Parametrization of stabilizing controllers subject to subspace constraints

We consider the problem of designing optimal stabilizing decentralized controllers subject to arbitrary subspace constraints. Recent work considered structural constraints, where each part of the controller has access to some measurements but not others, and developed a parametrization of the stabilizing structured controllers such that the objective is a convex function of the parameter, with the parameter subject to a single quadratic equality constraint. Here we further show that other types of controller constraints arising in decentralized control and other areas of constrained control can be simultaneously encapsulated in this framework with one additional quadratic equality constraint.