The purpose of this project is to test the feasibility of game level generation using machine learning and artificial intelligence. Others have previously achieved this on a 2 dimensional scale; However, not much research has been put into AI generated content in a 3 dimensional space.
The program within this project will consist of a level editor, a method of encoding and decoding levels into a numerical format, and the game itself that will run said levels. The numerical representation of these levels will be critical for creating a ML and AI pipeline that can predict and create a playable and fun level. The interactable game available for the end user is being developed in Unity.
With the ability to manipulate game data outside of the engine, this affords us the option to use widely available methods and tools like scikit-learn and TensorFlow within Python. While other pipelines may be considered, working natively within the engine or with the project files directly is unnessesary because content will not be generated in real-time within the game.
Some of the tools in order to complete this project and accompying documentation include: