Detectability via observability in a nonuniform framework: dual relationship with controllability and stabilizability

Ignacio Huerta, Pablo Monzón · 2025

Read Detectability via observability in a nonuniform framework: dual relationship with controllability and stabilizability by Ignacio Huerta, Pablo Monzón (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-2508-09918.txt.

Introduction

Ignacio Huerta and Pablo Monzón study observability in control systems in this 2025 math paper. Their systems change over time and are not uniform, which makes standard tools weak. They define a new property called nonuniform complete observability.

The authors prove a dual link between this property and nonuniform complete controllability. That link is the bridge to their main result. It shows that the new observability property is enough to ensure detectability with exponential decay. Detectability matters because it lets a system's hidden state be tracked from its outputs. The work extends classic control theory to a broader, time-varying setting.

Catalog record

CREATORIgnacio Huerta, Pablo Monzón
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE56 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-09918

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Detectability via observability in a nonuniform framework: dual relationship with controllability and stabilizability Ignacio Huerta, Pablo Monzón Abstract In this paper we propose a new observability property for nonautonomous linear control systems in finite dimension: the nonuniform complete observability, which is more general than the uniform complete observability. A dual relationship is established between this new notion of observability and the recently defined nonuniform complete controllability property, with the aim of obtaining the main result, which proves that nonuniform complete observability guarantees the nonuniform exponential detectability, a concept related to exponential stability in this framework. 1 Introduction In this work we deal with linear time-varying systems of the form \displaystyle\dot{x}(t)=A(t)x(t)+B(t)u(t), (1.1a) \displaystyle y(t)=C(t)x(t)+D(t)u(t), (1.1b) where x(t)\in\mathbb{R}^{n} is the state of the system, u(t)\in\mathbb{R}^{p} is the input , y(t)\in\mathbb{R}^{m} is the output and A(t),B(t),C(t),D(t) are measurable, bounded on finite intervals, matrix valued functions from \mathbb{R} to M_{n\times n}(\mathbb{R}) , M_{n\times p}(\mathbb{R}) , M_{m\times n}(\mathbb{R}) and M_{m\times p}(\mathbb{R}) respectively. On the one hand, we present new results on observability, i.e., the ability of the system to infer in finite time the actual or the initial state, through the observation of the output and the input, while on the other hand we present a result on detectability, which allows us to design an asymptotically estimator of the state. The article focus on nonuniform time-varying linear systems and extends the original results of Rudolph Kalman [ 12 ] , introducing a new class of systems: nonuniform completely observable systems. In a previous work we explore a new controllability property [ 8 ] and we related it with the classical Kalman definitions and new results coming from the recent qualitative theory of linear nonautonomous differential equations: the bounded growth property and the dichotomies theory [ 15 ] . In that work, we were able to present the Nonuniform Complete Controllability property which is more general than the classical Uniform Complete Controllability and less restrictive than simple Complete Controllability [ 12 ] . Besides, we characterized this new concept in terms of Gramians and we presented some non-trivial examples. In this manuscript we extend those ideas to the observability problem, using the adjoint and dual systems of the original one. We recall the following notation which will be used in this paper. The inner product of two vectors x,y\in\mathbb{R}^{n} is denoted by \langle y,x\rangle=x^{T}y and the Euclidean norm of a vector will be denoted by |x|=\sqrt{\langle x,x\rangle} . A symmetric matrix M=M^{T}\in M_{n}(\mathbb{R}) is semi–positive definite if x^{T}Mx=\langle Mx,x\rangle\geq 0 for any real vector x\neq 0 , and this property will be denoted as M\geq 0 . In case that the above inequalities are strict, we say that M=M^{T} is positive definite. Given two matrices M,N\in M_{n}(\mathbb{R}) , we write M\leq N if N-M\geq 0 . 1.1 Observability Observability analysis of ( 1.1a )-( 1.1b ) can be reduced to the study of the simpler system \displaystyle\dot{x}(t)=A(t)x(t), (1.2a) \displaystyle y(t)=C(t)x(t), (1.2b) and the mapping from the state to the output and its inverse, which allows the possibility of recovering the state from the output [ 6 ] . We formalize some definitions. The system ( 1.2a )-( 1.2b ) is observable at time t_{0} if and only if there exists t_{f}>t_{0} such that M(t_{0},t_{f})>0 . In addition, it is completely observable (CO) if and only if for any t_{0} , there exists t_{f}>t_{0} such that M(t_{0},t_{f})>0 . We refer the reader to [ 4 , 12 , 14 ] for a detailed description. For linear time-invariant systems, observability is equivalent to the complete rank of the observability matrix [ 12 ] : \mathcal{O}=\left[\begin{array}[]{c}C\\ CA\\ CA^{2}\\ \vdots\\ CA^{n-1}\end{array}\right]\in M_{nm\times n}(\mathbb{R}) and also to the invertibility of the observability Gramian M(t_{0},t_{1})=\displaystyle\int_{t_{0}}^{t_{1}}e^{A^{T}(s-t_{0})}C^{T}Ce^{A(s-t_{0})}\;ds, for any value t_{1}>t_{0} . The former idea can not be directly extended to the time-varying case, but the latter leads to the invertibility of the general observability Gramian M(t_{0},t_{f})=\displaystyle\int_{t_{0}}^{t_{f}}\Phi_{A}^{T}(s,t_{0})C^{T}(s)C(s)\Phi_{A}(s,t_{0})\;ds. (1.3) where \Phi_{A}(s,t_{0}) is the transition matrix of ( 1.2a ). A concept that is more restrictive than complete observability, and which in turn implies it, corresponds to the case where t_{0}=t and t_{f}=t+\sigma , for a fixed \sigma>0 , which corresponds to the uniform complete observability (UCO), whose formal definition will be given later (see [ 11 , Definition 3.3.3] and [ 17 , Section 1.5] and [ 18 , Section 2] for more detailed references). 1.2 Detectability The study of detectability for linear systems has a long tradition, closely tied to the design of observers and the stabilization of dynamical systems via output feedback. In the time-invariant setting, detectability is fully characterized by algebraic conditions and its equivalence to the existence of an exponentially convergent observer is well established. However, when moving to linear time-varying systems, the problem becomes significantly more intricate: the explicit time dependence of the dynamics demand a reformulation of both observability and detectability in terms of the state-transition operator and associated Lyapunov and Gramian criteria. By considering the system ( 1.2a )-( 1.2b ), the exponential detectability consists of guaranteeing the existence of an observer L:\mathbb{R}\to M_{n\times p}(\mathbb{R}) described by the equation \dot{\hat{x}}(t)=A(t)\hat{x}(t)+L(t)[y(t)-C(t)\hat{x}(t)], such that the linear system \dot{e}(t)=[A(t)-L(t)C(t)]e(t) (1.4) will be exponentially stable, with e(t)=x(t)-\dot{\hat{x}}(t) being the estimation error (see [ 16 ] ). Given the nature of different stability behaviors in the time-varying framework, uniform exponential detectability is defined based on the uniform exponential stability of the system ( 1.4 ) (see [ 21 ] ). For further details on the notion of detectability in this context, we refer the reader to [ 1 , 22 ] . 1.3 Novelty of this work In this paper we introduce and characterize the concept of nonuniform complete observability (NUCO), which is more general than the classical Kalman property of uniform complete observability, but more restrictive than complete observability. We also present the corresponding nonuniform exponential detectability idea (NUED), related to the existence of an nonuniformly exponentially stable observer. The first novel result corresponds to Theorems 5 and 6 , in which, based on the concepts of adjoint and dual systems, a close relationship on nonuniform complete controllability and nonuniform complete observability is established, allowing to flow from one to the other and their respective consequences. Then, Theorem 7 follows directly, extending a well-known uniform property related to the observability. The second novel result of this work is Theorem 9 , which states that NUCO implies NUED. Several intermediate and auxiliary results complete the work, including examples of NUCO systems and a counterxample of the reciprocal of Theorem 9 . 1.4 Structure of the article Section 2 reviews the main results on uniform and nonuniform complete controllability, including the Gramian characterization and the relationship with the bounded growth property. In Section 3 , the idea of complete observability is introduced, in both uniform and nonuniform senses. After a brief review on dual and adjoint systems, the main relationships between nonuniform controllability and nonuniform observability are presented. Section 4 presents some examples showing that nonuniform complete observability is effectively an intermediate stage between uniform complete observability and complete observability. Nonuniform exponential detectability is addressed in Section 5 , along with the idea of nonuniform exponential stability. Several auxiliary results are presented here and are used in Section 6 , where the main result

Read the complete text