The dissertation will be defended at the public meeting of the Dissertation Defence Council of Informatics Engineering in the SRA-I Meeting Hall of Vilnius Gediminas Technical University at 10 a.m. on 2 May 2024.
Procedural generation in video games refers to the automatic creation of game content, such as levels, environments, and characters, through algorithmic processes rather than manual design. This approach enables developers to achieve diverse video game scene patterns, enhancing player experiences. Multi-criteria decision-making methods are employed in procedural generation to balance multiple objectives, such as gameplay variety, aesthetics, and a fluid combination of abstract video game-level features. Neutro-sophic sets, a mathematical framework dealing with indeterminate and uncertain information, offer a way to handle ambiguous elements in procedural generation, adding a unique creative dimension to the process. The dissertation consists of an introduction, three main chapters, general conclusions, and a list of references. The first chapter performs a literature review on creative procedural generation methods for video games and formulates the dissertation’s objectives. The second chapter proposes a novel approach for procedural video game scene generation, which uses genetic algorithms, employs MCDM methods for fitness function, and models creativity-based criteria. Proposed methods include WASPAS-SVNS and CoCoSo fitness functions for the genetic algorithm, regional object morph algorithm and modelling of Gestalt design principles for the fitness functions. The third chapter evaluates, explores and presents the generated result artefacts of the proposed creative procedural generation method. The case study results show how the algorithm can increase the creative value of the generated artefacts and reduce the time for manual decision-making of creative tasks. The method reduces the number of repetitive game scene patterns and generates a significant number of unique game object layout patterns. MCDM methods and neutrosophic sets ensure the combination of fluid-conflicting criteria. Generated artefact features are easy to distinguish and do not make generated iterations chaotic by not employing every criterion identically in a single algorithm run. One generated game scene can employ more than one visual design pattern if there is a possibility in the initial genetic algorithm seed and random mutation direction. When combined for different design rules, cellular automata-based rules with local neighbourhood check agents can generate varied video game scene patterns relatively quickly. The final algorithm employs an above-average ability to generate creative value.
Doctoral dissertation readers can search via VILNIUS TECH Virtual Library.
