ProceduralGeneration

My Bachelor's Thesis

PCG Toolkit

For my Bachelor's thesis I built a modular toolkit for procedurally generating 2D tile-based maps, and ran a scientific comparison of three generation methods:

The backend runs in Go, with an interactive browser frontend in JavaScript. Key metrics — entropy, clustering, autocorrelation and frequency distribution — are analysed in real time.

PCG Toolkit interface

To keep the methods comparable, every algorithm works on standardised input (25×25 grids, 8 terrain types, fixed seeds), evaluated across seven quantitative metrics for diversity, coherence and structural complexity.

Perlin noise scale comparison

Each method has its strengths: MLCA offers fine control, Perlin produces natural transitions, and WFC yields the most varied results — at a higher computational cost. Hybrid combinations show strong potential.

Wave Function Collapse output

The result is a reproducible framework for evaluating and comparing procedural generators, with visual output, interactive tooling and full documentation.

Conway's Game of Life