This thesis explores the use of ontological reasoning to improve artificial intelligence, AI, for non-player characters, NPCs, in video games. Faced with the limitations of traditional AI techniques in games, we propose an approach combining logic programming and ontologies.

The study begins with a state of the art review of AI methods commonly used in the video game industry, highlighting their strengths and weaknesses. We then examine the theoretical foundations of logical reasoning and ontologies, shedding light on their potential for NPC decision-making.

Our main contribution lies in the development of a knowledge engineering methodology integrating Prolog and ontologies for NPC decision-making by reasoning. This approach aims to create complex and adaptive behaviors, while having good maintainability and extensibility of the code and knowledge base.

To validate our approach, we implement and test our system in the Wumpus World game, an environment popularized by S. Russell et Norvig's book: Artificial Intelligence : A Modern Approach. We then compare the performance and quality of the behaviors obtained with those generated by traditional methods.

The results demonstrate the potential of ontological reasoning for creating complex behaviors in comparison to conventional techniques such as conditional instructions, finite state machines, or behavior trees, which suffer from scalability issues. This thesis thus contributes to the advancement of AI techniques applied to the video game domain, offering an alternative to conventional approaches.

KEYWORDS: Artificial Intelligence, Logic Programming, Ontologies, Reasoning, Non-Player Character, Video Games.

Jury :

Président du jury : Sébastien Salva
Rapportrice et rapporteur : Anne-Gwenn Bosser, Laurent Perrussel
Examinateurs : Tristan Cazenave, Florian Richoux
Encadrants : Christophe Rey, Loïc Yon, Bruno Bachelet.