Dynamic Programming with Non-Classical Information Structures
- J. Swigart and S. Lall.
- Mathematical Theory of Networks and Systems (online/DVD only), 2008.
We consider the problem of Partially Observed Markov Decision Processes with a non-classical information structure. Under a particular constraint on the information structure, optimal decision policies can be found via a dynamic programming approach. We also consider state space systems with linear dynamics and quadratic cost objectives, and provide sufficient conditions on the information structure under which optimal control policies can be found analytically.