Defining and Using Ideal Teammate and Opponent Agent Models

摘要

A common challenge for agents in multiagent systems is trying to predict what other agents are going to do in the future. Such knowledge can help an agent determine which of its current action options is most likely to achieve its goals. There is a long history in adversarial game playing of using a model of an opponent which assumes that it always acts optimally. Our research extends this strategy to adversarial domains in which the agents have incomplete information, noisy sensors and actuators, and a continuous action space. We introduce \\ideal-model-based behavior outcome prediction" (IMBBOP) which models the results of other agents' future actions in relation to their optimal actions based on an ideal world model. Our technique also includes a method for relaxing this optimality assumption. IMBBOP was a key component of our successful CMUnited-99 simulated robotic soccer application. In this paper, we dene IMBBOP and illustrate its use within the simulated robot...