Static decentralized team problems: Sufficient conditions, algorithms, and the exponential cost criterion

The stationary conditions of Radner are shown under relaxed conditions to be sufficient for the global optimum of the static team problem with convex cost. This extension of Radner's theorem establishes the global optimality of the optimal affine laws for the exponential of a quadratic performance index. A formulation of the team problem as a constrained parameter optimization problem simplifies the derivation of the necessary conditions for optimal affine team laws. The solution of this constrained problem expresses the optimal decentralized team control law as an explicit projection of the optimal centralized law for both the quadratic and the exponential of a quadratic performance indices.