Evolutionary generation of game levels





Procedural Content Generation, Creative Computing, Novelty Generation, Video Game Design, Genetic Algorithms, Computational Creativity


This paper outlines an approach for evolutionary procedural generation of video game content. The study deals with the automatic generation of game level designs using genetic algorithms and the development of a fitness function that describes the playability of the game level. The research explores whether genetic algorithms have the ability to produce outcomes that demonstrate characteristics that arise through human creativity, and whether these automated approaches offer any benefits in terms of time and effort involved in the design process. The approach is compared to a random method and the results show that the genetic algorithm is more consistent in finding levels; however analysis of the game levels indicates that the fitness function is not fully capturing level playability. The ability to produce playable levels decreases as the play area increases, however there is potential to produce larger maps that are both playable and arguably creative through a recombination method.




How to Cite

Connor AM, Greig TJ, Kruse J. Evolutionary generation of game levels. EAI Endorsed Trans Creat Tech [Internet]. 2018 Apr. 10 [cited 2024 Apr. 18];5(15):e4. Available from: https://publications.eai.eu/index.php/ct/article/view/1504