Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

It looks nice. What I like about procedural generation is the great freedom in design. It's a great opportunity to tackle the problem of managing complexity in a visually rewarding way.

There is the top down approach where you start laying out the road networks, it defines neighborhoods, which you split up into individual houses. It can be made as nice as the programmatic time you are willing to invest, ultimately it's realistic, but quite repetitive and boring.

There is the scoring based approach where you get points when certain type of building are near other type of buildings. Then you optimize more or less. Quite often it's not as repetitive and boring, but not very realistic, so you can add some statistic matching penalties to make it more realistic (eventually using neural networks like GAN).

There is the more ambitious and computer-intensive bottom-up approach, where you start defining the rules of the game (like a sim-city generator) and have your city evolved by playing the game eventually from the point of view of the various agents in the game. The city will have a more organic feel and look. By defining the right rules, you can even witness some emergent behavior. You have less control but it'll give you a world deeper than it looks.



Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: