Q-Learning in Simulated Robotic Soccer - Large State Spaces and Incomplete Information

In this paper we show how Bayesian networks (BNs) can be used for modeling other agents in the environment. BNs are a compact representation of a joint probability distribution. More precisely we will have special attention to the problem of large state spaces and incomplete information. To test our techniques experimentally, we will consider the robotic soccer simulation. Robotic soccer clients will learn through Q-learning, a form of reinforcement learning. The longterm goal of this research is to define generic techniques that allow agents to learn in largescaled multi-agent systems.