Abstract
In recent years, while AI technology has continued to evolve, the amount of data required and the difficulty of creating an environment in which it can be executed are also increasing. This current situation is disadvantageous for individual developers and beginners who do not have much know-how or materials, and creates a large gap between them and developers who have the know-how and materials. Therefore, in this paper, we decided to research an automatic generation method for 2D action game maps, focusing on two points for individual developers and beginners: "It can be implemented with a small amount of data" and "It is easy to use even for beginners". In the course of research, we succeeded in converting a roughly arranged game map into coordinate data and using that coordinate data to generate a game map from a single map data. Among these methods, the one that was particularly excellent was the method using Markov chains, which makes it possible to generate a map that reproduces the characteristics of the original map while also having a new personality.
https://ror.org/05tc07s46 
