arXiv reader

Dual Representations and Proxy Characterization of H_2 Norm for Partial Integral Equations

Danilo Braghini, Tristan Thomas, Sachin Shivakumar, Javad M. Velni and Matthew M. Peet

arXiv
2610.21355v1 · PDF · 15 October 2026
Categories
math.OC
Licence
CC BY 4.0 The site shows the text.

Abstract

Duality and dual representations are essential tools for systems theory and controller synthesis. For any finite-dimensional causal linear time-invariant (CLTI) system, there exists a dual CLTI system with identical H_\infty and H_2 norm. These relations are found by simply taking the adjoint of the transfer function. For nonlinear, time-varying or infinite-dimensional systems, however, such a transfer function may not exist or be difficult to obtain. For duality in the H_\infty norm, this obstacle is overcome by using the equivalent time-domain, induced L_2-norm characterization. For H_2 norms there is no such equivalent characterization and hence proxy norms such as induced energy to peak and impulse to energy norms are used. These proxy norms bound the H_2 norm in the CLTI case. However, neither of these H_2 proxy norms equal the same H_2 proxy of the canonical dual representation. In this paper, we overcome this by proposing a dual representation whose induced energy to peak proxy equals the impulse to energy proxy of the primal and vice-versa. This relation is developed in the context of Partial Integral Equation (PIE) representations. The result is then used to establish primal-dual operator inequalities for both proxy norms similar in form to the primal-dual LMIs used to characterize the H_2 norm.

Items 29

Notation

Notation

0YRM

We will use uppercase calligraphic font \mathcal{A} for operators. Given Hilbert spaces X and Y, \mathcal{L}(X,Y) denote the set of bounded linear operators from X to Y. We denote the Hilbert adjoint by \mathcal{A}^* \in \mathcal{L}(Y,X). L_{\infty}^{n}[\Omega], and L_2^n[\Omega] are the spaces of \mathbb{R}^n–valued Lebesgue bounded, and Lebesgue square-integrable functions, respectively, on the domain \Omega, with the standard inner product. W^{n}_d[\Omega] is the Sobolev space of d times L_2–differentiable \mathbb{R}^{n}–valued functions on \Omega.

Definition

Definition 1

HXG1

Given a matrix P and polynomials Q_1,Q_2,R_0, R_1, R_2, we say that \mathcal{P} is a 4-PI operator if

\begin{aligned}\@setfontsize{\@setfontsize{\@setfontsize{\@setfontsize{\@setfontsize{\@setfontsize{\@setfontsize{\@setfontsize{\small}{10}{12pt}}{10}{12pt}}{10}{12pt}}{10}{12pt}}{10}{12pt}}{10}{12pt}}{10}{12pt}}{10}{12pt} &\left(\mathcal{P} \begin{bmatrix} x\\\mathbf{x} \end{bmatrix} \right)(s) := \begin{bmatrix} Px + \int_{a}^{b}Q_1(\theta)\mathbf{x}(\theta)d\theta\\Q_2(s)x+ \left(\mathcal{R}\mathbf{x}\right) (s) \end{bmatrix}, \text{where}~\left(\mathcal{R}\mathbf{x}\right)(s)\\ &\hspace{-0.5mm}=\hspace{-0.5mm}R_0(s) \mathbf{x}(s) +\hspace{-1.5mm}\int_{a}^s \hspace{-1.5mm}R_1(s,\theta)\mathbf{x}(\theta)d \theta+\hspace{-1.5mm}\int_s^b \hspace{-1.5mm}R_2(s,\theta)\mathbf{x}(\theta)d \theta. \end{aligned}

We denote the map from P,Q_i,R_i \mapsto \mathcal{P} as

\mathcal{P}= \hspace{0.5mm}\scalebox{0.9}[1.2]{$\mathit{\Pi}$}\hspace{-1.mm}\left[\begin{array}{c|c} P&Q_1\\\hline Q_2 & \left\{R_i\right\} \end{array}\right]\subset \mathcal{L}(\mathbb{R}^m \times L_2^{n},\mathbb{R}^p \times L_2^{q})

and, \mathcal{P} \in \pmb{\Pi} if \mathcal{P} is a 4-PI operator.

Definition

Definition 2

9D2G

Given \mathcal{T},\mathcal{A},\mathcal{B},\mathcal{C},\mathcal{H} \in \pmb{\Pi} and v, w, we say z=G(\mathcal{T},\mathcal{A},\mathcal{B},\mathcal{C},\mathcal{H})(w,v) if there exists x(t) such that

\begin{aligned} \!\partial_t( \mathcal T x(t)) \!= \!\mathcal A x(t)\!+\!\mathcal{B}w(t), \mathcal T x(0) \!=\! \mathcal{H} v, z(t) = \mathcal Cx(t), \end{aligned}
Note

Note

9WR9

The operators \mathcal{T},\mathcal{A},\mathcal{B},\mathcal{C},\mathcal{H} determine the dimensions of v,w,z,x. In this paper, however, we presume v(t),w(t),z(t) are real vector-valued functions

Definition

Definition 3

SKRV

Given \mathcal{T,A,B, C, H}\in\pmb{\Pi}, \{x,z\} is a solution of G(\mathcal{T},\mathcal{A},\mathcal{B},\mathcal{C},\mathcal{H}) if \mathcal{T}x is Fréchet differentiable, and \{z, x, w,v\} satisfy Eqn. (2).

Definition

Definition 4

CMT6

We say G(\mathcal{T},\mathcal{A},\mathcal{B},\mathcal{C},\mathcal{H}) is well-posed if, for every w\in L_2 and v \in \mathbb{R}^{n_v}, there exists a solution \{x,z\} of G(\mathcal{T},\mathcal{A},\mathcal{B},\mathcal{C},\mathcal{H}).

Definition

Definition 5 (Induced impulse-to-energy norm)

9GOP

Given \mathcal{T,A,B,C} \in \pmb{\Pi} with G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B}) well-posed, define

\|G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B})\|_{2,i}:= \sup_{T\ge 0} \sup_{\substack{z,\mathbf{x}\, \text{satisfy~\eqref{eq:primal_aux}}\\ \left\lVert{v}\right\rVert_2=1}} \left\lVert{z}\right\rVert_{L_2[0,T]}.
Definition

Definition 6 (Induced energy-to-peak norm)

MPVN

Given \mathcal{T,A,B,C} \in \pmb{\Pi} with G(\mathcal{T},\mathcal{A},\mathcal{B},\mathcal{C},0) well-posed, define

\|G(\mathcal{T},\mathcal{A},\mathcal{B},\mathcal{C},0)\|_{\infty,2}=\sup_{T >0} \sup_{\substack{v,x\, \text{satisfy~\eqref{eq:dualPIE}}\\\left\lVert{v}\right\rVert_{L_2[0,T]}=1} } \|z (T)\|.
Proposition

Proposition 7

P7U9

Given \mathcal{T,A,B, C}\in\pmb{\Pi}, where the domain of \mathcal{B} is \mathbb{R}^{n_w}, and the image of \mathcal{C} is \mathbb{R}^{n_z}, suppose \hat G(s)=\mathcal{C}(s\mathcal{T}-~\mathcal{A})^{-1}\mathcal{B} exists.

Then we have the following.

\hspace{-2mm} \|G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B})\|_{2,i} \le\hspace{-1mm} \|\hat G\|_{H_2}\hspace{-1mm}\le \sqrt{n_w} \|G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B})\|_{2,i}
\hspace{-3mm} \|G(\mathcal{T},\mathcal{A},\mathcal{B},\mathcal{C},0)\|_{\infty,2} \le\hspace{-1mm} \|\hat G\|_{H_2}\hspace{-1mm}\le \sqrt{n_z} \|G(\mathcal{T},\mathcal{A},\mathcal{B},\mathcal{C},0)\|_{\infty,2}
Lemma

Lemma 8

6FAY

Given \mathcal{T,A,B,C} \in \pmb{\Pi}, suppose x,v satisfy Eqn. (?) for \{\mathcal{T},\mathcal{A},\mathcal{B},\mathcal{C}\}, and \bar x,\bar v satisfy Eqn. (?) for \{\mathcal{T}^*, \mathcal{A}^*,\mathcal{C}^*,\mathcal{B}^*\}. Then, for any T>0, we have the following

\begin{aligned} \bigl(\mathcal{B}^* \bar x(T) \bigr)^T v= \int_0^T\left\langle \bar v(T-t),\mathcal{C} x(t)\right\rangle dt. \end{aligned}
Used by Proof
Proof of Lemma 8

Suppose x,v satisfy Eqn. (?), and \bar v,\bar x satisfy Eqn. (?). Define V(t) =\left\langle \bar x(t),\mathcal T x(T-t) \right\rangle for t \in [0,T]. Then

\begin{aligned} \dot V(t)&=\left\langle \mathcal T^*\dot{\bar x}(t),x(T-t)\right\rangle - \left\langle \bar x(t),\mathcal T\dot x(T-t)\right\rangle \\ &= \left\langle \mathcal A^*\bar x(t) +\mathcal C^*\bar v(t),x(T-t)\right\rangle- \left\langle \bar x(t),\mathcal A x(T-t)\right\rangle\\ &= \left\langle \bar v(t), \mathcal{C}x(T-t) \right\rangle. \end{aligned}

Hence, integrating in t over [0,T] we have

\begin{aligned} &V(T)-V(0)=\int_0^T\left\langle \bar v(t), \mathcal{C}x(T-t) \right\rangle dt\\ &=-\int_T^0\left\langle \bar v(T-s), \mathcal{C}x(s) \right\rangle ds=\int_0^T\left\langle \bar v(T-s), \mathcal{C}x(s) \right\rangle ds \end{aligned}

where we used the change of variables s=T-t. Now from Eqn. (?), we have \mathcal{T} x(0)= \mathcal B v and from Eqn. (?), we have \mathcal{T}^*\bar x(0)=0. Thus

\begin{aligned} &\bigl(\mathcal{B}^* \bar x(T) \bigr)^T v=\left\langle \bar x(T),\mathcal{B} v\right\rangle=\left\langle \bar x(T),\mathcal{T} x(0)\right\rangle\\& =\left\langle \bar x(T),\mathcal{T} x(0)\right\rangle-\left\langle \mathcal{T}^*\bar x(0),x(T)\right\rangle\\ &=\left\langle \bar x(T),\mathcal{T} x(0)\right\rangle-\left\langle \bar x(0),\mathcal{T}x(T)\right\rangle=V(T)-V(0)\\ &=\int_0^T\left\langle \bar v(T-t),\mathcal{C} x(t)\right\rangle dt \end{aligned}

as desired.

Theorem

Theorem 9

2ZX9

Given \mathcal{T,A,B,C}\in \pmb{\Pi}, let G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B}) and G(\mathcal{T}^*,\mathcal{A}^*,\mathcal{C}^*,\mathcal{B}^*,0) be well-posed. Then we have that

\|G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B})\|_{2,i} = \|G(\mathcal{T}^*,\mathcal{A}^*,\mathcal{C}^*,\mathcal{B}^*,0)\|_{\infty,2}.
Used by Proof, Proof, Proof
Proof of Theorem 9

Recall that

\|G(\mathcal{T}^*,\mathcal{A}^*,\mathcal{C}^*,\mathcal{B}^*,0)\|_{\infty,2}= \sup_{T>0} \sup_{\substack{v,x\, \text{satisfy~\eqref{eq:dualPIE}}\\\left\lVert{v}\right\rVert_{L_2[0,T]}=1} } \|\mathcal{B}^* x(T)\|,
\text{and }\;\|G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B})\|_{2,i}:= \sup_{T>0} \sup_{\substack{v,z,\mathbf{x}\text{satisfy~\eqref{eq:primal_aux}}\\ \left\lVert{v}\right\rVert_2=1}} \left\lVert{z}\right\rVert_{L_2[0,T)}.

To show that \|G(\mathcal{T}^*,\mathcal{A}^*,\mathcal{C}^*,\mathcal{B}^*,0)\|_{\infty,2}\le \|G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B})\|_{2,i}, we establish that for any T>0 and any \bar{v}, \bar{x} which satisfy satisfy Eq. (5) with \left\lVert{\bar{v}}\right\rVert_{L_2[0,T]}=1, we have that |\mathcal{B}^* x(T)|\le \|G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B})\|_{2,i}.

Specifically, for T>0 and any \bar{v}, \bar{x} which satisfy Eq. (5) with \left\lVert{\bar{v}}\right\rVert_{L_2[0,T]}=1, let v = \frac{1}{\left\lVert{\mathcal{B}^*\bar{x}(T)}\right\rVert}\mathcal{B}^*\bar{x}(T) and x,z satisfy Eq. (2) for this v. Then from Lem. 8,

\begin{aligned} \left\lVert{\mathcal{B}^*\bar{x}(T)}\right\rVert &= (\mathcal{B}^*\bar{x}(T))^Tv= \int_0^T\left\langle v(T-t),\mathcal{C}x(t)\right\rangle dt\\ &\hspace{-1.5cm}\le\! \left\lVert{v}\right\rVert_{L_2[0,T]}\!\left\lVert{z}\right\rVert_{L_2[0,T]}\! =\!\left\lVert{z}\right\rVert_{L_2[0,T]}\!\le\! \|G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B})\|_{2,i} \end{aligned}

as desired, where the final inequality holds from the definition of \|G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B})\|_{2,i} and the fact that v,x,z satisfy Eq. (2) with \left\lVert{v}\right\rVert=1. Since the bound holds uniformly for any T>0 and any \bar{v}, \bar{x} which satisfy Eq. (5) with \left\lVert{\bar{v}}\right\rVert_{L_2[0,T]}=1, we conclude that

\begin{aligned} \|G(\mathcal{T}^*,\mathcal{A}^*,\mathcal{C}^*,\mathcal{B}^*,0)\|_{\infty,2}&= \sup_{T>0} \sup_{\substack{v,x\, \text{satisfy~\eqref{eq:dualPIE}}\\\left\lVert{v}\right\rVert_{L_2[0,T]}=1} } |\mathcal{B}^* x(T)|\\ &\le \|G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B})\|_{2,i}. \end{aligned}

Now, to show that \|G(\mathcal{T},\mathcal{A},0, \mathcal{C},\mathcal{B})\|_{2,i}\le \|G(\mathcal{T}^*,\mathcal{A}^*,\mathcal{C}^*,\mathcal{B}^*,0)\|_{\infty,2}, we similarly establish that for any T>0 and any v, z x which satisfy Eq. (?) with \left\lVert{v}\right\rVert=1, we have that \left\lVert{z}\right\rVert_{L_2[0,T)} \le \|G(\mathcal{T}^*,\mathcal{A}^*,\mathcal{C}^*,\mathcal{B}^*,0)\|_{\infty,2}.

Specifically, for T>0 and any v, z x which satisfy Eq. (?) with \left\lVert{v}\right\rVert=1, let \bar v(t) = \frac{1}{\left\lVert{z}\right\rVert_{L_2[0,T)}}z(T-t) and \bar x satisfy Eq. (?) for this \bar v. Then from Lem. 8,

\begin{aligned} \left\lVert{z}\right\rVert_{L_2[0,T]}&=\frac{1}{\left\lVert{z}\right\rVert_{L_2[0,T)}}\int_0^T \left\langle z(t),z(t)\right\rangle dt\\ &\hspace{-1.4cm}=\int_0^T \left\langle \bar v(T-t),\mathcal{C} x (t)\right\rangle dt=\bigl(\mathcal{B}^* \bar x(T) \bigr)^T v\\ &\hspace{-1.4cm}\le \left\lVert{\mathcal{B}^* \bar x(T) }\right\rVert\!\left\lVert{ v}\right\rVert\!=\!\left\lVert{\mathcal{B}^* \bar x(T) }\right\rVert\!\le\! \|G(\mathcal{T}^*,\mathcal{A}^*,\mathcal{C}^*,\mathcal{B}^*,0)\|_{\infty,2} \end{aligned}

where the final inequality holds from the definition of \|G(\mathcal{T}^*,\mathcal{A}^*,\mathcal{C}^*,\mathcal{B}^*,0)\|_{\infty,2} and the fact that \bar v,\bar x satisfy Eq. (?) with \left\lVert{\bar v}\right\rVert=1. Since the bound holds uniformly for any T>0 and any v, z x which satisfy Eq. (?) with \left\lVert{v}\right\rVert=1, we conclude that

\begin{aligned} \|G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B})\|_{2,i}&= \sup_{T>0} \sup_{\substack{v,z,\mathbf{x}\text{satisfy~\eqref{eq:primal_aux}}\\ \left\lVert{v}\right\rVert_2=1}} \left\lVert{z}\right\rVert_{L_2[0,T)}\\ &\le \|G(\mathcal{T}^*,\mathcal{A}^*,\mathcal{C}^*,\mathcal{B}^*,0)\|_{\infty,2} \end{aligned}

as desired. We conclude that \|G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B})\|_{2,i}=\|G(\mathcal{T}^*,\mathcal{A}^*,\mathcal{C}^*,\mathcal{B}^*,0)\|_{\infty,2}.

Uses Lemma 8
Theorem

Theorem 10

WPJC

Given \mathcal{T,A,B,C} \in \pmb{\Pi}, suppose there exist \mathcal{P}=\mathcal{P}^* \in \pmb{\Pi}, and \gamma, \epsilon>0, such that \mathcal{P} \succcurlyeq \epsilon I, satisfying

\begin{aligned} \mathcal A^*\mathcal P\mathcal T +\mathcal T^*\mathcal P\mathcal A + \mathcal C^*\mathcal C &\preccurlyeq 0,\; \mathcal B^*\mathcal P\mathcal B \preccurlyeq \gamma I. \end{aligned}

Then, \left\lVert{G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B})}\right\rVert_{2,i}^2 \leq \gamma.

Proof of Theorem 10

We prove that if the inequalities hold for some \mathcal{P}\succcurlyeq \epsilon I, \gamma>0, then \left\lVert{G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B})}\right\rVert_{2,i}^2 \leq \gamma, recalling that

\|G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B})\|_{2,i}:= \sup_{T>0} \sup_{\substack{v,z,\mathbf{x}\text{satisfy~\eqref{eq:primal_aux}}\\ \left\lVert{v}\right\rVert_2=1}} \left\lVert{z}\right\rVert_{L_2[0,T)}.

Suppose there exist \mathcal{P}\succcurlyeq \epsilon I, \gamma>0 satisfying Eqn. (?). For any T>0, let z, x, v satisfy Eqn. (?) for t\in[0,T], given some initial condition with \left\lVert{v}\right\rVert_2=1. Define the storage functional V(x(t))=\left\langle x(t),\mathcal{T^*PT}x(t)\right\rangle_{L_2[a,b]}. Since \mathcal{P} is coercive, V(x)\ge \epsilon \left\lVert{\mathcal{T}x}\right\rVert^2 for any x\in L_2. Then, taking time derivative along solutions of Eqn. (?), we have that

\begin{aligned} \dot V(x(t))&=\left\langle x(t),(\mathcal{A^*PT+T^*PA}), x(t)\right\rangle_{L_2[a,b]}, \end{aligned}

and, since \mathcal{P} satisfies Eqn. (?), we further have that

\dot V( x(t)) \leq -\left\langle x(t),\mathcal{C^*C} x(t)\right\rangle_{L_2[a,b]}.

Integrating \dot{V} w.r.t. t over [0,T], we have that

\begin{aligned} \left\lVert{z}\right\rVert_{L_2[0,T]}^2&=\int_0^T\left\langle x(s),\mathcal{C^*C}x(s)\right\rangle ds \\ &\leq V(x(0))- V(x(T))\leq V(x(0))\\ &=\left\langle x(0),\mathcal{T^*PT}x(0)\right\rangle=\left\langle \mathcal{B} v,\mathcal{PB} v\right\rangle \end{aligned}

Then, since \mathcal{P}, \gamma satisfy Eqn. (?), we have that

\left\lVert{z}\right\rVert_{L_2[0,T]}^2 \leq \bigl|\left\langle \mathcal{B} v, \mathcal{PB} v\right\rangle\bigr| \Bigr|\leq \gamma \left\lVert{v}\right\rVert_2^2=\gamma.

Since the inequality holds uniformly for any T>0 and solution of Eqn. (?), by the definition of the impulse-to-energy norm, we conclude that \left\lVert{G(\mathcal{T,A},0,C,B)}\right\rVert_{2,i}^2 \leq \gamma, as desired.

Theorem

Theorem 11

5MEO

Given \mathcal{T,A,B,C} \in \pmb{\Pi}, suppose there exist \mathcal{P}=\mathcal{P}^* \in \pmb{\Pi}, and \gamma, \epsilon>0, such that \mathcal{P} \succcurlyeq \epsilon I, satisfying

\begin{aligned} \begin{bmatrix} -\gamma I & \mathcal{B^*PT}\\ \mathcal{T^* P B} & \mathcal{A^*PT+T^*PA} \end{bmatrix} &\preccurlyeq0, \; \mathcal{C^*C}-\mathcal{T^*PT} \preccurlyeq0. \end{aligned}

Then, \left\lVert{G(\mathcal{T,A,B,C},0)}\right\rVert_{\infty,2}^2 \leq \gamma

Proof of Theorem 11

We prove that if there exist \mathcal{P} \succcurlyeq \epsilon I, \gamma >0 satisfying Eqn. (?), then \|G(\mathcal{T},\mathcal{A},\mathcal{B},\mathcal{C},0)\|_{\infty,2}^2 \leq \gamma, recalling that

\|G(\mathcal{T},\mathcal{A},\mathcal{B},\mathcal{C},0)\|_{\infty,2}= \sup_{T >0} \sup_{\substack{v,x\, \text{satisfy~\eqref{eq:dualPIE}}\\\left\lVert{v}\right\rVert_{L_2[0,T]}=1} } \|z (T)\|.

Suppose there exist \mathcal{P}\succcurlyeq \epsilon I, \gamma>0 satisfying Eqn. (?). For any T>0, suppose z,x, v satisfy Eqn. (?), given some input with \left\lVert{v}\right\rVert_{L_2[0,T]} = 1. Define the storage functional V(x(t))=\left\langle x(t), \mathcal{T^* PT} x(t)\right\rangle. Then, V(x)\ge \epsilon \left\lVert{\mathcal{T}\bar{x}}\right\rVert^2 and 1012pt1012pt1012pt1012pt1012pt1012pt1012pt1012pt

\begin{aligned} \dot V(x(t))=\left\langle x(t),(\mathcal{A^* PT+T^*PA}) x(t)\right\rangle+2\left\langle v,\mathcal{B^*PT} x(t)\right\rangle, \end{aligned}

implying that, since \mathcal{P} and \gamma satisfy Eqn. (?),

\begin{aligned} \dot V(x(t))&-\gamma \left\lVert{v(t)}\right\rVert_2^2=\\ &\hspace{-4mm}\left\langle \begin{bmatrix} v(t)\\ x(t) \end{bmatrix}, \begin{bmatrix} -\gamma I & \mathcal{B^*PT}\\\mathcal{T^*PB}& \mathcal{A^*PT+T^*PA} \end{bmatrix}\begin{bmatrix} v(t)\\ x(t) \end{bmatrix}\right\rangle \leq 0. \end{aligned}

Integrating \dot{V} w.r.t. t over [0,T], we get

\begin{aligned} V(x(T))-V(x(0))-\gamma \left\lVert{v}\right\rVert_{L_2[0,T]}^2 \leq 0. \end{aligned}

Then, since \mathcal{P} and \gamma satisfy Eqn. (?), we have that

\begin{aligned} \left\lVert{z(T)}\right\rVert_2^2 &= \left\langle x(T), \mathcal{C^*C}x(T)\right\rangle\leq\left\langle x(T),\mathcal{T^*PT} x(T)\right\rangle\\&= V(x(T)) =V(x(T)-V(x(0))\leq\gamma\left\lVert{v}\right\rVert^2=\gamma \end{aligned}

Since the inequality holds uniformly for any T>0 and solutions of Eqn. (?), by the definition of the energy-to-peak norm, we conclude that \left\lVert{G(\mathcal{T,A,B,C},0)}\right\rVert_{\infty,2}^2 \leq \gamma, as desired.

Lemma

Lemma 12

JH91

Given Hilbert spaces X and Y, \mathcal{R}\in\mathcal{L}(X), \mathcal{S}\in\mathcal{L}(Y), and \mathcal{U}\in\mathcal{L}( Y, X) with \mathcal{S} \succcurlyeq \epsilon I for some \epsilon>0. Then the following are equivalent.

  1. \displaystyle \begin{bmatrix} \mathcal{S}&\mathcal{U}\\\mathcal{U}^*&\mathcal{R} \end{bmatrix}\succcurlyeq 0.

  2. \displaystyle \mathcal{R}-\mathcal{U}^* \mathcal{S}^{-1} \mathcal{U} \succcurlyeq 0.

Used by Proof
Corollary

Corollary 13

JMBF

Given \mathcal{T,A,B,C} \in \pmb{\Pi}, suppose there exist \mathcal{P}=\mathcal{P}^* \in \pmb{\Pi}, and \gamma, \epsilon>0, such that \mathcal{P} \succcurlyeq \epsilon I, satisfying

\begin{aligned} \begin{bmatrix} -\gamma I & \mathcal{CPT^*}\\ \mathcal{T P C^*} & \mathcal{APT^*+TPA^*} \end{bmatrix} &\preccurlyeq0,\; \mathcal{BB^*}-\mathcal{TPT^*} \preccurlyeq0. \end{aligned}

Then, \left\lVert{G(\mathcal{T,A},0,\mathcal{C},\mathcal{B})}\right\rVert_{2,i}^2 \leq \gamma

Proof of Corollary 13

As before, we first prove that if Eqn. (?) hold for some \mathcal{P}\succcurlyeq \epsilon I, \gamma>0, then \left\lVert{G(\mathcal{T^*,A^*},\mathcal{C}^*,\mathcal{B}^*,0)}\right\rVert_{\infty,2}^2 \leq \gamma, and use the duality result in Thm. 9. Suppose there exist \mathcal{P}\succcurlyeq \epsilon I, \gamma>0 satisfying Eqn. (?). Now, let \mathcal{T'=T^*}, \mathcal{A'=A^*}, \mathcal{B'=C^*}, and \mathcal{C'=B^*}. Since \mathcal{T,A,B,C} \in \pmb{\Pi} \subset \mathcal{L}(X) and (A^*)^*=A for any A\in \mathcal{L}(X), we can verify that the conditions of Thm. ?, ? hold for \mathcal{T}',\mathcal{A}',\mathcal{B}',\mathcal{C}' since

\begin{aligned} \begin{bmatrix} -\gamma I & \mathcal{(B')^*PT'}\\ \mathcal{(T')^* P B'} & \mathcal{(A')^*PT'+(T')^*PA'} \end{bmatrix} &\preccurlyeq0\;\\ \mathcal{(C')^*C'}-\mathcal{(T')^*PT'}& \preccurlyeq 0, \end{aligned}

implying \left\lVert{G(\mathcal{T}',\mathcal{A}',\mathcal{B}',\mathcal{C}',0)}\right\rVert_{\infty,2}^2\leq \gamma. Now, from Thm. 9, we can conclude that

\begin{aligned} \gamma&\ge \left\lVert{G(\mathcal{T}'\!,\!\mathcal{A}'\!,\!\mathcal{B}'\!,\mathcal{C}',0)}\right\rVert_{\infty,2}^2\\ &=\left\lVert{G(\mathcal{(T')^*\!,\!(A')^*}\!,0,\!\mathcal{(B')^*\!,\!(C')^*})}\right\rVert_{2,i}^2\\ &=\left\lVert{G(\mathcal{(T^*)^*,(A^*)^*},0,\mathcal{(C^*)^*,(B^*)^*})}\right\rVert_{2,i}^2\\ &=\left\lVert{G(\mathcal{T,A},0,\mathcal{C,B})}\right\rVert_{2,i}^2. \end{aligned}
Theorem

Theorem 14

IW8L

Given \mathcal{T,A,B,C} \in \pmb{\Pi}, suppose at least one of the following statements hold.

  1. There exist \mathcal{P}=\mathcal{P}^* \in \pmb{\Pi}, and \gamma, \epsilon>0, such that \mathcal{P} \succcurlyeq \epsilon I, satisfying,

    \begin{aligned} \begin{bmatrix} -\gamma I & \mathcal{C} \\\mathcal{C}^* &\mathcal A^*\mathcal P\mathcal T +\mathcal T^*\mathcal P\mathcal A \end{bmatrix} &\preccurlyeq 0, \, \begin{bmatrix} \mathcal{P} &\mathcal{PB}\\ \mathcal{B^*P} &\gamma I \end{bmatrix} &\succcurlyeq 0 \end{aligned}
  2. There exist \mathcal{P}=\mathcal{P}^* \in \pmb{\Pi}, and \gamma,\epsilon>0, such that \mathcal{P} \succcurlyeq \epsilon I, satisfying

    \begin{aligned} \begin{bmatrix} -\gamma I & \mathcal{CPT^*}\\ \mathcal{T P C^*} & \mathcal{APT^*+TPA^*} \end{bmatrix} &\preccurlyeq0, \, \begin{bmatrix} \gamma I & \mathcal{B}^*\\\mathcal{B} & \mathcal{TPT^*} \end{bmatrix} \succcurlyeq 0. \end{aligned}

Then, \left\lVert{G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B})}\right\rVert_{2,i} \leq \gamma.

Proof of Theorem 14

We prove each statement separately

Proof of 1): Suppose 1) holds for some \mathcal{P} \succcurlyeq \epsilon I, \gamma>0. Then, \mathcal{P} \in \pmb{\Pi} \subset \mathcal{L}(X), and \mathcal{P},\gamma satisfy Eqn. (?). Now, we apply Lem. 12 using \mathcal{U}=-\mathcal{C} \in \mathcal{L}(X), \mathcal{R}=-(\mathcal{A^*PT+T^*PA}) \in \mathcal{L}(X), and \mathcal{S} = \gamma I \in \mathcal{L}(X). Since \mathcal{S} \succcurlyeq \epsilon I for any \epsilon\in(0,\gamma], the conditions of Lem. 12 are satisfied. Then, we have that

\begin{bmatrix} -\mathcal{S}&-\mathcal{U}\\-\mathcal{U}^*&-\mathcal{R} \end{bmatrix}= \begin{bmatrix} -\gamma I & \mathcal{C}\\\mathcal{C}^*& \mathcal{A^*PT+T^*PA} \end{bmatrix}\preccurlyeq0,
\text{and hence}\,\begin{bmatrix} \mathcal{S}&\mathcal{U}\\\mathcal{U}^*&\mathcal{R} \end{bmatrix}\succcurlyeq0.

Then, statement 1) of Lem. 12 holds. This implies

-\Bigl( \mathcal{A^*PT+T^*PA+C^*C}\frac{1}{\gamma}\Bigr)=\mathcal{R-U^*}\mathcal{S}^{-1}\mathcal{U} \succcurlyeq 0.

Furthermore, \mathcal{P},\gamma satisfy Eqn. (?).

Now, we apply Lem. 12 using \mathcal{U}=\mathcal{PB} \in \mathcal{L}(X), \mathcal{R}=\gamma I \in \mathcal{L}(X), and \mathcal{S} = \mathcal{P} \in \mathcal{L}(X). Then, since \mathcal{S} =\mathcal{P} \succcurlyeq \epsilon I, the conditions of Lem. 12 are satisfied. Then, we have that

\begin{bmatrix} \mathcal{S}&\mathcal{U}\\\mathcal{U}^*&\mathcal{R} \end{bmatrix}= \begin{bmatrix} \mathcal{P} & \mathcal{PB}\\\mathcal{B^*P}&\gamma I \end{bmatrix}\succcurlyeq0,

and hence, statement 1) of Lem. 12 holds. This implies

\gamma I -\mathcal{B^*PB} = \gamma I -\mathcal{B^*PP}^{-1}\mathcal{PB}=\mathcal{R-U^*}\mathcal{S}^{-1}\mathcal{U} \succcurlyeq 0.

Next, let \hat{\mathcal{P}} =\mathcal{P}\gamma. Then, we have that \hat{\mathcal{P}} \succcurlyeq \gamma \epsilon I and

\begin{aligned} \mathcal{A^*\hat{P}T+T^* \hat{P} A+C^*C} &\preccurlyeq0, \; \gamma^2 I-\mathcal{B^*}\hat{\mathcal{P}}\mathcal{B} \succcurlyeq 0, \end{aligned}

and hence \mathcal{\hat P}=\mathcal{P} \gamma, and \gamma^2 satisfy the conditions of Thm. ?, ?. This implies that \left\lVert{G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B})}\right\rVert_{2,i}^2 \leq \gamma^2, and hence \left\lVert{G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B})}\right\rVert_{2,i} \leq \gamma, as desired.

Proof of 2): Next, we prove statement 2) by showing that if there exist \mathcal{P} \succcurlyeq \epsilon I, \gamma >0 satisfying 2), then \|G(\mathcal{T}^*,\mathcal{A}^*,\mathcal{C}^*,\mathcal{B}^*,0)\|_{\infty,2} \leq \gamma, recalling that

\|G(\mathcal{T},\mathcal{A},\mathcal{B},\mathcal{C},0)\|_{\infty,2}= \sup_{T >0} \sup_{\substack{v,x\, \text{satisfy~\eqref{eq:dualPIE}}\\\left\lVert{v}\right\rVert_{L_2[0,T]}=1} } \|z (T)\|.

Suppose there exist \mathcal{P}\succcurlyeq \epsilon I, \gamma>0 satisfying Eqn. (?).

Now, we apply Lem. 12, using \mathcal{U} =\mathcal{B^*} \in \mathcal{L}(X), \mathcal{R}=\mathcal{TPT^*} \in \mathcal{L}(X), and \mathcal{S}=\gamma I \in \mathcal{L}(X). Since \gamma I \succcurlyeq \epsilon I for any \epsilon\in(0,\gamma], the conditions of Lem. 12 are satisfied. We have

\begin{bmatrix} \mathcal{S}&\mathcal{U}\\\mathcal{U}^*&\mathcal{R} \end{bmatrix}=\begin{bmatrix} \gamma I & \mathcal{B}^*\\\mathcal{B} & \mathcal{TPT^*} \end{bmatrix}\succcurlyeq 0,

and hence, statement 1) of Lem. 12 holds. This implies

\mathcal{TPT^*} -\gamma^{-1}\mathcal{BB^*} =\mathcal{R}- \mathcal{U}^* \mathcal{S}^{-1}\mathcal{U} \succcurlyeq 0.

Next, let \hat{\mathcal{P}} =\mathcal{P}\gamma. Then, we have that \hat{\mathcal{P}} \succcurlyeq \gamma \epsilon I, and

\begin{aligned} \begin{bmatrix} -\gamma^2 I & \mathcal{C\hat PT^*}\\ \mathcal{T \hat P C^*} & \mathcal{A\hat PT^*+T\hat PA^*} \end{bmatrix} &\preccurlyeq0, \;\mathcal{T\hat PT^*} -\mathcal{BB^*}\succcurlyeq 0, \end{aligned}

and hence \mathcal{\hat P}=\mathcal{P} \gamma, and \gamma^2 satisfy the conditions of Cor. ?, ?. This implies \left\lVert{G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B})}\right\rVert_{2,i}^2\leq \gamma^2, and hence \left\lVert{G(\mathcal{T},\mathcal{A},0,\mathcal{C},\mathcal{B})}\right\rVert_{2,i}\leq \gamma is proven.

Note

Note

1T7N

Items 1) of Thm. ?, ? and Thm. ?, ? can be shown to be equivalent.

Corollary

Corollary 15

SOHP

Given \mathcal{T,A,B,C} \in \pmb{\Pi}, suppose there exist \mathcal{P}=\mathcal{P}^* \in \pmb{\Pi}, and \gamma,\epsilon>0, such that \mathcal{P} \succcurlyeq \epsilon I, satisfying

\begin{aligned} \mathcal A\mathcal P\mathcal T^* +\mathcal T\mathcal P\mathcal A^* + \mathcal B\mathcal B^* &\preccurlyeq 0, \;\; \mathcal{CPC^*} \preccurlyeq \gamma I. \end{aligned}

Then, \left\lVert{G(\mathcal{T,A,B,C},0)}\right\rVert_{\infty,2}^2 \leq \gamma.

Proof of Corollary 15

We prove that if Eqn. (?) hold for some \mathcal{P}\succcurlyeq \epsilon I, \gamma>0, then \left\lVert{G(\mathcal{T^*,A^*},0,\mathcal{B}^*,\mathcal{C}^*)}\right\rVert_{2,i}^2 \leq \gamma. We use duality, in Thm. 9, to assert that \left\lVert{G(\mathcal{T,A,B,C},0)}\right\rVert_{\infty,2}^2 \leq \gamma. Suppose there exist \mathcal{P}\succcurlyeq \epsilon I, \gamma>0 satisfying Eqn. (?). Let \mathcal{T'=T^*}, \mathcal{A'=A^*}, \mathcal{B'=C^*}, and \mathcal{C'=B^*}. By Thm. 11.3 of Chapter II of [gohbergVol1], (A^*)^*=A for any A\in \mathcal{L}(X). Since \mathcal{T,A,B,C} \in \pmb{\Pi} \subset \mathcal{L}(X), we have (\mathcal{T}')^*=\mathcal{T}, (\mathcal{A}')^*=\mathcal{A}, (\mathcal{C}')^*=\mathcal{B}, and (\mathcal{B}')^*=\mathcal{C} and consequently, conditions of Thm. ?, ? hold,

\begin{aligned} (\mathcal A')^*\mathcal P\mathcal T' +(\mathcal T')^*\mathcal P\mathcal A' +(\mathcal C')^*\mathcal C' \preccurlyeq 0,\; \mathcal{(B')^*PB'} \preccurlyeq \gamma I, \end{aligned}

and hence we conclude that \left\lVert{G(\mathcal{T}',\mathcal{A}',0,\mathcal{C}',\mathcal{B}')}\right\rVert_{2,i}^2\leq \gamma. Lastly, from Thm. 9, we have that

\begin{aligned} \gamma&\ge \left\lVert{G(\mathcal{T}'\!,\!\mathcal{A}'\!,0,\!\mathcal{C}'\!,\!\mathcal{B}')}\right\rVert_{2,i}^2\\ &=\left\lVert{G((\mathcal{T}')^*\!,\!(\mathcal{A}')^*\!,\!(\mathcal{C}')^*\!,\!(\mathcal{B}')^*\!,\!0)}\right\rVert_{\infty,2}^2\\&=\left\lVert{G(\mathcal{(T^*)^*,(A^*)^*,(B^*)^*,(C^*)^*},0)}\right\rVert_{\infty,2}^2\\ &=\left\lVert{G(\mathcal{T,A,B,C},0)}\right\rVert_{\infty,2}^2. \end{aligned}
Theorem

Theorem 16

U5R7

Given \mathcal{T,A,B,C} \in \pmb{\Pi}, suppose at least one of the following statements hold.

  1. There exist \mathcal{P}=\mathcal{P}^* \in \pmb{\Pi}, and \gamma,\epsilon>0, such that \mathcal{P} \succcurlyeq \epsilon I, satisfying

    \hspace{-6mm}\begin{bmatrix} -\gamma I & \mathcal{B}^* \\\mathcal{B}&\mathcal A\mathcal P\mathcal T^* +\mathcal T\mathcal P\mathcal A^* \end{bmatrix}\preccurlyeq 0 \quad \begin{bmatrix} \mathcal{P} &\mathcal{PC^*}\\ \mathcal{CP}&\gamma I \end{bmatrix} \succcurlyeq 0
  2. There exist \mathcal{P}=\mathcal{P}^* \in \pmb{\Pi}, and \gamma,\epsilon>0, such that \mathcal{P} \succcurlyeq \epsilon I, satisfying

    \hspace{-6mm}\begin{bmatrix} -\gamma I & \mathcal{B^*PT}\\ \mathcal{T P B} & \mathcal{A^*PT+T^*PA} \end{bmatrix} \preccurlyeq0 \quad \begin{bmatrix} \gamma I & \mathcal{C}\\\mathcal{C}^* & \mathcal{T^*PT} \end{bmatrix}\succcurlyeq 0

Then, \left\lVert{G(\mathcal{T,A,B,C},0)}\right\rVert_{\infty,2} \leq \gamma.

Proof of Theorem 16

We use the duality result of Thm. 9, to prove that the LPIs in statements 1) and 2) imply \left\lVert{G(\mathcal{T,A,B,C},0)}\right\rVert_{\infty,2} \leq \gamma. First, we show 1) implies \left\lVert{G(\mathcal{T}',\mathcal{A}',0,\mathcal{C}',\mathcal{B}')}\right\rVert_{2,i} \leq \gamma.

Suppose 1) holds for some \mathcal{P}\succcurlyeq \epsilon I,\gamma>0. Then, \mathcal{P}, \gamma satisfy Eqn. (?).

Now, let \mathcal{T'=T^*}, \mathcal{A'=A^*}, \mathcal{B'=C^*}, and \mathcal{C'=B^*}. Then, statement 1) of Thm. ?, ? hold for G(\mathcal{T}',\mathcal{A}',0,\mathcal{C}',\mathcal{B}'):

\begin{aligned} \begin{bmatrix} -\gamma I & \mathcal{C}' \\(\mathcal{C}')^* &(\mathcal A')^*\mathcal P\mathcal T' +(\mathcal T')^*\mathcal P\mathcal A' \end{bmatrix} \!\preccurlyeq\! 0, \;\begin{bmatrix} \gamma I&(\mathcal{B}')^*\mathcal{P}\\ \mathcal{P} \mathcal{B}'&\mathcal{P} \end{bmatrix} \!\succcurlyeq\! 0, \end{aligned}

which implies that \left\lVert{G(\mathcal{T}',\mathcal{A}',0,\mathcal{C}',\mathcal{B}')}\right\rVert_{2,i}\leq \gamma.

Next, we show that 2) implies \left\lVert{G(\mathcal{T}',\mathcal{A}',0,\mathcal{C}',\mathcal{B}')}\right\rVert_{2,i} \leq \gamma. Suppose 2) holds for some \mathcal{P}\succcurlyeq \epsilon I,\gamma>0. Then, \mathcal{P}, \gamma satisfy Eqn. (?). Hence, statement 2) of Thm. ?, ? hold for G(\mathcal{T}',\mathcal{A}',0,\mathcal{C}',\mathcal{B}'). Specifically, we have that 1012pt1012pt1012pt1012pt1012pt1012pt1012pt1012pt

\begin{aligned} \begin{bmatrix} -\gamma I &\hspace{-2 mm} \mathcal{C'P(T')^*}\\ \mathcal{T' P (C')^*} &\hspace{-2 mm} \mathcal{A'P(T')^*+T'P(A')^*} \end{bmatrix} \preccurlyeq0, \begin{bmatrix} \gamma I &\hspace{-1 mm} (\mathcal{B}')^*\\\mathcal{B}' &\hspace{-1 mm} \mathcal{T'P(T')^*} \end{bmatrix}\succcurlyeq 0, \end{aligned}

which implies that \left\lVert{G(\mathcal{T}',\mathcal{A}',0,\mathcal{C}',\mathcal{B}')}\right\rVert_{2,i}\leq \gamma.

Now, we are in position to invoke Thm. 9, using G(\mathcal{T}',\mathcal{A}',0,\mathcal{C}',\mathcal{B}'), which states that

\begin{aligned} \left\lVert{G(\mathcal{T}',\mathcal{A}',0,\mathcal{C}',\mathcal{B}')}\right\rVert_{2,i}&=\left\lVert{G(\mathcal{T}^*,\mathcal{A}^*,0,\mathcal{B}^*,\mathcal{C}^*)}\right\rVert_{2,i}\\&\hspace{-9mm}=\left\lVert{G(\mathcal{(T')^*,(A')^*,(B')^*,(C')^*},0)}\right\rVert_{\infty,2}. \end{aligned}

Since \mathcal{T,A,B,C} \in \pmb{\Pi} \subset \mathcal{L}(X), we have that, by Thm. 11.3 of Chapter II of [gohbergVol1], (\mathcal{T}')^*=\mathcal{T}^{**}=\mathcal{T}, (\mathcal{A}')^*=\mathcal{A}^{**}=\mathcal{A}, (\mathcal{B}')^*=\mathcal{B}^{**}=\mathcal{B}, and (\mathcal{C}')^*=\mathcal{C}^{**}=\mathcal{C}, and hence

\left\lVert{G(\mathcal{T}',\mathcal{A}',0,\mathcal{C}',\mathcal{B}')}\right\rVert_{2,i}=\left\lVert{G(\mathcal{T,A,B,C},0)}\right\rVert_{\infty,2}.

Since both 1) and 2) imply \left\lVert{G(\mathcal{T}',\mathcal{A}',0,\mathcal{C}',\mathcal{B}')}\right\rVert_{2,i}\leq \gamma, both imply \left\lVert{G(\mathcal{T,A,B,C},0)}\right\rVert_{\infty,2}\leq \gamma, as desired.

Example

Example 1

IJB9

First consider the scalar, single delay system, where the non-zero matrices are A_1=-1, B=2, and C_0=2, with the delay \tau = 0.5. This system has an analytic bound for the H_2 norm, which is 3.6724.

Example

Example 2

8SOY

Next, consider the two-delay system with \tau_1 = \pi/10, \tau_2 = 1, so that the delays are incommensurate, and

\begin{aligned} &A_0 = \begin{bmatrix} -1 & 1 & 2\\ 1 & -3 & 2\\ 0 & 0 & -1 \end{bmatrix},\qquad A_1 = \frac{1}{5}\begin{bmatrix} -3 & 0 & 1\\ 1 & -2 & 0\\ 0 & 2 & -2 \end{bmatrix},\\ & A_2 = \frac{1}{5}\begin{bmatrix} -4 & 1 & 0\\ 0 & -2 & 1\\ 2 & 1 & -3 \end{bmatrix}, \quad B = \begin{bmatrix} 1 \\ 1 \\1 \end{bmatrix}, \; C_0 = \begin{bmatrix} 1 & 1 &1 \end{bmatrix}. \end{aligned}

The method in [jarlebring2011] and the bounds from the LPIs agree to four significant figures. However, 10th order Padé approximation yields a higher value of of 3.9275 for the H_2–norm.

Source and licence

The text on this page is from arXiv:2610.21355v1 by Danilo Braghini, Tristan Thomas, Sachin Shivakumar, Javad M. Velni and Matthew M. Peet, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.

The record of each item is in analysis/objects/2610/2610.21355.jsonl.