On Structured Realizability and Stabilizability of Linear Systems

We study the notion of structured realizability for linear systems defined over graphs. A stabilizable and detectable realization is structured if the state-space matrices inherit the sparsity pattern of the adjacency matrix of the associated graph. In this paper, we demonstrate that not every structured transfer matrix has a structured realization and we reveal the practical meaning of this fact. We also uncover a close connection between the structured realizability of a plant and whether the plant can be stabilized by a structured controller. In particular, we show that a structured stabilizing controller can only exist when the plant admits a structured realization. Finally, we give a parameterization of all structured stabilizing controllers and show that they always have structured realizations.

💡 Research Summary

The paper investigates the interplay between graph‑induced sparsity constraints and the classical concepts of realizability and stabilizability for linear time‑invariant (LTI) systems. A “structured realization” is defined as a state‑space model (A,B,C,D) whose non‑zero entries follow exactly the adjacency pattern of a given directed graph G. The authors first show that, contrary to the unrestricted case, not every transfer matrix that respects the graph’s sparsity can be represented by a state‑space model with the same sparsity. By introducing the notion of “structured degree,” they prove that when the transfer matrix contains independent dynamical blocks associated with different graph components, the minimal order required to embed all blocks in a single state vector exceeds the sum of the individual minimal orders, making a structured realization impossible.

Having established that structured realizability is not guaranteed, the paper then examines its consequences for controller synthesis. A “structured stabilizing controller” is a controller K(s) whose own state‑space matrices obey the same graph pattern and that renders the closed‑loop system stable. The central theorem (Theorem 3) states that the existence of such a controller is a necessary condition for the plant to admit a structured realization. The proof proceeds by contradiction: if the plant cannot be realized with the prescribed sparsity, any controller constrained to the same sparsity cannot alter the pole locations of the plant in a way that respects the graph, and thus cannot achieve stabilization. This result highlights a fundamental departure from classical centralized control, where any stabilizable plant can be stabilized regardless of implementation constraints.

When a structured realization does exist, the authors provide a complete parameterization of all structured stabilizing controllers. They adapt the Youla‑Kučera framework, imposing that the free parameter Q(s) shares the graph’s zero‑pattern. Every admissible Q(s) yields a controller K(s) that stabilizes the plant, and conversely every structured stabilizing controller can be expressed in this form. Moreover, they prove that each such K(s) possesses a state‑space realization that respects the graph, guaranteeing that the controller can be implemented in a distributed fashion without violating the underlying communication or physical interconnection constraints.

The practical implications are twofold. First, for networked systems such as smart grids, multi‑robot teams, or large‑scale sensor networks, the paper warns that attempting to design a distributed controller for a plant that lacks a structured realization is futile; the structural constraints must be verified at the modeling stage. Second, when the plant is structurally realizable, the presented parameterization enables designers to pursue performance‑oriented objectives (e.g., H₂, H∞ optimization) while automatically satisfying sparsity constraints, because any choice of Q(s) within the admissible subspace yields a controller that can be physically deployed on the given network topology.

In summary, the work introduces a new prerequisite—structured realizability—for the existence of structured stabilizing controllers, establishes a rigorous link between realizability and stabilizability under graph constraints, and delivers a constructive, all‑encompassing description of the set of admissible structured controllers. This bridges a gap between abstract control theory and the concrete realities of distributed implementation, offering both a diagnostic tool for feasibility and a synthesis framework for network‑constrained control design.