GRUE : an architecture for agents in games and other real-time environments

摘要: This thesis presents an architecture, which we call GRUE, for intelligent agents in real-time dynamic worlds. Such environments require to be able flexibly adjust their behaviour take into account changes the environment or other agents’ actions. Our architecture is based on work done robotics (Nilsson, 1994; Benson and Nilsson, 1995; Benson, 1996), also deals with complex, environments. focuses goal arbitration, method used by agent choose appropriate current situation, re-evaluate when situation changes. In process, have developed a representing items environment, resources, terms of properties. allows specify needed object required properties use available objects interchangeably. We show that GRUE can successfully both typical AI test bed commercial game environment. addition, undertaken experimentally effects features included our comparing using standard one more removed find these do improve performance where expected.