Decentralized control with structural and information flow constraints

In this tutorial paper we give an up to date presentation of norm optimal controller synthesis for linear and time invariant systems under information constraints. The conventional type of constraints arises from imposing the computation of the output feedback control law to be made having access to only a specified subset of measurements. This in turn translates to imposing a certain sparsity pattern on the input-output map of the controller. A novel approach is to allow the controller to be implemented as an interconnection of linear and time invariant sub-controllers, i.e. as a linear and time invariant network that routes the available measurements between the sub-controllers. The topology of the controller´s network is given as a pre-specified adjacency matrix (of a graph), which this time translates into imposing sparsity constraints on a certain left factorization of the stabilizing controller.