Geofictional Algorithms

Posted by Luciano on 22 August 2016 in Korean (한국어)

A somewhat different topic than usual.

If anyone is looking for mapping inspiration, or just out of curiosity, there is a twitter account called @unchartedatlas. It is a "bot" which produces imaginary maps "randomly" - obviously the design of the algorithm is probably pretty complex - the creator discusses it here. Some of the maps aren't very realistic, but others are remarkable.

Happy mapping.

Comment from refried sushi on 22 August 2016 at 02:06

I'm torn.

One the one hand the consistency of the product without being repetitive is wonderful. The bot has a pleasant style. On the other hand, all of these maps feel like drafts, like the bot needs another "render" pass to work out the finer topography and detail; I keep seeing straight line rivers and jagged coasts. These are not quite believable IMO but sometimes scary close.

Excellent find.

Hide this comment

Comment from dryerlint on 22 August 2016 at 03:39

Somewhat similarly, the game Dwarf Fortress is known for its in-depth worldbuilding. When you create a new region, it'll simulate terrain generation, civilization spread, climate, etc. I'm not talking about simplistic Minecraft-style "world generation". DF will actually simulate things like rain shadows and volcanoes. This page and this page should give you an idea of how deep the whole process goes.

I recommend trying out the game, it's pretty enjoyable even though the graphics and UI are awful.

Hide this comment

Comment from Szpw on 22 August 2016 at 04:15

While it's learning curve is insane and the UI is poorly designed, Dwarf Fortress has one of the most complex and detailed world generation systems I've ever seen. It literally simulates everything

Hide this comment

Comment from wangi on 22 August 2016 at 16:12

Any plans to procedurally generate terrain for the new continents? ;)

Hide this comment

Comment from isleño on 23 August 2016 at 16:15

I'd love to see something like this...

Hide this comment

Comment from Luciano on 23 August 2016 at 23:20

I think if anyone working at OGF has come close to the idea of algorithmically generating map detail, it would Paxtar and the huge project to make contours for that country. I wouldn't know where to begin on such a project - I might have some of the technical skills, but I lack the background in image manipulation that could make the output work graphically.

I think wangi and isleño are on to something, though: it seems unlikely, given the pace at which most of us work, that we will "fill" our OGF world with detail in our lifetimes, unless we develop some tools for automated (semi-automated?) mapping. I suppose the "landcover" polygons I've placed in Ardisphere are a sort of "low tech" move in that direction. I copy-paste them across the landscape, giving a pseudo-random feel to things that I can then go in and tweak fairly rapidly to get a desired landscape effect. If I was more committed to the procedure, I could make a script that did it in JOSM. But so far it's just an experiment, and my recent work in Tárrases has convinced me that I was on the wrong track.

Hide this comment

Comment from wangi on 24 August 2016 at 14:39

I had done a fair bit of digging on procedural city generation a while back, but came to the conclusion that it's not necessarily the right approach.

But I think generating the terrain is. If you have terrain to drape your imaginary city & countryside on to then it's a massive enabler to improving realism...

...Adding on land cover atop of the terrain is then a very good next step. Using slope analysis, climate rules etc to pick appropriate types.

This then leaves you a with a great starting point for adding in culture.

Hide this comment

Comment from isleño on 25 August 2016 at 02:43 seems unlikely, given the pace at which most of us work, that we will "fill" our OGF world with detail in our lifetimes, unless we develop some tools for automated (semi-automated?) mapping.

This has been on my mind too: finding faster ways to create large areas of detail. So far, my best attempt has been on the island of Sankt Theodorus in Niulutan-Riu. It's not automated at all, but still much quicker than drawing each land use shape individually.

Basically I used a stylus to sketch out the land use in Inkscape, super fast, just like drawing fields on a piece of paper. Then I translated the negative areas enclosed by the lines into polygons, saved them as .svg, imported the file into JOSM, and assigned farmland/meadow/orchard randomly en masse. It was much quicker and easier than anything I had ever tried before.

Maybe we should start a wiki page for sharing stuff like this? I'd love to see the various "shortcuts" that different people are exploring. Exchanging ideas and getting feedback might help us reach breakthroughs more quickly, and it would be great to showcase a variety of methods people might be able to choose from, to try out in their own countries.

Hide this comment

Comment from Ūdilugbulgidħū on 26 August 2016 at 11:22

Very interesting. Has anyone contacted @unchartedatlas to see if they would be interested in applying some of this to OGF themselves?

Hide this comment

Comment from isleño on 26 August 2016 at 18:54

@Udi: Not to my knowledge. Of course you're welcome to contact them if you like. It would be interesting to see how it might translate into an OGF country.

Hide this comment

Comment from Luciano on 26 August 2016 at 22:10

Based on looking at the creator's website and some of his other projects, I don't think he's particularly interested in geofiction. My impression is that he's looking at all kind of "automated" and AI-type creative pursuits: automated novel-writing, dadaesque jokebots, abstract art making programs, etc.

Then again, giving him a country (if he wanted one), it is highly unlikely to end up less realistic than some of what other users have already done.

Hide this comment

Leave a comment

Parsed with Markdown

  • Headings

    # Heading
    ## Subheading

  • Unordered list

    * First item
    * Second item

  • Ordered list

    1. First item
    2. Second item

  • Link

  • Image

    ![Alt text](URL)

Login to leave a comment