Abstract
Generating immersive game content is one of the ultimate goals for a game designer. This goal can be achieved taken into account that players’ perceptions
of the same game differ according to a number of factors including: players’
personality, playing styles, expertise and cultural background. One promising
avenue towards optimizing the gameplay experience for individual game players
- and thereby attempt to close the affective loop in games - is to automatically
tailor the game content in real-time. To realize player-driven procedural content generation one needs to specify the aspects of the game that have a key influence on the gameplay experience, identify the relationship between these aspects and player experience and define a mechanism for tailoring the game content to each individual needs.
In this dissertation we attempt to address the following research questions towards the aim of generating personalized content for the player: how can we
measure player experience, how can we represent game content, playing style
and the in-game interaction, what features should be used to capture player experience and how can they be extracted, how can we model the unknown function between game content, player behavior and affect, how can we generate game content that is tailored to particular player needs and style, how often game content should be adapted, and how the adaptation mechanism can be tested?
We focus on 2D platform game genre as a testbed for our player-driven procedural content generation framework. Crowd-sourcing experiments are designed to collect gameplay data, subjective and objective indicators of experience from human players: three datasets differing in the number of participants and the types of features collected and analyzed. Computational models of player experience are built on game content, gameplay, and visual reaction features capturing various aspects of the in-game interaction. Different forms of representation are considered for capturing frequencies, temporal and spatial content and behavioral events.
As soon as models of player experience are built, a real-time adaptation framework is designed which is guided by the models. The models are used as heuristics in the search of personalized content. Two adaptation mechanisms were tested in this thesis: the first is based on exhaustive search while genetic search is employed in the second. The mechanisms were tested with artificial agents and humans players.
The key findings of the thesis demonstrate the ability of the player-driven procedural content generation framework to recognize playing behavior differences and to generate player-centered content that optimizes particular aspects of player experience.
of the same game differ according to a number of factors including: players’
personality, playing styles, expertise and cultural background. One promising
avenue towards optimizing the gameplay experience for individual game players
- and thereby attempt to close the affective loop in games - is to automatically
tailor the game content in real-time. To realize player-driven procedural content generation one needs to specify the aspects of the game that have a key influence on the gameplay experience, identify the relationship between these aspects and player experience and define a mechanism for tailoring the game content to each individual needs.
In this dissertation we attempt to address the following research questions towards the aim of generating personalized content for the player: how can we
measure player experience, how can we represent game content, playing style
and the in-game interaction, what features should be used to capture player experience and how can they be extracted, how can we model the unknown function between game content, player behavior and affect, how can we generate game content that is tailored to particular player needs and style, how often game content should be adapted, and how the adaptation mechanism can be tested?
We focus on 2D platform game genre as a testbed for our player-driven procedural content generation framework. Crowd-sourcing experiments are designed to collect gameplay data, subjective and objective indicators of experience from human players: three datasets differing in the number of participants and the types of features collected and analyzed. Computational models of player experience are built on game content, gameplay, and visual reaction features capturing various aspects of the in-game interaction. Different forms of representation are considered for capturing frequencies, temporal and spatial content and behavioral events.
As soon as models of player experience are built, a real-time adaptation framework is designed which is guided by the models. The models are used as heuristics in the search of personalized content. Two adaptation mechanisms were tested in this thesis: the first is based on exhaustive search while genetic search is employed in the second. The mechanisms were tested with artificial agents and humans players.
The key findings of the thesis demonstrate the ability of the player-driven procedural content generation framework to recognize playing behavior differences and to generate player-centered content that optimizes particular aspects of player experience.
Originalsprog | Engelsk |
---|
Forlag | IT-Universitetet i København |
---|---|
Antal sider | 270 |
ISBN (Trykt) | 978-87-7949-285-1 |
Status | Udgivet - 2013 |
Navn | ITU-DS |
---|---|
Nummer | 90 |
ISSN | 1602-3536 |