Generating narrative through intelligent agents in digital games

[摘要] An ongoing problem in game design is how to create story-based games that allow the player to have a variety of experiences if the game is played more than once, preferably without burdening the designer with a prohibitive workload. In this project, I approach the problem of creating a game with a mutable narrative from an AI perspective, designing a system called CharacterSimulator that generates a population of non-player characters (NPCs) with which the player will interact and assigns the NPCs a set of goals to carry out. Varying the set of NPCs and their objectives will create a different narrative experience for the player when the game is replayed. Although the NPC behaviors were originally modeled on Braitenberg vehicles, I have largely moved away from that model in the final version, focusing more on assigning NPCs goals that result in narratively interesting interactions.

[发布机构] Massachusetts Institute of Technology