A genetic approach in procedural content generation for platformer games level creation

Abstract

In this article we used a genetic algorithm approach for generating and evaluating rhythms for creating levels of 2D runner platformer games. After generating rhythms, we used a grammar based approach to generate geometry based on these rhythms. We used a novel fitness function for the genetic algorithm in the area of PCG. This approach also minimizes the amount of the content that must be manually authored. Our results show that this method can produce a variety of levels with controlled difficulty between two levels and all generated levels are fully playable. We believe that the presented method is potentially applicable to commercial platformer games.