A SCALABLE APPROACH TO BELIEVABLE NON PLAYER CHARACTERS IN MODERN VIDEO GAMES

Recognizing the importance of artificial intelligence to modern video games, considerable efforts have been directed towards creating more believable non player characters in games, complete with personalities, emotions, relationships, and other psychosocial elements. To date, research in the field has primarily focused on the quality of character behaviour, and not on aspects of performance and scalability. This is increasingly a problem as we strive for more complex, dynamic, and realistic behaviour in an ever-growing population of characters in a virtual game world. In this paper, we explore the performance and scalability of believable non player characters in modern video games. We propose an approach to scalability that is able to improve performance and increase the ability to support a larger quantity of characters without sacrificing the believability or quality of behaviour. We then discuss a prototype implementation of this approach, as well as experiences in experiments and simulations using this prototype. Results to date have been quite positive, and show tremendous potential for continued work in the future.