On decentralized linear stochastic control problems with quadratic cost

A decentralized stochastic optimal control problem is considered where agents have different a priori information on the system initial state. Agents are assumed to exchange their control values but not their state vector observation values. It is shown that this leads to a suboptimal control law, with correction terms being added to the well-known optimal proportional feedback control signal obtainable under the information centralization assumption of linear dynamics with quadratic cost, when person-by-person satisfactory team decisions are considered.