The Phantom of Davis-Wielandt Shell: A Unified Perspective for Graphical Stability Analysis of MIMO LTI Systems

arXiv:2507.19918 · eess.SY, cs.SY, math.OC, math.RA · Submitted 2025-07-26 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.

Rosa: We're looking at the paper: "The Phantom of Davis-Wielandt Shell".

Dev: Here's the gist of the paper. This paper presents a unified framework based on Davis-Wielandt (DW) shell for graphical stability analysis of multi-input and multi-output linear timeinvariant feedback systems The proposed rotated scaled relative graph (θSRG) concept is shown to be the least conservative closed-loop stability criterion among existing 2-D graphical conditions for bi-component feedback loops Unified Framework and Geometric Intuition The framework establishes connections between DW shells and various graphical representations, as well as gain and phase measures through an intuitive geometric perspective This approach allows for the mapping of relationships and relative conservatism among various separation conditions The core idea is to leverage the DW shell to build a unified framework that derives aforementioned sets and conditions via intuitive geometric transformations The Rotated Scaled Relative Graph (θ-SRG) The paper proposes the concept of θ-SRG, which is a mixed gain-phase representation This concept is derived from the DW shell by considering mappings such as SRGθ(A):= e iθSRG(e-iθA) The θ-SRG yields the least conservative stability condition among all for bi-component feedback systems Characterization of Separation Conditions The paper uncovers connections between DW shells and various lower-dimensional graphical representations, examining the respective emphasis and conservatism of different 2-D graphical stability conditions The separation of DW shells of two matrices characterizes the nonsingularity of the unitary orbits formed around these two matrices Theorem IV.1 provides a direct characterization of the nonsingularity of I + AU HBU in terms of DW shell separation Numerical Characterization and Visualization The paper develops an SDP-based algorithm for plotting rotated SRGs, which offers insights into visualizing other members of the DW family of graphical representations This algorithm is introduced in Algorithm 1 for plotting θ-SRG Furthermore, the concept of texture is proposed as a generalization of convexity to characterize non-convex sets numerically Stability Conditions and Implications The paper derives a set of sufficient graphical conditions for the nonsingularity of I + A∆B in Corollary 1 These conditions include small gain, large gain, numerical range, sectorial phase, SRG phase, normalized numerical range, segmental phase, and scaled relative graph The final result is the uniparameter θ-SRG phase condition in Corollary 2 which is equivalent to the biparameter condition in Corollary 1 Stability Verification via Frequencywise Analysis The final section demonstrates how the conditions can be used for closed-loop stability by relating them to frequencywise analysis of LTI systems Corollary 4 provides a sufficient condition for internal stability based on the separation of SRG-1θ(ω)+(G(iω)) and -

µ∈[0,1: µSRGθ(ω)−(H(iω))] This leads to Corollary 5, which provides an explicit mixed gain-phase condition The authors conclude by showing that the DW shell perspective provides a cohesive interpretation of existing stability criteria and reveals their relative conservatism Future work includes extending the proposed framework to nonlinear systems, as well as to semistable linear systems using Davis’s original shell defined on linear relations The paper concludes with an example demonstrating a gap between Corollary 4 and Corollary 5.

Rosa: Pruning redundant reasoning tokens The results in Sections III and IV are summarized in Figure 2 This figure illustrates the connection between DW shells and lower-dimensional representations (gray boxes), including various 2-D sets, such as numerical range and SRG From this perspective, in this section we introduce the concept of θ-SRGs and study their basic properties How the training signal is built The mappings to lower-dimensional spaces are all well-defined functions, and separation in codomain implies separation of their preimages Table III provides explicit formulas for these inverse sets in terms of the representations of the original matrix Where the gains come from The gain and phase conditions discussed above can all be interpreted as using 3-D regions (see Figure 6) to cover (or ‘approximate’) the DW−1(A) and DW(−B) The θ-SRG emerges as a natural candidate for 2-D mixed gain-phase representation Pruning redundant reasoning tokens The authors thank Di Zhao (Nanjing University) and Wei Chen (Peking University) for helpful discussions The paper concludes with an example of interconnection that falls within the gap between Corollary 4 and Corollary 5 How the training signal is built The construction of the DW tomography-based algorithm for plotting θ-SRGs involves solving a sequence of SDPs The computation of the endpoints of the segment can be formulated as a DW shell tomography problem and solved by SDP solvers Where the gains come from The analysis shows that for normal matrices, their DW shells are convex hulls of the lifted spectrum points The boundary points of their θ-SRGs all stem from line segments in 3-D connecting these lifted points Pruning redundant reasoning tokens The authors thank Di Zhao (Nanjing University) and Wei Chen (Peking University) for helpful discussions The paper concludes with an example demonstrating a gap between Corollary 4 and Corollary 5 How the training signal is built The relationship among the main results proposed in this section are as follows The final result is that by searching over θ ∈ [0, π) (essentially adding back the missing dimension), separation in terms of θ-SRGs fully recovers the DW separation condition Where the gains come from The spectral containment property of SRGθ(A) emerges as a natural consequence The gain and phase conditions discussed above all leverage specific µ-independent sets to overapproximate the two sets to be separated in Corollary 4 This avoids an explicit contractive parameter µ in the conditions Pruning redundant reasoning tokens The paper concludes with an example demonstrating a gap between Corollary 4 and Corollary 5 The final result is that by searching over θ ∈ [0, π) (essentially adding back the missing dimension), separation in terms of θ-SRGs fully recovers the DW separation condition Where the gains come from The gain and phase conditions discussed above all leverage specific µ-independent sets to overapproximate the two sets to be separated in Corollary 4 This avoids an explicit contractive parameter µ in the conditions How the training signal is built The relationship among the main results proposed in this section are as follows The final result is that by searching over θ ∈ [0, π) (essentially adding back the missing dimension), separation in terms of θ-SRG phase conditions fully recovers the DW separation condition Where the gains come from

Title and authors: Rosa: Let’s talk about the people who wrote this, Ding Zhang, Xiaokan Yang, Axel Ringh, Li Qiu. They are doing this unified framework using the Davis-Wielandt shell for analyzing multi-input and multi-output linear timeinvariant feedback systems.

Dev: I’ve seen a lot of work on stability analysis with matrix methods before, but this paper seems to be building something that ties the underlying linear algebra structure—the DW shell—directly into graphical representations and specific measures like gain and phase.

Taro: I wonder if this unified perspective is accessible enough for someone who isn't deeply immersed in abstract linear algebra to actually use it when they are testing real-world failure modes, you know, the messy stuff.

Rosa: Well, the authors suggest that by using this DW shell approach, they can build connections between these various graphical representations and gain or phase measures through an intuitive geometric perspective. This is about mapping out relationships and relative conservatism among different separation conditions.

Dev: So they are taking abstract properties of these matrix shells—like how far apart two shells are—and translating that into something visual that we can actually use to assess stability in a closed-loop system, right?

Taro: If it’s truly geometric, maybe it gives us a better intuition than just crunching numbers and looking at standard charts. It moves the understanding from purely algebraic to something more spatial.

Rosa: Right. And they propose this rotated scaled relative graph, or theta SRG concept, as a mixed gain-phase representation that leads directly to the stability criterion they say is the least conservative among existing 2D graphical conditions for bi-component feedback loops <ref:2507.19918#pg1,graphical conditions for bi-component feedback loops>.

Dev: That’s the big payoff—a single condition that covers a whole family of others. It sounds like they are using this new theta SRG to give us a more efficient way to define safety margins in control loops.

The paper's summary: Rosa: So, what does the paper actually summarize? Essentially, they show how the separation of Davis-Wielandt shells between two matrices characterizes the nonsingularity of I plus A B HBU. This is a direct way to check if a closed-loop system is stable or not.

Dev: That’s interesting because it connects this shell separation directly to an important mathematical property, which means we can use this geometric separation to estimate the spectrum of A B by looking at how far apart those shells are.

Taro: If you can use shell separation to find things like the eigenvalues or the spectrum, that’s valuable for understanding what happens internally in a system before you even run a simulation. It gives us insight into the structure itself.

Rosa: The paper also talks about how this shell separation helps characterize different lower-dimensional graphical representations—the 2D conditions—examining their emphasis and conservatism <ref:2507.19918#pg1>. They show connections between DW shells and things like the numerical range or SRG, which tells you which one is more conservative for a given system setup.

Dev: So they are not just giving us a new tool; they’re comparing this tool to the old ones to say, "Hey, this new geometric way of looking at it actually gives us a better bound on stability than these other common ways."

Taro: That comparison is key. If you can quantify *why* one method is better than another—for instance, showing that the theta SRG condition is superior—that helps us choose the right analysis tool for the specific type of system we're modeling.

The paper's improvements: Rosa: The improvements they propose are pretty substantial. They suggest a set of sufficient graphical conditions for the nonsingularity of I plus A Delta B, which includes things like small gain, large gain, numerical range, and scaled relative graph.

Dev: That’s a lot of conditions bundled up into one result in Corollary one <ref:2507.19918#pg1>. And they culminate with this final result in Corollary two: the uniparameter theta SRG phase condition which is equivalent to that biparameter condition mentioned earlier <ref:2507.19918#pg1>.

Taro: So they are saying that by focusing on this single theta SRG phase condition, you can get all the information from those other more complex conditions. That’s a big simplification for real-time analysis where you need quick checks.

Rosa: They also developed an SDP-based algorithm to plot these rotated SRGs, which gives us a way to visualize other members of the Davis-Wielandt family of graphical representations, and they even proposed a concept called texture as a way to generalize convexity for characterizing non-convex sets numerically.

Dev: So the visualization part is practical. Having an algorithm that can actually plot these theta SRGs using semidefinite programs means researchers can see how these geometric concepts map out in three dimensions, which is much harder to do manually.

Taro: And that texture concept for non-convex sets—that sounds like a way to handle the messy, real-world shapes of stability regions that don't fit nice smooth mathematical boundaries. It’s about characterizing things when they aren't perfectly convex.

Conclusion: Rosa: So, wrapping up the whole thing with "The Phantom of Davis-Wielandt Shell: A Unified Perspective for Graphical Stability Analysis of MIMO LTI Systems," the main implication is that this DW shell perspective provides a cohesive interpretation of existing stability criteria and reveals their relative conservatism.

Dev: It shows that we can use geometric transformations to derive these sets and conditions intuitively, which helps us understand the relationship between different analytical tools without having to re-derive everything from scratch every time.

Taro: For me, it’s about having a systematic way to look at stability checks that doesn't just rely on picking one method arbitrarily; it gives us a hierarchy of methods based on how conservative they are and what geometric shape they represent.

Rosa: And for the listener, it means we have this new theta SRG phase condition which is equivalent to the biparameter condition, giving us a clear path toward verifying internal stability using frequencywise analysis.

Dev: Before we wrap up this one, Taro, what’s your take on how this framework fits into the bigger picture of control engineering problems?

Taro: I think it provides a rigorous way to check the system structure itself—the underlying mathematical relationship—rather than just checking if a certain gain or phase value happens to work for one specific test case.

Rosa: It gives us that structural understanding, and it's all tied together by this unified framework based on the Davis-Wielandt shell. That’s the big picture here.

Dev: Yeah, it’s a solid piece of work connecting deep linear algebra to practical stability assessment for MIMO systems. We'll keep an eye on how they extend this concept to nonlinear systems next.

Peking University · Chalmers University of Technology and University of Gothenburg · The Chinese University of Hong Kong, Shenzhen

eess.SY, cs.SY, math.OC, math.RA

Submitted: 2025-07-26

Updated: 2026-11-10

Comments: 16 pages, 10 figures. Revised version

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 87/100

The gist: This paper presents a unified framework based on Davis-Wielandt (DW) shell for graphical stability analysis of multi-input and multi-output linear timeinvariant feedback systems The proposed rotated…

Key concepts

Davis-Wielandt (DW) Shell
The DW shell is a mathematical tool used to characterize the set of matrices that satisfy certain stability conditions. It provides a geometric way to visualize and understand the relationships between different matrix properties, such as those related to closed-loop stability, by mapping them onto specific regions in space.
Rotated Scaled Relative Graph ($ heta$-SRG)
The $ heta$-SRG is a mixed gain-phase representation derived from the DW shell. It is proposed as the least conservative stability criterion among existing 2-D graphical conditions for bi-component feedback systems, offering a comprehensive view combining both gain and phase information in a single concept.
Separation of DW Shells
The separation of two DW shells directly characterizes the nonsingularity of the expression $I + A riangle B$. This separation condition is crucial for determining the necessary and sufficient conditions for closed-loop stability in MIMO LTI systems by relating it to matrix properties.
Texture (Numerical Characterization)
Texture is introduced as a generalization of convexity used to numerically characterize sets, such as SRGs. It helps determine if a set is non-convex or convex by identifying specific matrix independent textures, providing a robust method for numerical analysis beyond simple geometric shapes.

Terminology

Summary

This paper presents a unified framework based on Davis-Wielandt (DW) shell for graphical stability analysis of multi-input and multi-output linear timeinvariant feedback systems The proposed rotated scaled relative graph (θSRG) concept is shown to be the least conservative closed-loop stability criterion among existing 2-D graphical conditions for bi-component feedback loops

Unified Framework and Geometric Intuition

The framework establishes connections between DW shells and various graphical representations, as well as gain and phase measures through an intuitive geometric perspective This approach allows for the mapping of relationships and relative conservatism among various separation conditions The core idea is to leverage the DW shell to build a unified framework that derives aforementioned sets and conditions via intuitive geometric transformations

The Rotated Scaled Relative Graph (θ-SRG)

The paper proposes the concept of θ-SRG, which is a mixed gain-phase representation This concept is derived from the DW shell by considering mappings such as SRGθ(A):= e iθSRG(e-iθA) The θ-SRG yields the least conservative stability condition among all for bi-component feedback systems

Characterization of Separation Conditions

The paper uncovers connections between DW shells and various lower-dimensional graphical representations, examining the respective emphasis and conservatism of different 2-D graphical stability conditions The separation of DW shells of two matrices characterizes the nonsingularity of the unitary orbits formed around these two matrices Theorem IV.1 provides a direct characterization of the nonsingularity of I + AU HBU in terms of DW shell separation

Numerical Characterization and Visualization

The paper develops an SDP-based algorithm for plotting rotated SRGs, which offers insights into visualizing other members of the DW family of graphical representations This algorithm is introduced in Algorithm 1 for plotting θ-SRG Furthermore, the concept of texture is proposed as a generalization of convexity to characterize non-convex sets numerically

Stability Conditions and Implications

The paper derives a set of sufficient graphical conditions for the nonsingularity of I + A∆B in Corollary 1 These conditions include small gain, large gain, numerical range, sectorial phase, SRG phase, normalized numerical range, segmental phase, and scaled relative graph The final result is the uniparameter θ-SRG phase condition in Corollary 2 which is equivalent to the biparameter condition in Corollary 1

Stability Verification via Frequencywise Analysis

The final section demonstrates how the conditions can be used for closed-loop stability by relating them to frequencywise analysis of LTI systems Corollary 4 provides a sufficient condition for internal stability based on the separation of SRG-1θ(ω)+(G(iω)) and -[µ∈[0,1]µSRGθ(ω)−(H(iω))] This leads to Corollary 5, which provides an explicit mixed gain-phase condition

Conclusion and Future Work

The authors conclude by showing that the DW shell perspective provides a cohesive interpretation of existing stability criteria and reveals their relative conservatism Future work includes extending the proposed framework to nonlinear systems, as well as to semistable linear systems using Davis’s original shell defined on linear relations The paper concludes with an example demonstrating a gap between Corollary 4 and Corollary 5.

Pruning redundant reasoning tokens

The results in Sections III and IV are summarized in Figure 2 This figure illustrates the connection between DW shells and lower-dimensional representations (gray boxes), including various 2-D sets, such as numerical range and SRG From this perspective, in this section we introduce the concept of θ-SRGs and study their basic properties

How the training signal is built

The mappings to lower-dimensional spaces are all well-defined functions, and separation in codomain implies separation of their preimages Table III provides explicit formulas for these inverse sets in terms of the representations of the original matrix

Where the gains come from

The gain and phase conditions discussed above can all be interpreted as using 3-D regions (see Figure 6) to cover (or ‘approximate’) the DW−1(A) and DW(−B) The θ-SRG emerges as a natural candidate for 2-D mixed gain-phase representation

The authors thank Di Zhao (Nanjing University) and Wei Chen (Peking University) for helpful discussions The paper concludes with an example of interconnection that falls within the gap between Corollary 4 and Corollary 5

The construction of the DW tomography-based algorithm for plotting θ-SRGs involves solving a sequence of SDPs The computation of the endpoints of the segment can be formulated as a DW shell tomography problem and solved by SDP solvers

The analysis shows that for normal matrices, their DW shells are convex hulls of the lifted spectrum points The boundary points of their θ-SRGs all stem from line segments in 3-D connecting these lifted points

The authors thank Di Zhao (Nanjing University) and Wei Chen (Peking University) for helpful discussions The paper concludes with an example demonstrating a gap between Corollary 4 and Corollary 5

The relationship among the main results proposed in this section are as follows The final result is that by searching over θ ∈ [0, π) (essentially adding back the missing dimension), separation in terms of θ-SRGs fully recovers the DW separation condition

The spectral containment property of SRGθ(A) emerges as a natural consequence The gain and phase conditions discussed above all leverage specific µ-independent sets to overapproximate the two sets to be separated in Corollary 4 This avoids an explicit contractive parameter µ in the conditions

The paper concludes with an example demonstrating a gap between Corollary 4 and Corollary 5 The final result is that by searching over θ ∈ [0, π) (essentially adding back the missing dimension), separation in terms of θ-SRGs fully recovers the DW separation condition

The gain and phase conditions discussed above all leverage specific µ-independent sets to overapproximate the two sets to be separated in Corollary 4 This avoids an explicit contractive parameter µ in the conditions

The relationship among the main results proposed in this section are as follows The final result is that by searching over θ ∈ [0, π) (essentially adding back the missing dimension), separation in terms of θ-SRG phase conditions fully recovers the DW separation condition

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Unified Graphical Stability Analysis Framework: The system can establish intuitive geometric connections between various graphical representations, such as Nyquist plots, eigenloci of return ratios, and singular value-based gain measures through the Davis-Wielandt shell concept.

  2. Least Conservative Stability Criterion: The framework proposes the rotated scaled relative graph (theta-SRG) concept to derive a closed-loop stability criterion that is shown to be the least conservative among the existing 2-D graphical conditions for bi-component feedback loops.

  3. Generalizable Visualization Algorithm: A reliable and generalizable algorithm is proposed for visualizing theta-SRGs, including standard SRGs as a special case, using techniques derived from DW shells.

  4. Robust Stability Condition Derivation: The system can derive the DW separation condition, which serves as a necessary and sufficient characterization of closed-loop stability for MIMO LTI systems by relating it to the nonsingularity of I + AB.

  5. Spectrum Estimation: The system can estimate the spectrum of AB by utilizing separation theorems derived from DW shells, such as if DW−1(A) ∩ DW(−B) = ∅, then-1 ∈/ Λ (AB).

  6. Mixed Gain-Phase Condition Generation: The system can generate a frequencywise theta-SRG condition that serves as a sufficient condition for internal stability by combining gain and phase measures.

  7. Frequency Dependent Stability Testing: For negative feedback systems with LTI components G(s) and H(s), the system can test closed-loop stability by checking if for each omega in R+ ∪ 0, there exists a theta(omega) in R such that SRG−1 theta(omega)+(G(i omega)) ∩ −[mu in [0,1] mu SRGtheta(omega)−(H(i omega)) = ∅.

  8. Numerical Characterization of Sets: The system can compute the boundary points of DW shells using a DW tomography-based algorithm that involves solving a sequence of semidefinite programs (SDPs).

  9. Texture-Based Non-Convexity Handling: The system can characterize non-convex sets, such as SRGs, by identifying matrix independent textures like Ltheta and Ohei theta, which are used to determine if a set is numerically characterizable.

Sources

Related papers