Optimal Sparse Output Feedback Control Design: a Rank Constrained Optimization Approach
We consider the problem of optimal sparse output feedback controller synthesis for continuous linear time invariant systems when the feedback gain is static and subject to specified structural constraints. Introducing an additional term penalizing the number of non-zero entries of the feedback gain into the optimization cost function, we show that this inherently non-convex problem can be equivalently cast as a rank constrained optimization, hence, it is an NP-hard problem. We further exploit our rank constrained approach to define a structured output feedback control feasibility test with global convergence property, then, obtain upper/lower bounds for the optimal cost of the sparse output feedback control problem. Moreover, we show that our problem reformulation allows us to incorporate additional implementation constraints, such as norm bounds on the control inputs or system output, by assimilating them into the rank constraint. We propose to utilize a version of the Alternating Direction Method of Multipliers (ADMM) as an efficient method to sub-optimally solve the equivalent rank constrained problem. As a special case, we study the problem of designing the sparsest stabilizing output feedback controller, and show that it is, in fact, a structured matrix recovery problem where the matrix of interest is simultaneously sparse and low rank. Furthermore, we show that this matrix recovery problem can be equivalently cast in the form of a canonical and well-studied rank minimization problem. We finally illustrate performance of our proposed methodology using numerical examples.