Artificial intelligence of non-playable characters in video games
Friberg, Jesse (2025)
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-202502112620
https://urn.fi/URN:NBN:fi:amk-202502112620
Tiivistelmä
Artificial Intelligence created by game developers for non-playable characters is one of the most important parts of developing a fully fleshed video game. Even though the subject is an important part of the industry, it is not talked about enough and documentation on the subject in general is rather lacking. The objective of this thesis was to seek out the most common solutions to creating an artificial intelligence that feels intuitive and helps the player immerse themselves in the game they are playing. The goal was to study these solutions and find out how they are used.
The method of research used in this thesis started with looking at developer logs, articles, and conferences held by game developers, and then gathering these sources to a spreadsheet. Data would then be extracted from these sources and be summarized. After enough data was collected, conclusion was drawn on which solutions were the most common and would be studied further in this thesis.
As a result of this study, it was shown that the four most common solutions used were Goal-Oriented Action Planning (technique used to create goal-oriented AI), Hierarchical Finite State Machines (nested Finite State Machines), Navigation mesh (abstract data structure), and A* (A-star, search algorithm). These four solutions were the most common and many of the other solutions were derived from one of these four which shows that the reason they are so popular is their flexibility and ability to be derived.
The method of research used in this thesis started with looking at developer logs, articles, and conferences held by game developers, and then gathering these sources to a spreadsheet. Data would then be extracted from these sources and be summarized. After enough data was collected, conclusion was drawn on which solutions were the most common and would be studied further in this thesis.
As a result of this study, it was shown that the four most common solutions used were Goal-Oriented Action Planning (technique used to create goal-oriented AI), Hierarchical Finite State Machines (nested Finite State Machines), Navigation mesh (abstract data structure), and A* (A-star, search algorithm). These four solutions were the most common and many of the other solutions were derived from one of these four which shows that the reason they are so popular is their flexibility and ability to be derived.