The reviewed record of science sign in
Pith

arxiv: 2606.28557 · v1 · pith:M44JBW2G · submitted 2026-06-26 · math.OC

A New Noise Model for Data-driven Control: Generalized Frobenius Norm Bounds

Huayuan Huang , M. Kanat Camlibel , Henk J. van Waarde This is my paper

Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel 2026-06-30 00:55 UTCgrok-4.3pith:M44JBW2Grecord.jsonopen to challenge →

classification math.OC
keywords data-driven controlnoise modelFrobenius normS-lemmaquadratic stabilizationH2 controlHinf controldissipativity
0
0 comments X

The pith

A generalized Frobenius norm bound on noise samples yields necessary and sufficient conditions for data-driven control design.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces a new noise model that treats the matrix of noise samples through a generalized Frobenius norm bound. This model produces a tighter overapproximation than prior quadratic matrix inequality descriptions when the noise is instantaneously bounded. The authors then use the model to obtain necessary and sufficient conditions for several data-driven control problems and for related analysis tasks. A supporting technical result is a new S-lemma that equates a quadratic matrix inequality to a quadratic inequality on the vectorized noise matrix.

Core claim

The central claim is that the new generalized Frobenius noise model, together with the accompanying S-lemma, supplies necessary and sufficient conditions under which a controller designed from noisy data satisfies quadratic stabilization, H2 performance, or Hinf performance, and likewise supplies conditions for data-driven stabilizability and dissipativity analysis.

What carries the argument

The generalized Frobenius norm bound on the noise-sample matrix, which replaces a quadratic matrix inequality description and is shown to be less conservative for instantaneously bounded noise.

If this is right

  • Necessary and sufficient LMI conditions exist for data-driven quadratic stabilization.
  • The same framework supplies necessary and sufficient conditions for data-driven H2 and Hinf control.
  • Data-driven analysis problems ranging from stabilizability to dissipativity also admit necessary and sufficient conditions under the new noise model.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The reduced conservatism may translate into smaller feasible sets for the controller parameters when the same data are used.
  • The vectorized S-lemma could be applied to other robust-control problems that currently rely on quadratic matrix inequalities.
  • Numerical comparisons on standard benchmark plants would quantify how often the new conditions are strictly less conservative than the older QMI formulation.

Load-bearing premise

The new S-lemma must hold in full generality so that a quadratic inequality on the vectorized noise matrix implies the desired quadratic matrix inequality.

What would settle it

A concrete counterexample consisting of a quadratic inequality on the vectorized noise matrix together with a quadratic matrix inequality that is not implied by it, or a data-driven control instance in which the derived conditions certify a controller that fails on the true plant.

Figures

Figures reproduced from arXiv: 2606.28557 by Henk J. van Waarde, Huayuan Huang, M. Kanat Camlibel.

Figure 1
Figure 1. Figure 1: The RLC circuit. VIII. ILLUSTRATIVE EXAMPLES In this section, we illustrate our theoretical results with three examples. The simulations are conducted in MATLAB, using YALMIP [38] with the solver MOSEK [39]. A. Dissipativity verification In this example, we consider an RLC circuit as depicted in [PITH_FULL_IMAGE:figures/full_fig_p011_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: An inverted pendulum on a cart sets of systems compatible with the data, making it more difficult to find a single stabilizer for all systems in the sets. In addition, the WF(Φϵ)-based method consistently achieves a higher feasibility rate than the Wqmi(Ψϵ)-based method across all five cases. This observation supports the discussion in Section VI-D regarding the reduced conservatism of the proposed approac… view at source ↗
Figure 3
Figure 3. Figure 3: Average H∞ performance as a function of the number of samples T . The blue curve corresponds to the Wqmi(Ψϵ)-based approach, the red curve corresponds to the WF(Φϵ)-based approach, and the black curve shows the optimal performance. is drawn independently from the same distribution. We collect T ∈ {20, 40, . . . , 200} input and state data samples. The entries of the noise samples w(t) are drawn independent… view at source ↗
read the original abstract

In this article, we introduce a new noise model for data-driven control. The model can be interpreted as a generalization of a Frobenius norm bound on the matrix of noise samples. For instantaneously bounded noise, the proposed model provides a less conservative overapproximation than an existing noise model based on a quadratic matrix inequality (QMI). Using the new model, we derive necessary and sufficient conditions for data-driven control. The framework covers a broad class of design problems, including quadratic stabilization, $\mathcal{H}_2$ control and $\mathcal{H}_{\infty}$ control, and is further extended to cover data-driven analysis problems, ranging for stabilizability to dissipativity. A key technical contribution is a new type of S-lemma that offers necessary and sufficient conditions under which a quadratic matrix inequality is implied by a quadratic inequality in vectorized variables.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript introduces a new noise model interpreted as a generalization of the Frobenius norm bound on noise samples. For instantaneously bounded noise this model yields a less conservative over-approximation than existing quadratic-matrix-inequality (QMI) descriptions. Using the model together with a new S-lemma, the authors derive necessary-and-sufficient LMI conditions for data-driven quadratic stabilization, H2 control and H∞ control, and extend the framework to data-driven analysis problems including stabilizability and dissipativity analysis.

Significance. If the new S-lemma holds in the stated generality, the work supplies exact (necessary and sufficient) convex conditions for a broad class of data-driven control and analysis problems under a noise model that is demonstrably less conservative than prior QMI bounds. The technical contribution of the S-lemma itself may be of independent interest for matrix-inequality problems that involve vectorized quadratic forms.

major comments (2)
  1. [Abstract and §3] Abstract and §3: the new S-lemma is asserted to furnish necessary and sufficient conditions under which a QMI is implied by a quadratic inequality in vectorized variables. The necessity direction is load-bearing for every subsequent “necessary and sufficient” claim (quadratic stabilization, H2/H∞ synthesis, dissipativity). The proof must be checked for the case in which the vectorized quadratic form fails to span all matrix directions admitted by the generalized Frobenius noise set; if any such direction is missed, necessity collapses.
  2. [§4–5] §4–5: the conversion of the generalized Frobenius noise model into the LMI conditions for H2 and H∞ control is performed via the S-lemma. Explicit verification that the resulting LMIs remain feasible precisely when the original (non-convex) problem is feasible would be required to substantiate the necessity claim.
minor comments (1)
  1. [Abstract] The phrase “ranging for stabilizability” in the abstract is presumably a typographical error for “ranging from stabilizability.”

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thorough and constructive review of our manuscript. We address the major comments below.

read point-by-point responses

  1. Referee: [Abstract and §3] Abstract and §3: the new S-lemma is asserted to furnish necessary and sufficient conditions under which a QMI is implied by a quadratic inequality in vectorized variables. The necessity direction is load-bearing for every subsequent “necessary and sufficient” claim (quadratic stabilization, H2/H∞ synthesis, dissipativity). The proof must be checked for the case in which the vectorized quadratic form fails to span all matrix directions admitted by the generalized Frobenius noise set; if any such direction is missed, necessity collapses.

    Authors: We appreciate the referee drawing attention to this critical point regarding the necessity in the S-lemma. In the proof of Theorem 3.2, we demonstrate that the quadratic form in the vectorized variables, derived from the generalized Frobenius norm bound, does indeed span all relevant matrix directions. This is achieved by verifying that the set of possible quadratic terms generated by the noise model is fully captured by the vectorization (detailed in the steps leading to equation (3.8)). Consequently, the necessity direction holds without collapse. To further clarify this for readers, we will add a short remark in the revised version. revision: partial

  2. Referee: [§4–5] §4–5: the conversion of the generalized Frobenius noise model into the LMI conditions for H2 and H∞ control is performed via the S-lemma. Explicit verification that the resulting LMIs remain feasible precisely when the original (non-convex) problem is feasible would be required to substantiate the necessity claim.

    Authors: The necessity and sufficiency are inherited from the S-lemma, as the control synthesis problems are equivalently cast as finding a quadratic form that satisfies the QMI implied by the noise model. The derivations in Sections 4 and 5 consist of a sequence of equivalences: the data-driven problem ⇔ nonconvex QMI ⇔ LMI via S-lemma. We believe this provides the required verification. Should the referee identify a specific scenario where this chain breaks, we would welcome the opportunity to address it. revision: no

Circularity Check

0 steps flagged

No circularity: new noise model and S-lemma are independent contributions

full rationale

The paper introduces a generalized Frobenius noise model and derives a new S-lemma as original technical contributions. Necessary-and-sufficient LMI conditions for quadratic stabilization, H2/H∞ control, and dissipativity follow from these elements plus standard matrix-inequality machinery. No step reduces a claimed prediction or first-principles result to a fitted parameter, self-citation chain, or definitional tautology. The derivation is self-contained; external verification of the S-lemma would be a correctness question, not a circularity issue.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim rests on the validity of the new S-lemma and on the interpretation of the generalized Frobenius bound as a valid over-approximation of noise; no free parameters or invented physical entities are visible in the abstract.

axioms (1)
  • standard math Standard algebraic properties of quadratic matrix inequalities and the classical S-lemma hold and can be extended to the vectorized case.
    The paper invokes a new variant of the S-lemma, which presupposes the background theory of quadratic forms and matrix inequalities.
invented entities (1)
  • Generalized Frobenius norm noise model no independent evidence
    purpose: To serve as a less conservative over-approximation of instantaneously bounded noise for data-driven control synthesis.
    The model is introduced in the paper; no external falsifiable prediction or independent measurement is mentioned.

pith-pipeline@v0.9.1-grok · 5682 in / 1369 out tokens · 39306 ms · 2026-06-30T00:55:04.931458+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

40 extracted references · 2 canonical work pages · 1 internal anchor

  1. [1]

    From model-based control to data-driven control: Survey, classification and perspective,

    Z.-S. Hou and Z. Wang, “From model-based control to data-driven control: Survey, classification and perspective,”Information Sciences, vol. 235, pp. 3–35, 2013

    2013

  2. [2]

    Behavioral systems theory in data-driven analysis, signal processing, and control,

    I. Markovsky and F. D ¨orfler, “Behavioral systems theory in data-driven analysis, signal processing, and control,”Annual Reviews in Control, vol. 52, pp. 42–64, 2021

    2021

  3. [3]

    Combining prior knowl- edge and data for robust controller design,

    J. Berberich, C. W. Scherer, and F. Allg ¨ower, “Combining prior knowl- edge and data for robust controller design,”IEEE Transactions on Automatic Control, vol. 68, no. 8, pp. 4618–4633, 2022

    2022

  4. [4]

    Bridging direct and indirect data-driven control formulations via regularizations and relaxations,

    F. D ¨orfler, J. Coulson, and I. Markovsky, “Bridging direct and indirect data-driven control formulations via regularizations and relaxations,” IEEE Transactions on Automatic Control, vol. 68, no. 2, pp. 883–897, 2022

    2022

  5. [5]

    Data-driven control: Part two of two: Hot take: Why not go with models?

    F. D ¨orfler, “Data-driven control: Part two of two: Hot take: Why not go with models?”IEEE Control Systems Magazine, vol. 43, no. 6, pp. 27–31, 2023

    2023

  6. [6]

    H. J. van Waarde, M. K. Camlibel, and H. L. Trentelman,Data-Based Linear Systems and Control Theory, 1st ed. Kindle Direct Publishing,

  7. [7]

    Available: https://henkvanwaarde.github.io/dblsct

    [Online]. Available: https://henkvanwaarde.github.io/dblsct

  8. [8]

    Formulas for data-driven control: Stabilization, optimality, and robustness,

    C. De Persis and P. Tesi, “Formulas for data-driven control: Stabilization, optimality, and robustness,”IEEE Transactions on Automatic Control, vol. 65, no. 3, pp. 909–924, 2020

    2020

  9. [9]

    An informativity approach to the data-driven algebraic regulator problem,

    H. L. Trentelman, H. J. Van Waarde, and M. K. Camlibel, “An informativity approach to the data-driven algebraic regulator problem,” IEEE Transactions on Automatic Control, vol. 67, no. 11, pp. 6227– 6233, 2021

    2021

  10. [10]

    Data-driven model predictive control with stability and robustness guarantees,

    J. Berberich, J. K ¨ohler, M. A. M ¨uller, and F. Allg ¨ower, “Data-driven model predictive control with stability and robustness guarantees,”IEEE Transactions on Automatic Control, vol. 66, no. 4, pp. 1702–1717, 2021. 14 IEEE TRANSACTIONS AND JOURNALS TEMPLATE

    2021

  11. [11]

    Data-driven dissipativity analysis: Application of the matrix S-lemma,

    H. J. Van Waarde, M. K. Camlibel, P. Rapisarda, and H. L. Trentelman, “Data-driven dissipativity analysis: Application of the matrix S-lemma,” IEEE Control Systems Magazine, vol. 42, no. 3, pp. 140–149, 2022

    2022

  12. [12]

    Robust data- driven state-feedback design,

    J. Berberich, A. Koch, C. W. Scherer, and F. Allg ¨ower, “Robust data- driven state-feedback design,” inProceedings of American Control Conference, 2020, pp. 1532–1538

    2020

  13. [13]

    Data-driven quadratic stabilization and LQR control of LTI systems,

    T. Dai and M. Sznaier, “Data-driven quadratic stabilization and LQR control of LTI systems,”Automatica, vol. 153, p. 111041, 2023

    2023

  14. [14]

    Data-driven system analysis of nonlinear systems using polynomial approximation,

    T. Martin and F. Allg ¨ower, “Data-driven system analysis of nonlinear systems using polynomial approximation,”IEEE Transactions on Auto- matic Control, vol. 69, no. 7, pp. 4261–4274, 2024

    2024

  15. [15]

    Decoupling parameter variation from noise: Biquadratic Lyapunov forms in data-driven LPV control,

    C. Verhoek, J. Eising, F. D ¨orfler, and R. T ´oth, “Decoupling parameter variation from noise: Biquadratic Lyapunov forms in data-driven LPV control,” in2024 IEEE 63rd Conference on Decision and Control (CDC). IEEE, 2024, pp. 6761–6766

    2024

  16. [16]

    When sampling works in data-driven control: Informativity for stabilization in continuous time,

    J. Eising and J. Cort ´es, “When sampling works in data-driven control: Informativity for stabilization in continuous time,”IEEE Transactions on Automatic Control, vol. 70, no. 1, pp. 565–572, 2025

    2025

  17. [17]

    Data-driven stabilization of switched and constrained linear systems,

    M. Bianchi, S. Grammatico, and J. Cort ´es, “Data-driven stabilization of switched and constrained linear systems,”Automatica, vol. 171, p. 111974, 2025

    2025

  18. [18]

    From noisy data to feedback controllers: non-conservative design via a matrix S-lemma,

    H. J. van Waarde, M. K. Camlibel, and M. Mesbahi, “From noisy data to feedback controllers: non-conservative design via a matrix S-lemma,” IEEE Transactions on Automatic Control, vol. 67, no. 1, pp. 162–175, 2020

    2020

  19. [19]

    On data-driven control: Informativity of noisy input-output data with cross-covariance bounds,

    T. R. Steentjes, M. Lazar, and P. M. Van den Hof, “On data-driven control: Informativity of noisy input-output data with cross-covariance bounds,”IEEE Control Systems Letters, vol. 6, pp. 2192–2197, 2022

    2022

  20. [20]

    Data-driven control via Petersen’s lemma,

    A. Bisoffi, C. De Persis, and P. Tesi, “Data-driven control via Petersen’s lemma,”Automatica, vol. 145, p. 110537, 2022

    2022

  21. [21]

    Quadratic matrix inequalities with applications to data-based control,

    H. J. van Waarde, M. K. Camlibel, J. Eising, and H. L. Trentelman, “Quadratic matrix inequalities with applications to data-based control,” SIAM Journal on Control and Optimization, vol. 61, no. 4, pp. 2251– 2281, 2023

    2023

  22. [22]

    Controller synthesis for input-state data with measurement errors,

    A. Bisoffi, L. Li, C. De Persis, and N. Monshizadeh, “Controller synthesis for input-state data with measurement errors,”IEEE Control Systems Letters, vol. 8, pp. 1571–1576, 2024

    2024

  23. [23]

    Data Informativity under Data Perturbation

    T. Kaminaga and H. Sasahara, “Data informativity under data perturba- tion,” 2025,arXiv:2505.01641

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  24. [24]

    Data-driven state-feedback controller synthe- sis for dissipativity: A dualization-based approach,

    P. Kristovi ´c and A. Joki ´c, “Data-driven state-feedback controller synthe- sis for dissipativity: A dualization-based approach,” in2024 American Control Conference (ACC). IEEE, 2024, pp. 1219–1224

    2024

  25. [25]

    Dissipativity-based data-driven decentralized control of interconnected systems,

    T. Nakano, A. Aboudonia, J. Eising, A. Martinelli, F. D ¨orfler, and J. Lygeros, “Dissipativity-based data-driven decentralized control of interconnected systems,” 2025,arXiv:2505.14047

    work page arXiv 2025

  26. [26]

    Data-driven analysis and design beyond common lyapunov functions,

    H. J. Van Waarde, M. K. Camlibel, and H. L. Trentelman, “Data-driven analysis and design beyond common lyapunov functions,” in2022 IEEE 61st Conference on Decision and Control (CDC). IEEE, 2022, pp. 2783–2788

    2022

  27. [27]

    Data-driven criteria for detectability and observer design for LTI systems,

    V . K. Mishra, H. J. Van Waarde, and N. Bajcinca, “Data-driven criteria for detectability and observer design for LTI systems,” in2022 IEEE 61st Conference on Decision and Control (CDC). IEEE, 2022, pp. 4846–4852

    2022

  28. [28]

    Provably robust verification of dissipativity properties from data,

    A. Koch, J. Berberich, and F. Allg ¨ower, “Provably robust verification of dissipativity properties from data,”IEEE Transactions on Automatic Control, vol. 67, no. 8, pp. 4248–4255, 2021

    2021

  29. [29]

    Dissipativity verification with guarantees for polynomial systems from noisy input-state data,

    T. Martin and F. Allg ¨ower, “Dissipativity verification with guarantees for polynomial systems from noisy input-state data,” in2021 American Control Conference (ACC). IEEE, 2021, pp. 3963–3968

    2021

  30. [30]

    Trade-offs in learning controllers from noisy data,

    A. Bisoffi, C. De Persis, and P. Tesi, “Trade-offs in learning controllers from noisy data,”Systems & Control Letters, vol. 154, p. 104985, 2021

    2021

  31. [31]

    A survey of the S-lemma,

    I. P ´olik and T. Terlaky, “A survey of the S-lemma,”SIAM review, vol. 49, no. 3, pp. 371–418, 2007

    2007

  32. [32]

    Data-driven control of positive linear systems using linear programming,

    J. Miller, T. Dai, M. Sznaier, and B. Shafai, “Data-driven control of positive linear systems using linear programming,” in2023 62nd IEEE Conference on Decision and Control (CDC). IEEE, 2023, pp. 1588– 1594

    2023

  33. [33]

    A semi-algebraic optimization approach to data- driven control of continuous-time nonlinear systems,

    T. Dai and M. Sznaier, “A semi-algebraic optimization approach to data- driven control of continuous-time nonlinear systems,”IEEE Control Syst. Lett., vol. 5, no. 2, pp. 487–492, 2020

    2020

  34. [34]

    Recursive state estimation for a set- membership description of uncertainty,

    D. Bertsekas and I. Rhodes, “Recursive state estimation for a set- membership description of uncertainty,”IEEE Transactions on Auto- matic Control, vol. 16, no. 2, pp. 117–128, 1971

    1971

  35. [35]

    System identification via membership set constraints with energy constrained noise,

    E. Fogel, “System identification via membership set constraints with energy constrained noise,”IEEE Transactions on Automatic Control, vol. 24, no. 5, pp. 752–758, 1979

    1979

  36. [36]

    Optimal kernel regression bounds under energy-bounded noise,

    A. Lahr, J. K ¨ohler, A. Scampicchio, and M. Zeilinger, “Optimal kernel regression bounds under energy-bounded noise,”Advances in Neural Information Processing Systems, vol. 38, pp. 100 411–100 444, 2026

    2026

  37. [37]

    S. Boyd, L. El Ghaoui, E. Feron, and V . Balakrishnan,Linear matrix inequalities in system and control theory. Philadelphia, PA, USA: SIAM, 1994

    1994

  38. [38]

    D. S. Bernstein,Matrix mathematics: theory, facts, and formulas. Princeton university press, 2009

    2009

  39. [39]

    Yalmip: A toolbox for modeling and optimization in matlab,

    J. L ¨ofberg, “Yalmip: A toolbox for modeling and optimization in matlab,” inProceedings of International Conference on Robotics and Automation, 2004, pp. 284–289

    2004

  40. [40]

    Version 10.0.30., 2022

    MOSEK ApS,The MOSEK optimization toolbox for MATLAB manual. Version 10.0.30., 2022

    2022