Procedural Content Generation for 2.5D Rogue-Lite Games: An Evolutionary Algorithm Approach

This paper introduces Ropburi, a 2.5D rogue-lite game that employs procedural content generation (PCG) to enhance player experience. The game utilizes a Weight-Based Genetic Algorithm (WBGA) to create balanced, engaging maps and a (1+1) Evolution Strategy (ES) for tuning difficulty levels by adjusting enemy attributes. Our approach surpasses traditional deterministic PCG methods by incorporating genetic algorithms, providing flexibility and efficiency in generating complex game environments. Experimental results demonstrate that smaller population sizes converge faster and require lower computational time, while optimized bots ensure precise difficulty adjustments. This research highlights the balance between optimization quality and computational efficiency, setting a new standard for automated level design in rogue-lite games and showcasing the potential of evolutionary algorithms in creating dynamic, immersive gaming experiences.