Procedural Generation of Video Game Cities for Specific Video Game Genres Using WaveFunctionCollapse

This research explores the application of the WaveFunctionCollapse (WFC) algorithm for the procedural generation of virtual cities tailored to specific video game genres. The primary objective is to bridge the gap between generic procedural city generation and the creation of environments that actively support specific gameplay requirements. By utilizing WFC, the study demonstrates how designers can translate small, manually created input patterns into larger, varied game levels, offering a potential framework for rapid prototyping in game development.

The methodology involves two distinct experimental phases. First, a parameter study replicates the city of Skara Brae from the classic role-playing game The Bard's Tale. By adjusting the N parameter—which dictates the size of input chunks—the study identifies how different values influence output characteristics; specifically, N=2 produces intricate, maze-like passages, while N=3 facilitates the creation of larger, open spaces. Second, a pilot user study involving eight participants evaluates how human designers approach level creation when provided with a genre-specific toolkit based on the tile set from Ultima I: The First Age of Darkness.

Key findings indicate that while WFC is effective for generating varied, locally similar environments, the algorithm does not inherently guarantee gameplay-specific constraints such as walkability or the logical placement of unique objectives. The study concludes that while WFC serves as a powerful tool for procedural content generation, it requires secondary passes to refine game-specific elements like enemy placement, building aesthetics, and objective accessibility. Future research is encouraged to explore deeper human-computer co-creativity and to expand the algorithm’s expressivity beyond current limitations to better align with human-level design intentions.