Part of Advances in Neural Information Processing Systems 16 (NIPS 2003)
Yu-han Chang, Tracey Ho, Leslie Kaelbling
In large multiagent games, partial observability, coordination, and credit assignment persistently plague attempts to design good learning algo- rithms. We provide a simple and efficient algorithm that in part uses a linear system to model the world from a single agent’s limited per- spective, and takes advantage of Kalman filtering to allow an agent to construct a good training signal and learn an effective policy.