A Novel Description of Linear Time--Invariant Networks via Structured\n Coprime Factorizations

Şerban Sabău, Cristian Oarâ, Sean Warnick, Ali JadbabaiearXiv 2013 · DOI · arXiv

Abstract

In this paper we study state-space realizations of Linear and Time-Invariant (LTI) systems. Motivated by biochemical reaction networks, Gonçalves and Warnick have recently introduced the notion of a Dynamical Structure Functions (DSF), a particular factorization of the system's transfer function matrix that elucidates the interconnection structure in dependencies between manifest variables. We build onto this work by showing an intrinsic connection between a DSF and certain sparse left coprime factorizations. By establishing this link, we provide an interesting systems theoretic interpretation of sparsity patterns of coprime factors. In particular we show how the sparsity of these coprime factors allows for a given LTI system to be implemented as a network of LTI sub-systems. We examine possible applications in distributed control such as the design of a LTI controller that can be implemented over a network with a pre-specified topology.

Introduction

In traditional control theory, systems are often treated as "black boxes" characterized by a Transfer Function Matrix (TFM). While the TFM effectively maps inputs to outputs, it obscures the internal signal flow and the specific interconnections between sub-systems. This lack of transparency becomes a significant hurdle when designing distributed controllers that must operate over pre-defined physical or communication topologies, such as vehicle platoons or complex chemical processing plants.

Previous approaches to distributed control often focused on making the TFM itself sparse (decentralized control). However, this is frequently too restrictive for practical applications. The fundamental challenge addressed in this research is: How can we design a controller that exhibits a dense input-output relationship while being internally composed of sparse, interconnected sub-units? The authors bridge this gap by connecting the "Signal Structure" of a system to the algebraic tools of Coprime Factorizations.

The Dynamical Structure Function Framework

The core of this new description lies in the Dynamical Structure Function (DSF) representation $(Q, P)$. In this framework, a system is described by the equation $Y = QY + PU$. Here, the matrix $Q$ explicitly represents the network topology—how outputs depend on one another—while $P$ represents the direct influence of inputs on those outputs.

By using DSFs, researchers can look "under the hood" of a dense transfer function to reveal a potentially sparse underlying network. This representation allows for the modeling of complex internal dependencies that are invisible in standard input-output models.

Methodology: Bridging DSFs and Coprime Factorizations

The paper establishes a formal mathematical link between DSFs and Coprime Factorizations, a cornerstone of classical control theory used for system stabilization and optimization. The methodology involves several key theoretical steps:

  1. Viable $(W, V)$ Pairs: The authors introduce a representation $L(\lambda) = (\lambda I - W)^{-1}V$, proving that for any LTI system, a class of these pairs can be parameterized to reflect the system's structure.
  2. Coprimeness Proofs: A major contribution is proving that these $(W, V)$ pairs are left coprime factorizations. This is essential because coprimeness is a prerequisite for advanced control techniques like the Youla parameterization.
  3. Sparsity-Preserving Stabilization: Initial $(W, V)$ factors may be unstable. The authors utilize a "Basic Pole Displacement" technique to transform these into stable factors $(M, N)$ while strictly preserving the original sparsity pattern of the network.
  4. Riccati Equations: Using non-symmetric algebraic Riccati equations, they demonstrate a one-to-one correspondence between stable coprime factorizations and unique DSFs.

Key Theoretical Results

  • Theorem IV.2: Formally confirms that any viable $(W, V)$ pair associated with a controllable LTI system constitutes a left coprime factorization.
  • Lemma IV.4 & Remark IV.5: Provide a constructive method to generate stable coprime factors $(M, N)$ that maintain the exact sparsity structure of the system's DSF. This implies that a system can be stabilized without altering its physical interconnection blueprint.
  • Theorem IV.9: Establishes that for almost every stable left coprime factorization, there exists a unique associated DSF. This allows engineers to perform synthesis in the coprime domain and translate the results directly back to a realizable network topology.

Discussion and Future Work

This research moves the field beyond simple decentralization toward sophisticated networked control, where sub-systems communicate in structured ways. By showing that sparsity is "hidden" within coprime factors, the authors provide a roadmap for implementing complex controllers over specific architectures.

While the paper provides sufficient conditions for these structured factorizations, identifying the necessary conditions remains an open area for research. Future work will focus on integrating these findings with a specialized version of the Youla parameterization to create a complete algorithm for synthesizing optimal controllers with pre-specified topologies from scratch.

Summary updated

Abstract

In this paper we study state-space realizations of Linear and Time-Invariant\n(LTI) systems. Motivated by biochemical reaction networks, Gon\\c{c}alves and\nWarnick have recently introduced the notion of a {\\em Dynamical Structure\nFunctions} (DSF), a particular factorization of the system's transfer function\nmatrix that elucidates the interconnection structure in dependencies between\nmanifest variables. We build onto this work by showing an intrinsic connection\nbetween a DSF and certain sparse left coprime factorizations. By establishing\nthis link, we provide an interesting systems theoretic interpretation of\nsparsity patterns of coprime factors. In particular we show how the sparsity of\nthese coprime factors allows for a given LTI system to be implemented as a\nnetwork of LTI sub-systems. We examine possible applications in distributed\ncontrol such as the design of a LTI controller that can be implemented over a\nnetwork with a pre-specified topology.\n
BibTeX
@misc{sabau2013novel,
  author = {Şerban Sabău and Cristian Oarâ and Sean Warnick and Ali Jadbabaie},
  title = {{A Novel Description of Linear Time--Invariant Networks via Structured\textbackslash\{\}n Coprime Factorizations}},
  year = {2013},
  eprint = {1303.2760},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1303.2760},
}

Loading related content...

Figure