Search-Based Procedural Generation for First-Person Shooter Maps

Dylan Ward and Reza Ryan Part of CW18

Over the past three decades, video games have become one of the most popular forms of entertainment in the world. This increase in popularity comes with a demand for frequent and quality content from consumers, however, delivering on this demand costs companies time and money.

The aim of this research is to identify, develop and evaluate a method of procedurally generating maps for multiplayer first-person shooters using a genetic algorithm (GA). Past research in the field of search-based procedural generation in first-person shooters (FPS) has allowed for little customization, not used evaluation techniques based on proven level design techniques and has not been verified via user testing. The objective of this research is to design an algorithm to generate maps automatically in a way that requires little manipulation from designers and test the generated maps on participants. The quality of generated maps is evaluated based on measuring the tension levels of an AI agent in a simulated match. Previous research in this field has allowed for little customization, this algorithm will be easily customizable, allowing designers to create levels of various shapes and sizes and allowing designers to use 3D asset packs for level construction. The algorithm has also been designed in such a way that it can be integrated into any real-time game engine with ease.