Automating Tower Defense Game Level Design with Evolutionary Algorithms

This paper explores the use of Genetic Algorithms (GA) and Customized Estimation of Distribution Algorithms (CEDA) to dynamically design enemy waves in a Tower Defense game. By varying population sizes (5 to 50), the study evaluates each algorithm’s efficiency and effectiveness in generating balanced and challenging enemy waves. Fitness scores are based on the severity of attacks on the player's base and the distance traveled by each monster. Results show that both GA and CEDA are capable of producing diverse and challenging gameplay experiences, with optimal performance achieved using moderate population sizes. This AI-driven approach introduces a new dimension to Tower Defense games, enabling dynamic and adaptive level design that enhances engagement and replayability.