3D Game - Enhancing Game AI with Machine Learning - Master of Science Applied Computer Science Capstone

Aug 2, 2026

Upload by Levi

Machine learning (ML) and artificial intelligence (AI) differ in scope, with ML forming a subset of AI focused on enabling computers to learn patterns. This project examines how video games commonly label NPC behavior as “AI” without using true ML, and proposes a reinforcement-learning approach instead. A 3D capture-and-hold game was built where bots use Q-learning to learn winning strategies, observe players’ tactics, and adapt counter-strategies. The system aims to increase player challenge through data-driven bot decision-making.

LanguageEnglish

CategoryTechnology

FormatPDF

Pages24

Reading Time1 h 0 min

AccessFree to download

Abstract

Introduction

Project Goal and Game Concept

Reinforcement Learning Approach

Bot Orchestration and Ensemble Modeling

Knowledge and Development Process

Grand Valley State University ScholarWorks@GVSU | Culminating Experience Projects | Graduate Research and Creative Practice | | --- | --- | | 12-15-2024
3D Game: Enhancing Game AI with Machine Learning Nicholas William English
Follow this and additional works at: [https://scholarworks.gvsu.edu/gradprojects](https://scholarworks.gvsu.edu/gradprojects)
Part of the Computer Sciences Commons | | ScholarWorks Citation English, Nicholas William, "3D Game: Enhancing Game AI with Machine Learning" (2024) . Culminating Experience Projects. 540. [https://scholarworks.gvsu.edu/gradprojects/540](https://scholarworks.gvsu.edu/gradprojects/540) This Project is brought to you for free and open access by the Graduate Research and Creative Practice at ScholarWorks@GVSU. It has been accepted for inclusion in Culminating Experience Projects by an authorized administrator of ScholarWorks@GVSU. For more information, please contact [scholarworks@gvsu.edu](scholarworks@gvsu.edu). 3D Game: Enhancing Game AI with Machine Learning Nicholas William English A Project Submitted to GRAND VALLEY STATE UNIVERSITY In Partial Fulfillment of the Requirements For the Degree of Master of Science in Applied Computer Science School of Computing December 2024 The signatures ofthe individuals below indicate that they have read and approved the project of Nicholas William English in partial fulfillment of the requirements for the degree of Master of Science in Applied Computer Science. | Dr. Byron DeVries, Project Advisor Date | | --- | | Dr. Robert Adams, Graduate Program Director Date | Dr. Robert Adams, Unit head Date Abstract Machine learning (ML) and artificial intelligence (AI) are terms that are often used synonymously, but they are ever-so slightly different. Machine learning is really a subset of artificial intelligence and involves creating an algorithm so that a computer can learn patterns. Artificial intelligence extends machine learning with the goal to go beyond pattern recognition by having a computer that is capable of mimicking human intelligence. Video games often use the term AI in reference to the bots or non-playable characters (NPCs) that players may interact with. Generally, these bots do not actually implement AI, nor do they use ML, but there has been some movement by indie game developers to develop games that implement large language models (LLMs). LLMs utilize machine learning to generate text based on a given context. In video games, these LLMs are used to generate dynamic conversations that players have with NPCs. While this advancement may allow for players to feel more immersed in the game, thereis not as much ofa push to use machine learning to challenge the players. Current games may utilize a difficulty setting for their bots, but this is generally used to control settings related to a bot’s health, damage, and aggressiveness. This just ensures that the bots are either pushovers or brutes, but they do not utilize any form of tactics or strategies to challenge the players. One of the goals of this project was to prove that bots in video games could implement machine learning to learn how to play the game and strategize the best way to win the game. The bots will learn what strategies players employ and will develop their own tactics to counteract players. Introduction The primary goal of this project was to create a 3D game that implements machine learning. In previously completed courses, 2D games were often utilized for the course projects as it was an interesting and creative way to implement ideas taught in the classroom. To expand on the knowledge that had to be acquired to understand how to program in a 2D environment, this capstone project went further by creating a game in a 3D environment. The game itself places a player on a team with friendly bots, who are competing against another team. The objective of the two teams is to capture and hold up to five posts in the game.

FAQs

What problem does the project target in existing video game “AI” behavior?

It targets that many games use difficulty settings or scripted NPC behavior instead of real machine learning, so bots lack tactics that challenge players strategically.

How do the bots learn and improve during gameplay?

Bots use reinforcement learning with the Q-learning algorithm, learning from past mistakes and victories and developing tactics that counter players.

How does the project decide where bots should go in the game?

A team-specific orchestrator determines bot decisions, using a boosting ensemble modeling technique to encourage better outcomes for the current state.

Aug 2, 2026

Upload by Levi