Implementing machine learning (AI) in game development with Unity

Abstract

The purpose of this thesis was to research the possibilities of machine learning in video game development and implement machine learning in a Unity environment. This thesis was self-commissioned, the author sought to improve his knowledge of machine learning and the potential implementation in video games. The thesis aims to answer the following questions: (1) What does machine learning in video games mean? (2) What will the machine learning agent replace (nonplayable character/procedurally generated content/simulation)? (3) How is machine learning implemented in Unity? The research questions were answered in the following way. First there is the theoretical framework research which covers the foundation of machine learning and why machine learning is a tool that should be investigated by game developers. The second part of the thesis is a practical implementation of machine learning using ml-agents and Unity highlighting the potential machine learning has. The result of this thesis is a theoretical framework and a practical demonstration of machine learning implemented in a video game using ML-Agents and Unity. This combination of research and demonstration proofs that machine learning has many potential use cases and is ready to be implemented in video game development. Developers should investigate implementing machine learning in their development process.