Scaled Graph Containment for Feedback Stability: Soft–Hard Equivalence and Conic Regions
Eder Baron-Prada, Julius P. J. Krebbekx, Adolfo Anta and Florian Dörfler · 2026
Read Scaled Graph Containment for Feedback Stability: Soft–Hard Equivalence and Conic Regions by Eder Baron-Prada, Julius P. J. Krebbekx, Adolfo Anta and Florian Dörfler (2026) 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-2604-05567.txt.
Introduction
Baron-Prada and three colleagues develop containment rules for scaled graphs inside multiplier-defined circular and conic regions. For circles, soft and hard containment coincide when the multiplier is positive-negative, letting soft computations certify hard stability. Tests on systems up to 300 states save 15 to 44 percent of compute, and conic regions bound nonsymmetric operators more tightly than circles.
Catalog record
| CREATOR | Eder Baron-Prada, Julius P. J. Krebbekx, Adolfo Anta and Florian Dörfler |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 38730 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.05567
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Scaled Graph Containment for Feedback Stability: Soft–Hard Equivalence and Conic Regions Eder Baron-Prada, Julius P. J. Krebbekx, Adolfo Anta and Florian Dörfler Abstract Scaled graphs (SGs) offer a geometric framework for feedback stability analysis. This paper develops containment conditions for SGs within multiplier-defined regions, addressing both circular and conic geometries. For circular regions, we show that soft and hard SG containment are equivalent whenever the associated multiplier is positive-negative. This enables hard stability certification from soft computations alone, bypassing both the positive semidefinite storage constraint and the homotopy condition of existing methods. Numerical experiments on systems with up to 300 states demonstrate computational savings of 15–44% for the circular containment framework. We further characterize which conic regions are hyperbolically convex, a condition our frequency-domain certificate requires, and demonstrate that such regions provide tighter SG bounds than circles whenever the operator SG is nonsymmetric. I Introduction Graphical methods have been central to feedback stability analysis since the work of Nyquist and Bode, offering intuitive geometric certificates. Extending this paradigm to MIMO LTI and nonlinear systems has been a long-standing goal. Scaled graphs (SGs), introduced in [ 20 ] and developed for feedback analysis in [ 6 ] , address this gap by representing operators as subsets of the complex plane that simultaneously encode gain and phase. Several SG formulations have been proposed, including signed [ 23 ] , soft [ 6 ] , and hard [ 7 ] variants. This study focuses on the latter two. The soft SG is defined on \mathcal{L}_{2} and characterizes the asymptotic behavior of \mathcal{L}_{2} -stable operators; the hard SG is defined on \mathcal{L}_{2e} , capturing finite-horizon behavior and accommodating persistent and unbounded trajectories. Two fundamental limitations currently constrain SG-based stability analysis. The first is geometric ; existing SG containment methods [ 8 , 17 , 13 ] rely on static multiplier regions, which describe only circular regions (disks, disk complements, or half-planes). Some specifications of practical interest, such as elliptical exclusion zones require conic containment regions that lie outside this class [ 11 ] . The second limitation is computational ; soft SGs are efficiently constructed via LMIs [ 8 ] or frequency-domain sampling [ 3 , 14 ] , but certifying stability from them requires checking separation conditions over infinitely many scaled sets [ 6 ] , a condition that is numerically expensive to verify. Hard SGs yield a simple geometric separation test [ 7 ] , but their computation demands a semidefinite constraint P\succeq 0 that becomes costly for large-scale systems [ 13 , 17 , 8 ] . This paper addresses both limitations. For the geometric limitation, we extend the containment framework from circular regions to general conic sections, following the frequency-domain inequality framework of [ 12 ] . We characterize exactly which conic regions are hyperbolically convex, the geometric property required for SG containment, and provide a frequency-domain certification condition for conic containment. For the computational limitation, we use the lens of Integral Quadratic Constraints (IQCs) [ 15 , 4 ] . We show that when the circular IQC multiplier is positive-negative [ 5 , 21 ] , soft and hard SG containment in a multiplier-defined region are equivalent for \mathcal{L}_{2} -stable systems. This enables a hybrid certification pipeline: compute SG regions via efficient soft LMIs, then certify stability via the simple hard separation condition, bypassing both the P\succeq 0 constraint and the homotopy sweep. The practical implications are significant. Numerical experiments on LTI systems with up to 300 states demonstrate computational savings of 15–44% from the soft LMI relaxation alone, enabling SG-based stability certification for large-scale systems, including power systems [ 2 ] and networked multi-agent systems [ 1 ] , where direct hard SG computation is currently prohibitive. Beyond computational efficiency, when the SG is elongated, a conic region can be shaped to fit it more closely than any disk: since stability certificates are derived from the containment region, tighter regions directly translate into less conservative stability margins. II Preliminaries II-A Signal Spaces, Systems, and Scaled Graphs Let \mathcal{L}_{2}^{n} denote the space of square-integrable signals u:\mathbb{R}_{\geq 0}\to\mathbb{R}^{n} with norm \lVert u\rVert^{2}:=\int_{0}^{\infty}u(t)^{\top}u(t)\,dt and inner product \langle u,y\rangle:=\int_{0}^{\infty}u(t)^{\top}y(t)\,dt . The extended space \mathcal{L}_{2e}^{n}:=\{u:\mathbb{R}_{\geq 0}\to\mathbb{R}^{n}\mid\mathcal{P}_{T}u\in\mathcal{L}_{2}^{n},\,\forall T\geq 0\} accommodates signals with unbounded energy, where \mathcal{P}_{T} truncates to [0,T] . A system H:\mathcal{L}_{2e}^{n}\to\mathcal{L}_{2e}^{n} is causal if \mathcal{P}_{T}(Hu)=\mathcal{P}_{T}(H(\mathcal{P}_{T}u)) for all T\geq 0 [ 24 ] . We also consider finite-dimensional LTI systems with zero initial condition, i.e., x(0)=0 , and \dot{x}(t)=Ax(t)+Bu(t),\;y(t)=Cx(t)+Du(t) , where x\in\mathbb{R}^{m} is the state vector, u\in\mathbb{R}^{n} is the input, and y\in\mathbb{R}^{n} is the output, with appropriately dimensioned matrices A , B , C , and D . The system has transfer function H(s)=C(sI-A)^{-1}B+D , where x\in\mathbb{R}^{m} , u,y\in\mathbb{R}^{n} . The system is \mathcal{L}_{2} -stable if \lVert H\rVert_{\infty}:=\sup_{\omega}\bar{\sigma}(H(\textup{j}\omega))<\infty . SGs generalize input–output pairs to gain–phase points in \mathbb{C}\cup{\infty} [ 6 ] . For u,y\in\mathcal{L}_{2}^{n} with u\neq 0 , define \displaystyle\rho(u,y):=\tfrac{\lVert y\rVert}{\lVert u\rVert},\quad\theta(u,y):=\arccos\!\left(\tfrac{\langle u,y\rangle}{\lVert u\rVert\lVert y\rVert}\right), (1) with \theta(u,y)=0 if y=0 . The soft SG of H:\mathcal{L}_{2}\to\mathcal{L}_{2} is \displaystyle\mathrm{SG}(H):=\left\{\rho(u,y)\,e^{\pm\mathrm{j}\theta(u,y)}\mid\forall u\in\mathcal{L}_{2}\setminus\{0\},\,y=Hu\right\}, and the hard SG of H:\mathcal{L}_{2e}\to\mathcal{L}_{2e} , using truncated gain \rho_{T}:=\rho(\mathcal{P}_{T}u,\mathcal{P}_{T}y) and phase \theta_{T}:=\theta(\mathcal{P}_{T}u,\mathcal{P}_{T}y) , is \displaystyle{\mathrm{SG}_{e}}(H)\!:=\!\left\{\rho_{T}(u,y)e^{\pm\mathrm{j}\theta_{T}(u,y)}\!\mid\!\forall u\!\in\mathcal{L}_{2e},y=Hu,\forall T\!>\!0\!\right\}. Let \mathbb{S}^{n} denote the real symmetric n\times n matrices. The open upper half-plane is \mathbb{C}_{+}:=\{z\in\mathbb{C}:\operatorname{Im}\{z\}>0\} . A set \mathcal{R}\subset\mathbb{C}_{+} is hyperbolically convex (h-convex) if for every pair of points in \mathcal{R} , the geodesic connecting them lies in \mathcal{R} (See [ 19 ] for details). II-B IQCs and J -Spectral Factorization A system H:\mathcal{L}_{2}^{n}\to\mathcal{L}_{2}^{n} satisfies a soft IQC defined by a Hermitian constant multiplier \Pi\in\mathbb{S}^{2n} if \displaystyle\int_{0}^{\infty}\begin{bmatrix}y\\ u\end{bmatrix}^{\top}\Pi\begin{bmatrix}y\\ u\end{bmatrix}dt\geq 0,\,\forall\,u\in\mathcal{L}_{2}^{n},\;y=Hu. (2) It satisfies a hard IQC if for any T\geq 0 \displaystyle\int_{0}^{T}\begin{bmatrix}y\\ u\end{bmatrix}^{\top}\Pi\begin{bmatrix}y\\ u\end{bmatrix}dt\geq 0,\,\forall\,u\in\mathcal{L}_{2e}^{n},\;y=Hu. (3) The passage from soft to hard IQCs is governed by the factorization structure of \Pi [ 15 ] . Definition 1 ( J -Spectral Factorization [ 21 ] ) . Let \Pi=\Pi^{\top}\in\mathbb{S}^{2n} be Hermitian. We say that \Pi admits a J -spectral factorization if there exists an invertible matrix \Psi\in\mathbb{R}^{2n\times 2n} such that \Pi=\Psi^{\top}J\Psi, where J:=\operatorname{diag}(I_{n},-I_{n}) . Definition 2 (Positive-Negative Matrix) . Let \Pi=\Pi^{\top}\in\mathbb{C}^{2n\times 2n} be partitioned as \Pi=\begin{bmatrix}\Pi_{11}&\Pi_{12}\\ \Pi_{21}&\Pi_{22}\end{bmatrix}, where \Pi_{ij}\in\mathbb{R}^{n\times n} . We say that \Pi is positive-negative if there exists \varepsilon>0 such that \Pi_{22}\succeq\varepsilon I_{n} and \Pi_{11}\preceq-\varepsilon I_{n}. Every positive-negative multiplier admits a J -spectral factorization \Pi=\Psi^{\top}J\,\Psi [ 21 ] ; moreover, this factorization is doubly hard [ 5 ] , which is required for the following soft–hard equivalence result based on [ 21 , 5 ] . Lemma 1 (Soft–hard IQC equivalence) . Let H:\mathcal{L}_{2}^{n}\to\mathcal{L}_{2}^{n} be a stable causal operator and let \Pi\in\mathbb{S}^{2n} be a constant positive-negative multiplier. Then H satisfies ( 2 ) if and only if H satisfies ( 3 ). Proof. (\Rightarrow) Since \Pi is constant and positive-negative, it admits a J -spectral factorization \Pi=\Psi^{\top}J\,\Psi where \Psi\in\mathbb{R}^{2n\times 2n} is a constant invertible matrix [ 21 , Lemma 4] . Setting z(t):=\Psi\Bigl[\begin{smallmatrix}(Hu)(t)\\ u(t)\end{smallmatrix}\Bigr] , the pointwise identity \bigl[\begin{smallmatrix}Hu\\ u\end{smallmatrix}\bigr]^{\top}\Pi\,\bigl[\begin{smallmatrix}Hu\\ u\end{smallmatrix}\bigr]=z^{\top}J\,z yields \displaystyle\int_{0}^{T}\begin{bmatrix}Hu\\ u\end{bmatrix}^{\top}\Pi\,\begin{bmatrix}Hu\\ u\end{bmatrix}\,dt=\!\int_{0}^{T}z^{\top}J\,z\,dt,\quad\forall\,T\leq\infty. (4) The soft IQC ( 2 ) is therefore equivalent to \int_{0}^{\infty}z^{\top}J\,z\,dt\geq 0 . Since \Psi is constant the factorization (\Psi,J) has no internal states; applying [ 21 , Lemma 2] with the storage \overline{M}=\underline{M}=0 (which holds as \Pi is positive-negative [ 5 , Thm 4.1] [ 21 , Thm 4] ), the truncated inequality \int_{0}^{T}z^{\top}J\,z\,dt\geq 0 holds for