⚡ electrical engineering

Scaled Graph Containment for Feedback Stability: Soft-Hard Equivalence and Conic Regions

This paper establishes that soft and hard scaled graph containment are equivalent for circular regions under positive-negative multipliers, enabling more efficient stability certification without storage or homotopy constraints, while also characterizing hyperbolically convex conic regions that yield tighter bounds for nonsymmetric operators.

Original authors: Eder Baron-Prada, Julius P. J. Krebbekx, Adolfo Anta, Florian Dörfler

Published 2026-04-08

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Eder Baron-Prada, Julius P. J. Krebbekx, Adolfo Anta, Florian Dörfler

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to keep a complex machine, like a power grid or a fleet of drones, running smoothly without it shaking apart or crashing. In engineering, this is called stability. To prove a machine is stable, engineers often use a mathematical "map" called a Scaled Graph (SG).

Think of the Scaled Graph as a shadow cast by the machine's behavior onto a piece of paper. If this shadow stays within a safe, drawn boundary, the machine is safe. If the shadow touches or crosses the boundary, the machine might become unstable.

This paper introduces two major improvements to how we draw these maps and boundaries to make the process faster and more accurate.

1. The "Soft" vs. "Hard" Shadow Problem

Traditionally, there were two ways to draw this shadow:

The "Soft" Shadow: This is like a quick sketch. It's easy to draw and very fast to calculate, but it only tells you about the machine's behavior after it has been running for a long time (steady state).
The "Hard" Shadow: This is a detailed, high-resolution scan. It accounts for every tiny glitch, even if the machine is just starting up or behaving wildly. It is the "gold standard" for safety, but calculating it is like trying to solve a massive puzzle while blindfolded—it takes a huge amount of computer power and time.

The Paper's Big Breakthrough:
The authors discovered a special trick. If the "boundary" you draw around the shadow is a specific type of shape (mathematically called a "positive-negative" region), then the quick sketch (Soft) and the detailed scan (Hard) are actually the same thing.

The Analogy: Imagine you are checking if a car fits in a garage. Usually, you need a laser scanner (Hard) to be 100% sure. But the authors found that if the garage is shaped just right, you can just use a quick tape measure (Soft) and be guaranteed that the laser scanner would give you the same result.
Why it matters: Engineers can now use the fast, easy method to prove safety for massive systems (like a whole city's power grid) without needing supercomputers. The paper shows this saves 15% to 44% of computing time.

2. The "Round" vs. "Egg-Shaped" Boundary Problem

For a long time, engineers could only draw circular boundaries (like a hula hoop) around these shadows.

The Problem: Real-world machine shadows are rarely perfect circles. They are often stretched, squashed, or egg-shaped. If you force a round hula hoop around an egg, you have to make the hoop huge to cover the egg's width. This leaves a lot of empty, wasted space.
The Consequence: Because the boundary is too loose (too big), the safety certificate is "conservative." It says, "This is safe," but it's being overly cautious, missing out on the machine's true potential.

The Paper's Second Breakthrough:
The authors developed a way to draw conic boundaries (shapes like ellipses, parabolas, or hyperbolas).

The Analogy: Instead of forcing a round hula hoop around an egg, they now let you draw a custom egg-shaped mold that fits the shadow perfectly.
Why it matters: A tighter fit means a more accurate safety margin. It allows engineers to push machines closer to their limits safely, knowing exactly where the edge is, rather than guessing with a giant circle.

Summary: What Does This Mean for You?

Faster Safety Checks: We can now verify that huge, complex systems (like smart grids or self-driving car networks) are stable much faster, saving time and money.
Better Precision: We can use custom-shaped safety zones instead of generic circles, allowing for more efficient and powerful system designs.
The "Magic" Condition: All of this works because the authors found a specific mathematical "key" (the positive-negative multiplier) that unlocks the equivalence between the fast method and the slow, rigorous method.

In short, this paper gives engineers a faster, sharper, and more flexible toolkit to ensure the complex machines of our future stay safe and stable.

1. Problem Statement

The paper addresses two fundamental limitations in Scaled Graph (SG)-based feedback stability analysis for Multi-Input Multi-Output (MIMO) Linear Time-Invariant (LTI) and nonlinear systems:

Geometric Limitation: Existing SG containment methods rely on static multiplier regions that are restricted to circular geometries (disks, disk complements, or half-planes). This prevents the analysis of systems requiring more complex exclusion zones, such as elliptical or general conic regions, which are often necessary for non-symmetric operators.
Computational Limitation:
- Soft SGs (defined on $L_2$ ) are computationally efficient to construct via Linear Matrix Inequalities (LMIs) or frequency sampling but require checking separation conditions over infinitely many scaled sets (homotopy condition) to certify stability.
- Hard SGs (defined on $L_{2e}$ ) allow for a simple, static geometric separation test but require solving LMIs with a positive semidefinite constraint ( $P \succeq 0$ ). This constraint is numerically expensive and often prohibitive for large-scale systems (e.g., power grids with hundreds of states).

The authors aim to bridge these gaps by establishing an equivalence between soft and hard containment under specific conditions and extending the framework to general conic regions.

2. Methodology

A. Soft–Hard Equivalence via Positive-Negative Multipliers

The core theoretical contribution is the proof that soft and hard SG containment are equivalent within a region defined by a positive-negative multiplier.

Positive-Negative Multiplier: A Hermitian matrix $\Pi$ is defined as positive-negative if its block structure satisfies $\Pi_{22} \succeq \epsilon I$ and $\Pi_{11} \preceq -\epsilon I$ .
J-Spectral Factorization: The authors leverage the fact that positive-negative multipliers admit a J-spectral factorization ( $\Pi = \Psi^\top J \Psi$ ) that is "doubly hard."
Result: For $L_2$ -stable systems, if a system satisfies a soft Integral Quadratic Constraint (IQC) defined by a positive-negative multiplier, it automatically satisfies the corresponding hard IQC. This implies that $SG(H) \subset S(\Pi) \iff SGe(H) \subset S(\Pi)$ .
Implication: One can compute the containment region using efficient soft LMIs (without the $P \succeq 0$ constraint) and use that result to certify hard stability, bypassing the expensive hard SG computation and the homotopy sweep.

B. Extension to Conic Regions

The paper generalizes the containment framework from circular regions to conic sections (ellipses, hyperbolas, parabolas).

Geometric Formulation: Instead of $2 \times 2$ multipliers (which yield circles), the authors use $3 \times 3$ real-symmetric matrices $\Theta$ to define regions via quadratic forms on the augmented vector $v(z) = [\text{Re}(z), \text{Im}(z), 1]^\top$ .
Hyperbolic Convexity (h-convexity): A critical requirement for SG containment is that the region must be h-convex. The authors derive a necessary and sufficient condition for a conic region $C(\Theta)$ to be h-convex: $\Theta_{11} \geq \Theta_{22}$ . This condition ensures the conic is "tall" (extended in the imaginary direction) or reduces to a disk/half-plane.
Frequency-Domain Certificate: A sufficient condition for SG containment in a conic region is provided as a frequency-domain inequality involving the Hermitian part of the transfer function, $H_s(\omega)$ .

C. Stability Certification Pipeline

The proposed workflow for stability analysis is:

Compute soft SG regions using efficient LMIs (relaxing $P \succeq 0$ ).
Verify that the multipliers are positive-negative.
Use the Hard SG Stability Theorem (strict geometric separation of $SGe^{-1}(H_1)$ and $-SGe(H_2)$ ) to certify stability, relying on the equivalence established in step 2.
For non-circular operators, fit the SG into a tight conic region (e.g., an ellipse) to reduce conservatism.

3. Key Contributions

Soft-Hard Equivalence Theorem: Proved that for positive-negative multipliers, soft SG containment implies hard SG containment. This allows large-scale systems to be certified using only soft computations, eliminating the need for the computationally heavy $P \succeq 0$ constraint.
Conic Region Characterization: Extended SG analysis beyond circles. The paper provides a rigorous characterization of which conic regions are h-convex ( $\Theta_{11} \geq \Theta_{22}$ ) and derives a frequency-domain certificate for containment within these regions.
Tighter Stability Margins: Demonstrated that conic regions (specifically ellipses) can fit elongated, non-symmetric SGs much more tightly than circles, leading to less conservative stability margins.
Computational Efficiency: Developed a hybrid certification pipeline that avoids the homotopy condition (required for soft-only stability) and the semidefinite constraint (required for hard-only computation).

4. Results

Numerical Experiments (Circular):
- Tested on LTI systems with state dimensions up to 300.
- The soft LMI formulation (without $P \succeq 0$ ) achieved computational speedups of 15% to 44% compared to the hard LMI formulation.
- The method successfully certified stability for systems where direct hard SG computation was previously too costly.
Numerical Experiments (Conic):
- Applied to systems with non-symmetric SGs.
- Ellipsoidal regions reduced the containment area by approximately 21% for one system and 9% for another compared to the smallest enclosing disk.
- This reduction directly translates to tighter (less conservative) stability margins.
Scalability: The approach scales effectively to large-scale MIMO systems (e.g., block-diagonal structures modeling power systems or multi-agent networks), making SG-based analysis feasible for problems previously considered intractable.

5. Significance

This paper represents a significant advancement in robust control theory and stability analysis:

Practicality for Large Systems: By removing the $P \succeq 0$ constraint requirement for stability certification, the method makes SG-based analysis viable for large-scale infrastructure systems (like power grids) where current methods are computationally prohibitive.
Reduced Conservatism: The introduction of conic regions allows engineers to model system behaviors that circular regions cannot capture efficiently, leading to more accurate and less conservative stability guarantees.
Unified Framework: It unifies the geometric intuition of soft SGs with the rigorous guarantees of hard SGs, providing a streamlined "best-of-both-worlds" approach for feedback stability analysis.

In summary, the authors provide a mathematically rigorous and computationally efficient framework that expands the geometric capabilities of Scaled Graphs while drastically reducing the computational burden of verifying feedback stability in large-scale systems.