Terrain generation algorithms

Abstract

Procedural terrain generation has become common in games as a whole and in indie games in particular. With procedural terrain generation developers can relatively easily create static or dynamically expanding game areas. Also it is more cost effective since large part of manual work can be automated which traditional game areas would require.

Goal of this thesis is to introduce and evaluate different algorithms that are used or have potential use cases in terrain generation. Such algorithms as various noise functions, which are widely used in the realm of terrain generation, a number of dungeon algorithms, which use variety of methods to generate the dungeon, fractal algorithm, and volumetric terrain generation algorithm which uses a combination of noise and fractal algorithms. Algorithms and techniques will be searched from various scientific articles and literary sources. Metrics used for terrain generation algorithm evaluation will also be introduced, and algorithms in this thesis will be evaluated using these metrics.

During evaluation it was noticed that the evaluated noise functions are generally capable of runtime terrain generation, but are lacking in customization and control since parameters are usually related to the algorithm rather than the resulting terrain. Albeit these shortcomings both Perlin and Simplex noise stand out for their ability to generate good quality terrains. On the other hand most of the evaluated dungeon generation algorithms are incapable of generating terrain during runtime with few exceptions. Also guaranteeing connectivity of rooms or areas in dungeon can be challenge in some algorithms. The introduced fractal algorithm is metrics wise similar to Perlin and Simplex noise even though it uses completely different method to generate the terrain. The volumetric terrain generation algorithm is the only algorithm capable of generating volumetric terrain and its high level of parametrization and customization is its strongest quality.