Treffer: Creation of an artificial intelligence system for a game with complex rules / ; Dirbtinio intelekto sistemos kūrimas žaidimui su sudėtingomis taisyklėmis.

The purpose of this project is twofold. The first objective is to design a video game AI for a computer-controlled opponent for a turn-based strategy game CubeWars. The second objective is to create an environment where this AI would be able to play against older algorithms and human players to evaluate its effectiveness. CubeWars has been in development since 2021. It is developed using Unity Engine, a comprehensive framework for game development that is available for free. The new AI system for the game was developed using machine learning methods. Because Unity programming is done in C# or JavaScript and the most popular machine learning libraries are available in Python, the Unity ML-Agents toolkit was used to accomplish this goal. ML-Agents is a third-party, open-source add-on for Unity that facilitates data exchange between the Unity game code and lifecycle methods and a neural network running in a Python environment. Several different models were trained using different parameters and evaluated in games against the benchmark script, which was the only AI algorithm used in CubeWars prior to this project.