Stabilizability and lower spectral radius for linear switched systems with singular matrices
Carl P. Dettmann, Chenmiao Zhang · 2025
Read Stabilizability and lower spectral radius for linear switched systems with singular matrices by Carl P. Dettmann, Chenmiao Zhang (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-2509-17799.txt.
Introduction
Carl P Dettmann and Chenmiao Zhang ask when a switched system can be steered back to calm. Their 2025 paper deals with step-by-step linear systems that swap between matrix rules. Some of those matrices are singular, so they crush some states to zero.
The pair give a new lower bound for the calm radius that works for any matrix set. For rank-one singular cases, that radius matches a joint spectral value. They work out the two-dim case in detail, with real and complex modes. They also note when the best switch path never repeats.
Catalog record
| CREATOR | Carl P. Dettmann, Chenmiao Zhang |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 35 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-17799
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Stabilizability and lower spectral radius for linear switched systems with singular matrices Carl P. Dettmann, Chenmiao Zhang Abstract We investigate the stabilizability of linear discrete-time switched systems with singular matrices, focusing on the spectral radius in this context. A new lower bound of the stabilizability radius is proposed, which is applicable to any matrix set. Based on this lower bound, more relationships between the stabilizability radius and joint spectral subradius are established. Detailed analysis of the stabilizability radius of a special kind of two-dimensional switched system, consisting of a singular matrix and a matrix with complex eigenvalues, is presented. The Hausdorff dimensions of the parameter sets such that the stabilizability radius of these systems equals a constant are also presented. Other properties of switched systems with singular matrices are also discussed along with examples. 1 Introduction The switched system is widely studied in theory and applications. Let us mention [ 19 ] and [ 20 ] for comprehensive descriptions of formation and stability of such systems. Ref. [ 3 ] contains more recent works on continuous time switched linear systems. There are many aspects of research on switched systems. For example, Ref. [ 16 ] consider the control of switched linear systems, the Lyapunov analysis of switched systems considered in [ 4 ] and [ 5 ] and Ref. [ 2 ] is about applications of switched systems to epidemiological spread. Despite the simple form of the system itself, the stability and stabilizability of such systems are hard to determine, even for linear ones. The discrete-time linear switched system of the given matrix set \mathcal{M}\subseteq\mathbb{R}^{n\times n},\;n\geq 2 is given by x(t+1)=M_{\sigma(t)}x(t),\quad x(0)=x_{0}, (1.1) where M_{\sigma(t)}\in\mathcal{M} , \sigma(t)\in\{1,\dots,m\} and m is the cardinality of \mathcal{M} . There are many general results on the stability and stabilizability of such system (for example, see [ 20 ] ). However, we are not aware of any study focusing on the case where the matrix set \mathcal{M} contains singular matrices. In the literature, the term “singular switched system” refers to a special kind of switched system with many known results (see [ 21 ] and [ 13 ] ), which is different from the one we consider in this paper. To describe the stability and stabilizability of system ( 1.1 ), the joint spectral properties are widely used. Let \mathcal{M}^{T} denote the set that contains all M_{\sigma(t)}\cdots M_{\sigma(1)} , the joint spectral radius of system ( 1.1 ) is defined as \hat{\rho}(\mathcal{M})=\lim_{T\rightarrow\infty}\sup_{A\in\mathcal{M}^{T}}\|A\|^{1/T}, (1.2) for any norm. The joint spectral subradius is defined as \check{\rho}(\mathcal{M})=\lim_{T\rightarrow\infty}\inf_{A\in\mathcal{M}^{T}}\|A\|^{1/T}, (1.3) for any norm. A new radius proposed in [ 11 ] , called the stabilizability radius, is defined as follows. \tilde{\rho}=\sup_{x_{0}\in\mathbb{R}^{n}}\inf\{\lambda\geq 0\,|\,\exists\sigma(\cdot),c>0\;s.t.\;\|x_{\sigma}(t)\|\leq c\lambda^{t}\|x_{0}\|\,\forall t\geq 0\}, (1.4) These three radii trivially satisfy \tilde{\rho}\leq\check{\rho}\leq\hat{\rho} and describe different properties of the switched system with different terminology in the literature. In this paper, we have • A switched system is called stable if \hat{\rho}(\mathcal{M})<1 (see, for example, [ 9 ] ). • A switched system is called uniformly stabilizable or open loop stabilizable in the literature if \check{\rho}(\mathcal{M})<1 (see, for example, [ 9 ] ). • A switched system is called pointwise stabilizable if \tilde{\rho}(\mathcal{M})<1 (proposition 2.5 in [ 11 ] ). Stanford first showed that pointwise stabilizability is different from uniform stabilizability in [ 17 ] . The stabilizability radius \tilde{\rho} , first introduced in [ 11 ] , had many implicit studies beforehand, for example, Ref. [ 12 ] considered this problem in measurement scheduling and other cases, and Ref. [ 7 ] considered this problem with applications in HIV drug control. For non-singular matrix sets, \tilde{\rho} has been proved to have several lower bounds in [ 6 ] . In applications mentioned above, the matrix set could contain singular matrices. In this paper, we have a new lower bound of the stabilizability radius for any finite matrix set. In the meantime, we do a detailed analysis of \tilde{\rho} of a two dimensional switched system with singular matrices, and show how to compute the stability radius \tilde{\rho} using continued fractions and Diophantine approximation. We also show that for the case \tilde{\rho}=0 , the set of such \mathcal{M} has zero Hausdorff dimension. Other properties of systems are mentioned as they appear. Some notations are used in this paper with their well-known meanings. The area denotes the n-1 dimensional Lebesgue measure of a set in \mathbb{R}^{n} . For two functions f and g , we write f(x)\sim g(x) as x\rightarrow 0 if \lim_{x\rightarrow 0}f(x)/g(x)=1 . The notation \|\cdot\| will now specifically denote the Euclidean norm in \mathbb{R}^{n} . In section 2 , we show the \tilde{\rho} is lower bounded by \check{\rho} divided by the cardinality of \mathcal{M} and give one more situation when \tilde{\rho}=\check{\rho} . Section 3 is dedicated to the stabilizability radius of the two dimensional switched system with a singular matrix and a matrix with complex eigenvalues. We also show the Hausdorff dimension of the parameter set such that \tilde{\rho}=c,0\leq c<1 is zero. Section 4 contains several examples of switched systems with singular matrices about relationship between smallest singular value and \tilde{\rho} . One strategy of finding the optimal switching law is also mentioned. 2 Lower Bound 2.1 Lower bound formation Let A_{t} denote the product M_{\sigma(t)}\cdots M_{\sigma(1)} . Suppose s^{2}_{1}(A_{t}),\dots,s^{2}_{n}(A_{t}) are ordered eigenvalues of A^{T}_{t}A_{t} and v_{1},\dots,v_{n} are corresponding orthonormal eigenvectors, where s_{1} is smallest singular value of A_{t} . We have two nice properties of v_{i} . 1. A_{t}v_{i} are orthogonal for i=1,\dots,n . 2. \|A_{t}v_{i}\|=s_{i}(A_{t}) . Suppose s_{1},\dots,s_{\kappa} are zero singular values, then v_{1},\dots,v_{\kappa} form an orthonormal basis for \mathbf{Ker}(A_{t}) . Let (x_{1},\dots,x_{n}) denote the coordinates of x\in\mathbb{R}^{n} in terms of v_{1},\dots,v_{n} . Start from the unit sphere. If the t iteration of the unit sphere under any A_{t} is a subset of a ball centered at the origin, we know that \|x(t)\|\leq\lambda\|x_{0}\| , where \lambda is the radius of the ball. Let B_{r} denote the ball with radius r centered at the origin and S^{n-1} denote the n dimensional unit sphere. We consider the intersection of the pre-image of A_{t} and S^{n-1} , i.e. S_{r,A_{t}}=\{x\in S^{n-1}|\|A_{t}x\|\leq r\} . As A_{t}v_{j}=0 for j=1,\dots,\kappa , then for any x\in S_{r,A_{t}} , we have \|A_{t}x\|^{2}=\|\sum_{i=1}^{n}x_{i}A_{t}v_{j}\|^{2}=\sum_{i=\kappa+1}^{n}s^{2}_{i}(A_{t})x_{i}^{2}\leq r^{2}. (2.1) For any nonzero A_{t} , consider the set S^{\prime}=\{x\in S^{n-1}||x_{n}|\leq r/s_{n}(A_{t})\} . Since \sum_{i=\kappa+1}^{n}s^{2}_{i}(A_{t})x_{i}^{2}\geq s^{2}_{n}(A_{t})x_{n}^{2}, (2.2) we have S_{r,A_{t}}\subseteq S^{\prime} , and S^{\prime} is a spherical segment. The area of S^{\prime} can be obtained using the area of spherical caps. A formula for area of caps is given by Lemma 2.1 . Lemma 2.1 . ((1) in [ 14 ] )The surface area of the cap of the unit sphere S^{n-1} , i.e. T_{n}(h)=\{x\in S^{n-1}|x_{1}\geq h\} with x_{1} be one of coordinates of x , is |T_{n}(h)|=\frac{1}{2}|S^{n-1}|I\left(1-h^{2};\frac{n-1}{2},\frac{1}{2}\right), (2.3) where I(h;\cdot,\cdot) denotes the regularized incomplete beta function, defined as I(h;a,b)=\frac{\int_{0}^{h}t^{a-1}(1-t)^{b-1}\text{d}t}{\int_{0}^{1}t^{a-1}(1-t)^{b-1}\text{d}t},\quad 0<h<1. (2.4) A useful equality of I(h;\cdot,\cdot) is given by I\left(h;\frac{n-1}{2},\frac{1}{2}\right)=1-I\left(1-h;\frac{1}{2},\frac{n-1}{2}\right). (2.5) One can prove this by changing variables in the integral definition of I(h;\cdot,\cdot) . Lemma 2.2 . The area of S_{r,A_{t}} is upper bounded by |S_{r,A_{t}}|\leq|S^{n-1}|I\left(\frac{r^{2}}{s^{2}_{n}\left(A_{t}\right)};\frac{1}{2},\frac{n-1}{2}\right). (2.6) For any t\in\mathbb{N}^{+} . Proof. The area of S^{\prime} is given by |S^{\prime}|=|S^{n-1}|-2\left|T_{n}\left(\frac{r}{s_{n}\left(A_{t}\right)}\right)\right|. (2.7) Hence |S_{r,A_{t}}|\leq|S^{n-1}|\left(1-I\left(1-\frac{r^{2}}{s^{2}_{n}\left(A_{t}\right)};\frac{n-1}{2},\frac{1}{2}\right)\right)=|S^{n-1}|I\left(\frac{r^{2}}{s^{2}_{n}\left(A_{t}\right)};\frac{1}{2},\frac{n-1}{2}\right). (2.8) Thus we have a bound of |S_{r,A_{t}}| . ∎ Theorem 1 . The stabilizability radius satisfies \frac{\check{\rho}}{m}\leq\tilde{\rho}\leq\check{\rho}. (2.9) Proof.