10 similar maps for one planet.

You could use much code from duck and ozonex who both published their code. How will you implement this? Easiest for me would be c# cos I'm good at it and all the preceding work is in c#. Then it would be harder to integrate with Linuxy systems. Could always host yourself and have a web service.

[Morax]

I'm not really a coder beyond some simple sensors and motor control, but would this not be extraordinarily huge AI-level effort to make something that is NOT as simple as some of the GW maps?

[Sllypper]

I'm not really a coder beyond some simple sensors and motor control, but would this not be extraordinarily huge AI-level effort to make something that is NOT as simple as some of the GW maps?

It's actually fairly easy to randomly generate something that doesn't look like natural terrain.

The GW map generator makes maps using realistic terrain generation. This approach i'm proposing would do the opposite, create the gameplay elements first.

For example, you first get an empty map and set the height levels to maximum. Then you sculpt out the map by generating "rooms" and setting paths between them. We could generate like 8 rooms per side and interconnect them so it doesn't look like a Voodoo map. You could have different types of paths like a lake, mountain, cliff; and different types of rooms, like a mass reclaim field, forest, plateau, second expansion...
Then you could make a second pass finding the empty area between paths and rooms and smooth it out.

It'd work kinda like the level generator from Spelunky (http://tinysubversions.com/spelunkyGen/) but less tile-y.
Edit: Or an oldschool roguelike map generation

[Sllypper]

Continuing the last post

It will not be easy to generate good maps. For example: if we tried to divide a map, let's say Hrungdaks Canyon, into rooms and paths it would looks something like this:



Team games would be difficult to balance how many mexes and mass each player would have, but as always, symmetry is our friend. It would be a lot easier to generate 1v1 maps. In hindsight, I should have done this to a 1v1 map instead

[Exotic_Retard]

i think that it would be a lot easier if the generator was not used live, but created a set of pregenerated maps instead - that way you can have the option of fine tuning beforehand:

lets say you have map confirmators or w.e.
these guys can either hit the generate button 10 times and when a good one comes up they say yeah its good.
or they can be given a choice of 10 generated maps and have to pick the best one.


given enough time and effort, the task of creating a map (just like many other tasks) can be subdivided into smaller more manageable things. creating such a project could be extremely cool because it could also create a set of mapping tools for the community to use in hand made maps. im not aware of what we have already but things i could see happening is:

basic texture painting based on cliff slope, water, and such
more intelligent reclaim/tree placement based on zones designated and not placing them on cliffs and such,
decal placement again based cliff slope, water edge, manually defined boundaries, ect.
"prettification" of cliff heightmaps and whatnot

with even more effort, since the above would be procedural you could make them more parametric and allow redefining boundaries and such, with the decals/whatever being redrawn to match.

would be an interesting but difficult project i would say.

[biass]

I had an idea to generate it from a hex system
make a few heightmap tiles, generate them randomly with a few rules and then load it into the editor with photoshop RAW, let a mapper do the rest