A new design for decentralized control with output feedbacks

A new approach for computing optimal output feedback gains for discrete dynamical systems is described. Each local controller optimizes a performance index defined by a proper decomposition of the global criterion, thus leading to a control law with local feedbacks plus a correcting term computed by the upper level. This coordination vector informs each agent of the effect of its own action on the other subsystems. In absence of coordination the solution is a good initialization to compute, by a gradient type algorithm involving a limited number of parameters, the best output feedbacks gains.