Learning models for decentralized decision making (Q1058971)

Models for decentralized decision making under uncertainty are introduced. Decision makers interact with each other and update their decisions using known learning schemes, rendering analysis of the group behavior tractable. Synchronous models are shown to lead to problems in stochastic games. Specific examples of such models are explored in detail and simulation results are given. Other situations in which only one decision maker acts at a time are also presented, and it is suggested that the decentralized control of Markov chains can be viewed in terms of such a model.