Non-player character decision-making in computer games

Abstract

One of the most overlooked challenges in artificial intelligence (AI) for computer games is to create non-player game characters (NPCs) with human-like behavior. Modern NPCs determine their actions in different situations using certain decision-making methods, enabling them to change the current state of the game world. In this paper, we survey current decision-making methods used by NPCs in games, identifying five categories. We give detailed overview of these five categories and determine the previous studies that belong to each of these categories. We also discuss the hybrid methods which are the combinations of different decision-making methods and the frameworks that are created for NPC decision-making. As a result of this analysis, we create a taxonomy table based on these covered studies. Lastly, the challenges faced in our study and future possibilities for improvement are described.