Sepideh Hassan-Moghaddam

Sepideh Hassan-Moghaddam, Mihailo R. Jovanović

We study an optimal control problem aimed at achieving a desired tradeoff between the network coherence and communication requirements in the distributed controller. Our objective is to add a certain number of edges to an undirected network, with a known graph Laplacian, in order to optimally enhance closed-loop performance. To promote controller sparsity, we introduce $\ell_1$-regularization into the optimal ${\cal H}_2$ formulation and cast the design problem as a semidefinite program. We derive a Lagrange dual, provide interpretation of dual variables, and exploit structure of the optimality conditions for undirected networks to develop customized proximal gradient and Newton algorithms that are well-suited for large problems. We illustrate that our algorithms can solve the problems with more than million edges in the controller graph in a few minutes, on a PC. We also exploit structure of connected resistive networks to demonstrate how additional edges can be systematically added in order to minimize the ${\cal H}_2$ norm of the closed-loop system.

Sepideh Hassan-Moghaddam, Mihailo R. Jovanović

Many large-scale and distributed optimization problems can be brought into a composite form in which the objective function is given by the sum of a smooth term and a nonsmooth regularizer. Such problems can be solved via a proximal gradient method and its variants, thereby generalizing gradient descent to a nonsmooth setup. In this paper, we view proximal algorithms as dynamical systems and leverage techniques from control theory to study their global properties. In particular, for problems with strongly convex objective functions, we utilize the theory of integral quadratic constraints to prove the global exponential stability of the equilibrium points of the differential equations that govern the evolution of proximal gradient and Douglas-Rachford splitting flows. In our analysis, we use the fact that these algorithms can be interpreted as variable-metric gradient methods on the suitable envelopes and exploit structural properties of the nonlinear terms that arise from the gradient of the smooth part of the objective function and the proximal operator associated with the nonsmooth regularizer. We also demonstrate that these envelopes can be obtained from the augmented Lagrangian associated with the original nonsmooth problem and establish conditions for global exponential convergence even in the absence of strong convexity.