Chemistry and Code

Over the last three days I’ve been working on an early prototype of the procedural island generation system for the Vertebrae project. It’s currently still very much a work in progress and focuses more on the simple aspects of land mass generation, future versions will expand on this early work, incorporating a much larger element of actual simulation, such as erosion, rainfall, humidity and so forth. With all this data I should hopefully be able to more accurately model a living world, with well placed biomes, aiding the intelligent placement of fauna and flora within the world.

Prototyping

The first version that sort-of looked like landmasses.

The prototypes I produce for this project will generally start with little planning, beyond a rough idea of the intended final product, as a result the interface quickly evolves as time goes on, expanding to include features I need at the time, including features that won’t make it into the final product. Using this system of rapid prototyping to test out the various aspects of Vertebrae makes it very easy for me to test out ideas in a self-contained manner.

Several features, such as the ability to undo/redo generations were implemented purely because generating worlds is fun and I wanted to make it less frustrating to play with the generator. I initially added support for undo and redo after generating an analogue of North and South America, arranged correctly, and then accidentally generating a new world, overwriting it. Sigh.

Thoughts on the Future

Landmasses starting to take shape.

Having implemented the generator using cellular automata, I still have three remaining algorithms to test out, based on midpoint displacement, Perlin noise and Voronoi diagrams, respectively. From what I’ve seen it is likely I’ll end up using a system based on Voronoi diagrams for my final world generator, this would seem to give me the most realistic looking terrain and the largest amount of data to work with (whereas midpoint displacements and cellular automata essentially leave me with a mass of pixels, the Voronoi diagrams gives me a large set of internal points that make good starting points for geographical and geological features.)