AI-assisted Game Level Generation based on Difficulty
Loading...
URL
Journal Title
Journal ISSN
Volume Title
Perustieteiden korkeakoulu |
Master's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.
Authors
Date
2023-12-11
Department
Major/Subject
Game Design and Production
Mcode
SCI3046
Degree programme
Master’s Programme in Computer, Communication and Information Sciences
Language
en
Pages
57+0
Series
Abstract
To provide an enjoyable gaming experience, game designers need to effectively manage the difficulty of game levels. Achieving the desired difficulty often requires numerous iterations and fine-tuning, which generally consumes a massive time. This thesis focuses on investigating the integration of artificial intelligence (AI) into a procedural dungeon generation(PDG) technique to generate levels that meet specific target difficulty criteria. PDG is a subarea of procedural content generation(PCG). PCG refers to using computational methods to generate game content, and can significantly reduce the time and cost of game development by creating assets automatically. As a subfield, PDG focuses on automatic dungeon generation. However, it is hard to ensure the generated level fits the desired difficulty. Artificial intelligence(AI) can create AI players to estimate the difficulty of levels. Thus, this research aims to apply AI players to assist the PDG method to control the generated levels' difficulty. This study creates a level-generation approach for a custom game. The game developed for the study is called Hack Into Chess. It is a puzzle game that requires players to collect items on the map when some enemies chase them. The level-generation approach implements cellular automata to generate levels and optimizes the levels according to the performance of an AI player which is based on Monte Carlo tree search. The generated levels are tested by human players to collect gameplay data and actual experience data. Finally, the thesis analyzes the correlation between the performance of the AI player and human players, which shows the potential of doing playtesting and predicting players' behavior by AI. The thesis also trains a prediction model to estimate the weight of factors that influence difficulty. It also helps level designers to find suitable parameters for generating levels with the desired difficulty.Description
Supervisor
Hämäläinen, PerttuThesis advisor
Hämäläinen, PerttuKeywords
game levels, procedural dungeon generation, artificial intelligence, difficulty control