Lyapunov and Riccati Equations from a Positive System Perspective
Dongjun Wu, Yankai Lin · 2025
Read Lyapunov and Riccati Equations from a Positive System Perspective by Dongjun Wu, Yankai Lin (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-17243.txt.
Introduction
Two classic matrix equations lie at the heart of control theory: the Lyapunov equation and the Riccati equation. Dongjun Wu and Yankai Lin read both through the lens of positive systems, systems whose variables stay nonnegative. Instead of attacking the equations head-on, they build positive systems whose long-run behavior encodes the answers. Claims about whether answers exist, and whether they are unique, then follow from positive systems theory.
For the Lyapunov case, under standard observability assumptions, the built system is strictly positive. It converges fast, in the Hilbert metric, to its Perron-Frobenius vector, which is tied to the solution. The Riccati case gives homogeneous positive systems with richer dynamics. Existence and uniqueness still follow there, but only slower, asymptotic convergence can be shown. The paper imports a fresh set of tools and intuitions into a well-trodden area.
Catalog record
| CREATOR | Dongjun Wu, Yankai Lin |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 33 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-17243
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Lyapunov and Riccati Equations from a Positive System Perspective Dongjun Wu, Yankai Lin Abstract This paper presents a new interpretation of the Lyapunov and Riccati equations from the perspective of positive system theory. We show it is possible to construct positive systems related to these equations, and then certain conclusions – such as the existence and uniqueness of solutions – can be drawn from positive systems theory. Specifically, under standard observability assumptions, a strictly positive linear system can be constructed for Lyapunov equations, leading to exponential convergence in Hilbert metric to the Perron-Frobenius vector – closely related to the solution of the Lyapunov equation. For algebraic Riccati equations, homogeneous strictly positive systems can be constructed, which exhibit more complex dynamical behaviors. While the existence and uniqueness of the solution can still be proven, only asymptotic convergence can be obtained. † † thanks: [ † † address: Department of Automatic Control, Lund University, Sweden † † address: School of Computer Science and Engineering, Wuhan Institute of Technology, Wuhan, PR China footnoteinfo]Corresponding author: Yankai Lin. This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme under grant agreement No 834142 (ScalableControl). , 1 Introduction The Lyapunov and algebraic Riccati equations play fundamental roles in control theory, underpinning applications in system analysis, model order reduction, stabilization and optimal control [ 9 , 11 , 18 , 1 , 14 ] . Their properties – such as existence, uniqueness, and numerical computation of solutions – have been extensively studied in the literature. In particular, it is well-known that Lyapunov equations for linear time-invariant systems admit explicit solutions - expressed as integrals for continuous-time systems or infinite sums for discrete-time systems. In contrast, algebraic Riccati equations rarely yield such explicit solutions, leading to a distinct treatment of these equations despite their strong connections. Recently, Pates [ 19 ] proposed viewing the Lyapunov equation as a fixed-point problem on a compact convex set, leveraging the Brouwer fixed-point theorem to establish solution existence and prove uniqueness under an observability assumption. However, this technique does not directly extend to Riccati equations. To address this, we propose a new approach to tackle Lyapunov and algebraic Riccati equations in a unified framework, namely, by associating these equations with dynamical systems, and the properties of the dynamical systems in turn characterize the solutions of these equations. Our main research finding is that both the Lyapunov and algebraic Riccati equations can be analyzed through the lens of positive systems theory – a framework that has emerged as a powerful tool in control theory [ 21 , 13 ] . To be specific, we first consider the simpler case of Lyapunov equations in both continuous-time and discrete-time. By introducing an auxiliary variable, we construct a linear matrix differential (difference) equation which is shown to be strictly positive under standard observability conditions. Following standard arguments from positive systems theory, we demonstrate exponential convergence of the system in Hilbert metric to the unique Perron-Frobenius vector, which further leads to the unique solution to the Lyapunov equation. We then apply similar techniques to algebraic Riccati equations, leading to homogeneous strictly positive systems. In such cases, we prove asymptotic convergence (in Hilbert metric) of a dynamical system to the unique solution of the algebraic Riccati equation. Our contributions are twofold: 1. We establish novel connections between positive systems theory and classical equations in control, namely Lyapunov and algebraic Riccati equations. A key technical step involves homogenizing these equations by introducing a scalar-valued linear function, which enables a unified treatment under the framework of positive systems. 2. Leveraging both linear and nonlinear positive systems theory, we characterize the solutions of Lyapunov and algebraic Riccati equations as steady states of suitably constructed positive dynamical systems. Furthermore, we analyze the convergence properties of these dynamics, thereby providing a dynamical systems perspective on solving such equations. The paper is organized as follows. In Section 2 , we recall some concepts from positive systems theory and state some useful results that will be used as the main tools in this paper. In Section 3 , we study Lyapunov equations by positive systems theory. Section 4 extends the results in Section 3 to algebraic Riccati equations. Notations: Given a set S , the set of interior points of S is denoted as {\rm Int}\,S . For a square matrix A , the trace of A is denoted as {\rm tr}(A) . Let A_{1},\cdots,A_{m} be some matrices, denote {\rm col}(A_{1},\cdots,A_{m}) as the column stack of these matrices. Given two matrices A\in\mathbb{R}^{n\times n} and C\in\mathbb{R}^{m\times n} , we say that (A,C) is observable if {\rm col}(C,CA,\cdots,CA^{n-1}) is full rank. The set of n\times n positive semi-definite matrices is denoted as \mathbb{S}_{+}^{n} , and we write X\succ Y ( X\succeq Y ) if X-Y is positive definite (semi-definite). For a square matrix A , we say that A is stable if the eigenvalues of A lie on the left open-half complex plane; we say that A is Schur stable if the eigenvalues of A lie in the open unit disk in the complex plane. For an ordinary differential equation \dot{x}=f(x) , the solution at time t with initial condition (t_{0}=0,x(0)=x_{0}) is denoted \phi(t,x_{0}) . 2 Preliminaries on Positive Systems Theory In this section, we shall introduce the necessary notations and recall basic concepts that will be used throughout the paper. We begin by defining the cones, and then, on this basis, briefly review relevant results from the theory of positive systems and fixed-point theorems. Definition 1 (Cone) Let \mathcal{V} be a finite-dimensional vector space equipped with a norm \|\cdot\| . The subset \mathcal{K}\subseteq\mathcal{V} is called a cone if the following three properties are simultaneously satisfied: (a) {\rm Int}\,\mathcal{K}\neq\emptyset ; (b) \lambda\in\mathbb{R}_{\geq 0} and x,\;y\in\mathcal{K} \Rightarrow \lambda x\in\mathcal{K} , and x+y\in\mathcal{K} ; (c) x\in\mathcal{K}\backslash\{0\} \Rightarrow -x\notin\mathcal{K} . In particular, the cone \mathcal{K} is called a normal cone if the following condition is further satisfied: (d) there exists some \delta>0 such that \|x+y\|>\delta whenever x,y\in\mathcal{K} , with \|x\|=\|y\|=1 . A partial ordering can be defined on \mathcal{K} : if x-y\in\mathcal{K} , we say x\leq_{\mathcal{K}}y . We also write x\leq y when clear from the context. Throughout this paper, we shall encounter two types of cones, namely, the positive orthant \mathbb{R}_{+}^{n}:=\{x\in\mathbb{R}^{n}:x_{i}\geq 0,\;\forall i\in\{1,\cdots,n\}\} and the set of semi-positive definite matrices \mathbb{S}_{+}^{n}=\{X\in\mathbb{R}^{n\times n}:X\succeq 0\} . The norm on \mathbb{S}_{+}^{n} is defined as the Frobenius norm \|P\|^{2}=\sum_{ij}P_{ij}^{2} , and the partial ordering is defined as X\preceq Y\Leftrightarrow X-Y\preceq 0 . Obviously, both \mathbb{R}_{+}^{n} and \mathbb{S}_{+}^{n} are normal cones Definition 2 (Hilbert metric) Given a cone \mathcal{K} and x,y\in\mathcal{K}\backslash\{0\} , let M(x/y):=\inf\{\lambda:x\leq\lambda y\} and m(x/y):=\sup\{\mu:\mu y\leq x\} . Then, the Hilbert metric on \mathcal{K} is defined as d_{\mathcal{H}}(x,y)=\ln\frac{M(x/y)}{m(x/y)}. The Hilbert metric d_{\mathcal{H}}(x,y) measures the distance between the rays \{rx;r\geq 0\} and \{ry:r\geq 0\} , in particular, d_{\mathcal{H}}(x,y)=0 if and only if x and y lie on the same ray. Hilbert metric plays a crucial role in positive system theory, for instance in consensus [ 24 , 22 ] . For more details about Hilbert metric, see [ 8 ] . Definition 3 (Order-preserving maps) An operator T:\mathcal{V}\to\mathcal{V} is called homogeneous if T(\alpha x)=\alpha T(x) for all x\in V, \alpha\in\mathbb{R} . Let