Pith. sign in

REVIEW 3 major objections 5 minor 60 references

Data-Driven Stochastic Control: Foundations and Guarantees

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Noisy finite-horizon trajectories suffice to synthesize safety controllers with quantified probabilistic guarantees for unknown stochastic systems.

desk verdict A plausible new data-driven synthesis pipeline for stochastic safety, but the advertised distribution-free confidence guarantee rests on a Gaussian fourth-moment calculation and is not established. read the letter →

arxiv 2507.23280 v2 pith:LOOAMGFR submitted 2025-07-31 eess.SY cs.SY

Abolfazl Lavaei This is my paper
classification eess.SYcs.SY
keywords data-drivencontrolstochasticbarriercertificatesnon-i.i.d.trajectoriessum-of-squaresoptimizationprobabilisticsafetyguaranteesunknownnonlinearsystemsmatrixS-lemmaprocessnoise
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper establishes that safety controllers for stochastic nonlinear control systems can be synthesized directly from noisy input-state trajectories, without identifying the system matrices $A$ and $B$. The framework allows process noise with arbitrary distribution, known only through upper bounds $\Gamma_\mu$ and $\Gamma_\Sigma$ on its mean and covariance, and the trajectories are temporally dependent rather than i.i.d. The design conditions are cast as a sum-of-squares (SOS) feasibility problem, a tractable polynomial optimization formulation, and a feasible solution yields a quadratic stochastic control barrier certificate $B(x)=x^\top P x$ together with a state-feedback controller $u=K(x)x$. If the conditions hold, the safety probability is quantified as at least $1-\beta_1$ and the certificate itself is correct with confidence at least $1-\beta_2=1-T\bar\beta_2$. A sympathetic reader would care because this moves formal safety guarantees from model-based and worst-case bounded-disturbance settings to a purely data-driven stochastic setting, and in one benchmark the robust analysis is infeasible while the stochastic certificate succeeds.

What carries the argument

The load-bearing machinery is a quadratic stochastic control barrier certificate $B(x)=x^\top P x$, a Lyapunov-like function whose expected one-step growth is controlled. To synthesize it from data, the paper builds the data-conformity matrix $R_{\mathrm{DC}}^j$ in (15) from the empirical second moments $\frac{1}{N}\sum_{i=1}^N \vec X_j^i \vec X_j^{i\top}$ and the noise bounds; with confidence $1-\bar\beta_2$ it characterizes every $\Phi=[A\;B]$ consistent with the data. A matrix S-lemma eliminates the unknown $\Phi$ and turns the stochastic barrier condition into the linear matrix inequality (13c), with multipliers $\alpha_j(x)$. The concentration estimate in Lemma 1 uses a Gaussian fourth-moment identity to bound $\mathbb{E}[\|\varsigma\|^4]$ in terms of $\Gamma_\mu$ and $\Gamma_\Sigma$, and Boole's inequality converts the per-step confidence into the horizon-wide confidence $\beta_2=T\bar\beta_2$.

What would settle it

Repeat the data-collection experiment in Algorithm 1 with a heavy-tailed noise (for example, Student-$t$ with five degrees of freedom) scaled to satisfy the same covariance bound $\Gamma_\Sigma$, and count how often the data-conformity inequality (10) fails across many independent data sets; if the failure frequency exceeds $\bar\beta_2$, the Gaussian fourth-moment calculation in Lemma 1 is not distribution-free and the stated confidence is not valid for that noise.

Watch

Extended reading notes

Core claim

The central claim is Theorem 2: under Assumptions 1-3, for a discrete-time stochastic polynomial system $x^+=AF(x)+BG(x)u+\varsigma$ with unknown $A,B$, known dictionaries $F,G$, and noise statistics satisfying $\mu\mu^\top\preceq\Gamma_\mu$ and $\Sigma\preceq\Gamma_\Sigma$, feasibility of the SOS/LMI conditions (13a)-(13c) implies that $B(x)=x^\top P x$ with $P=\bar P^{-1}$ is a stochastic control barrier certificate and $u=\bar K(x)P x$ is its safety controller, with $\psi=(1+\rho^{-1})\mathrm{Tr}(P\Gamma_\mu)+\mathrm{Tr}(P\Gamma_\Sigma)$ and confidence at least $1-\beta_2=1-T\bar\beta_2$. The certificate then gives the finite-horizon safety probability $1-\beta_1$ via Theorem 1. The inputs to the condition are only the empirical average of $N$ noisy trajectories and the noise-moment bounds; the matrices $A$ and $B$ are never estimated.

Load-bearing premise

The certified confidence rests on a bound on the noise's fourth moment that the proof computes exactly only for Gaussian noise, so for heavier-tailed distributions the same confidence level is not established by the paper.

Editorial extensions

If this is right

  • A feasible solution of (13) certifies before deployment that with probability at least $1-\beta_1$ trajectories starting in $X_\eta$ avoid $X_\delta$ over the horizon $T$, and this guarantee itself holds with confidence at least $1-T\bar\beta_2$.
  • The scenario approach's i.i.d. requirement is relaxed: the state trajectories may be temporally dependent, and the only independent randomness across the $N$ runs is the process noise, because the input sequence is fixed during data collection.
  • Noise with unbounded support is admissible whenever upper bounds on the first two moments are known, which is outside the reach of robust methods that require a bounded disturbance set.
  • The number of realizations required for a chosen confidence is explicit, $N\ge \bar\beta_2^{-1}\epsilon^{-2}(\mathrm{Tr}(\Gamma_\Sigma^2)+(\mathrm{Tr}\Gamma_\Sigma)^2+2\lambda_{\max}(\Gamma_\Sigma)\mathrm{Tr}(\Gamma_\mu)+2\mathrm{Tr}(\Gamma_\Sigma)\mathrm{Tr}(\Gamma_\mu))$, and the data-uncertainty layer $\bar\beta_2$ vanishes as $N\to\infty$.
  • In the spacecraft case study the robust worst-case analysis with $\varsigma\varsigma^\top\preceq\kappa^2 I_3$ finds no controller, while the stochastic formulation with the same noise bounds yields a controller and quantified guarantees, showing when the probabilistic relaxation converts an infeasible problem into a feasible one.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If Lemma 1 were re-proved with a generic bound on $\mathbb{E}[\|\varsigma\|^4]$ rather than the Gaussian identity, the same S-lemma machinery would carry over to heavy-tailed noise; the paper remarks that this replacement is possible but gives no general formula.
  • Because the Chebyshev route gives only polynomial decay of $\bar\beta_2$ in $N$, swapping in a matrix-Bernstein concentration inequality for sub-Gaussian or bounded noise would yield exponential sample-complexity improvements while leaving the synthesis conditions unchanged.
  • The same data-conformity and S-lemma template could plausibly be adapted to input-output data by replacing state measurability with an observability dictionary, since the inequalities only use empirical covariance blocks of trajectory segments.
  • One could turn the feasibility problem into an optimization by maximizing the certified safety probability $1-\beta_1$ over the level-set parameters, at the cost of the bilinear couplings the paper identifies.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes a data-driven, trajectory-based framework for synthesizing stochastic control barrier certificates (S-CBC) and safety controllers for discrete-time stochastic nonlinear polynomial systems with unknown system matrices A and B and process noise of arbitrary distribution. Data are collected as N finite-horizon trajectories generated from the same initial condition and input sequence under different noise realizations. The main result, Theorem 2, provides SOS conditions (13) whose feasibility implies, with confidence at least 1 - T * beta_bar_2, that B(x) = x^T P x is an S-CBC and u = K(x)x is its safety controller, where beta_bar_2 is defined in Lemma 1 using only upper bounds Gamma_mu, Gamma_Sigma on the noise mean and covariance. The paper also includes three simulation studies (Lorenz, Chen, spacecraft) and a comparison with robust worst-case analysis.

Significance. If the main theorem were correct as stated, the paper would make a useful contribution: it extends trajectory-based data-driven safety synthesis to stochastic systems with unbounded noise support, using only first- and second-moment bounds, and it demonstrates via a spacecraft example that stochastic analysis can succeed where robust worst-case analysis fails. The algorithmic pipeline (Algorithm 1) is concrete, the SOS formulation is tractable, and the experiments provide a useful sanity check. However, the central distribution-free claim is not supported by the proof: Lemma 1, which is the statistical foundation of Theorem 2, is derived under a Gaussian assumption and no distribution-free replacement is supplied. This is a load-bearing gap, not a matter of conservatism.

major comments (3)
  1. [Lemma 1 (Eq. 9) and Remark 9] Lemma 1 is stated for a Gaussian random vector and its proof computes E[||zeta||^4] using the Gaussian fourth-moment identity (20a). The statement of Theorem 2 and Problem 1, however, promise arbitrary noise distributions. Remark 9 asserts distribution-freeness because only Markov/Chebyshev inequalities are used, but it does not provide any upper bound on E[||zeta||^4] in terms of the assumed bounds Gamma_mu and Gamma_Sigma, and no such general bound exists: the fourth moment is not controlled by the first two moments. For example, for scalar zero-mean noise with variance 1, take Z = +/-a with probability p/2 each and Z = 0 otherwise, with a = 1/sqrt(p); then E[Z^4] = 1/p, which is unbounded as p -> 0. For fixed N and epsilon, choosing p sufficiently small makes the actual probability of a large deviation exceed the Gaussian-based beta_bar_2 = 2/(N epsilon^2). Consequently, the data-conformity constraint (10) in Lemma 2 and the certified confidence 1 - beta_2 in Theorem 2 are not guaranteed for arbitrary noise distributions. This gap is inherited by every downstream result and directly invalidates the paper's central claim.
  2. [Proof of Theorem 2, probabilistic argument] The proof's event decomposition is incorrect. It defines E1 = {inequality (34) holds} and E2 = {inequality (38) holds}, then asserts P(not E1) = 0 'since E1 is a deterministic inequality and holds true.' But (34) is a condition involving the unknown matrix Phi = [A B]; it is not a priori true. The correct reasoning is that (13c) together with E2 implies E1, so E2 is a subset of E1 and P(E1 intersect E2) = P(E2) >= 1 - beta_2. As written, the proof attributes zero failure probability to an event that is actually guaranteed only conditional on the data event E2. The final confidence claim can be repaired by replacing this step, but the current proof is formally invalid.
  3. [Assumption 2 vs. Lemma 1] Assumption 2 only postulates upper bounds on the mean and covariance. For non-Gaussian noise the fourth moment E[||zeta||^4] is needed to evaluate the variance in (19), and no such bound is assumed or derived. Thus even the corrected statement of Lemma 1 requires an additional assumption (e.g., known fourth-moment bound) that is absent from the problem formulation. This is a mismatch between the assumptions stated in Theorem 2 and the statistical tool used in its proof.
minor comments (5)
  1. [Remark 9] The phrase 'provided the fourth moment exists' is insufficient: existence alone does not give a numeric upper bound, so beta_bar_2 in (9) cannot be evaluated for a general distribution. The text should either restrict the scope to Gaussian noise or explicitly add a fourth-moment bound to Assumption 2 and revise (9) accordingly.
  2. [Proof of Lemma 2, Eq. (25)-(26)] The implication from the Frobenius-norm bound to the spectral-norm bound is correct, but the notation in (26) uses a strict inequality '<' while the conclusion (10) uses a non-strict '<='. Since Lemma 1 gives strict inequality with probability at least 1 - beta_bar_2, the final non-strict inequality follows; the proof should state this explicitly to avoid confusion.
  3. [Theorem 2 statement, Eq. (13c)] The text says (13c) becomes an LMI after 'selecting appropriate scalar values for rho > 0 and kappa in (0,1)', but rho appears as a variable in the term -(1+rho)^{-1} \bar{P}. Unless rho is fixed before solving, the condition is not an LMI. Please clarify that rho is fixed a priori or handled by line search.
  4. [Proof of Theorem 2, Eq. (30)-(31)] There is a mismatched parenthesis in 'Tr(P mu mu^top)' (the opening parenthesis after 'Tr' is not closed). In addition, the trace arguments use the cyclic property, but the presentation would benefit from a one-line derivation of Tr(P mu mu^top) <= Tr(P Gamma_mu) from mu mu^top <= Gamma_mu, since that inequality is used repeatedly.
  5. [Section IV, trade-off discussion] The lower bound on N given in Section IV contains the same expression as beta_bar_2 and is correct only if the Gaussian or fourth-moment assumption is in place. Please add a qualifying remark that, for non-Gaussian noise, the sample-size bound requires a known upper bound on E[||zeta||^4].

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the data-driven SOS synthesis conditions are derived from the stated moment bounds via a concentration argument and the S-lemma; the main caveat is a distribution-free soundness gap, not circularity.

full rationale

The derivation chain is not circular. Lemma 1 obtains a concentration bound on the Frobenius distance between the empirical second moment and E[zeta zeta^T] using Markov/Chebyshev inequalities and the moment upper bounds Gamma_mu and Gamma_Sigma; Lemma 2 converts this into the data-conformity constraint (10). Theorem 2 then uses (10) to constrain the unknown Phi through (37)-(38) and applies the S-lemma so that (13c) implies the S-CBC condition (4c). None of these steps reintroduces the conclusion as an input: P, K(x), and the multipliers alpha_j are synthesized decision variables, and the confidence 1 - beta_2 is an a priori bound on the data-conformity event, not a quantity fitted to the data. The citations to [46], [48], and [49] import the standard S-CBC-to-safety theorem from independent prior publications (Kushner 1967; Anand, Lavaei, and Zamani 2022; Salamati et al. 2024) with assumptions that do not include the present data-driven claim, so these self-citations are not load-bearing in a circular sense. The genuine weakness, acknowledged in Remark 9, is that Lemma 1 is stated for Gaussian vectors and uses the Gaussian fourth-moment identity (20a); the claimed extension to arbitrary distributions does not provide a general upper bound on E[||zeta||^4] in terms of Gamma_mu and Gamma_Sigma. That is a soundness gap in the arbitrary-distribution guarantee, not a circularity: the bound is not derived from the safety conclusion it is used to certify.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

The central claim rests on the polynomial dictionary assumption, known noise moment bounds, a Gaussian-derived concentration inequality, and the S-lemma. No parameters are fitted to match the safety outcomes, but the arbitrary-distribution confidence bound is an unsupported premise. No new entities are invented; the stochastic control barrier certificate is taken from prior literature.

free parameters (1)
  • Noise moment upper bounds Gamma_mu and Gamma_Sigma = User-chosen in the experiments, e.g., Gamma_Sigma=0.006I3 for Lorenz, 0.008I3 for Chen, 0.0075I3 for spacecraft…
    Assumption 2 treats these as known inputs. Remark 2 sketches estimation by inflating empirical statistics but gives no certified procedure. The safety probability beta_1, the confidence beta_2, and the noise allowance psi all depend directly on these values.
assumptions (5)
  • domain assumption The true dynamics are exactly of the form x^+ = A F(x) + B G(x) u + zeta with known, sufficiently rich dictionaries F and G and no unmodeled dynamics.
    Assumption 1 and Remark 1. If the dictionary is incomplete, the data-conformity constraint is mis-specified and the synthesized certificate may not apply to the real system.
  • domain assumption The noise process zeta is i.i.d. across time and across the N trajectory realizations, and only the first two moments need be bounded.
    Definition 1 and Assumption 2. The union bound over j=1..T in the proof of Theorem 2 uses independence across time and trajectories to obtain confidence 1-T beta_bar_2.
  • ad hoc to paper The fourth moment of the noise is bounded in a way that the Gaussian-derived expression in Lemma 1 bounds the actual concentration failure probability.
    Lemma 1 is proven for Gaussian zeta using Gaussian moment identities. Remark 9 asserts distribution-freeness but provides no general upper bound on E[||zeta||^4]. This is the load-bearing gap in the arbitrary-distribution claim.
  • domain assumption All states are directly measurable and the training trajectories are collected while maintaining safety.
    Assumption 3 and Remark 7. The method requires full-state measurements and safe data collection; extending to input-output data is left to future work.
  • standard math The matrix S-lemma applies with nonnegative polynomial multipliers alpha_j(x) over the state set X.
    Used in the proof of Theorem 2 around equation (39). The S-lemma is standard, but the nonnegativity of alpha_j(x) over X and the Slater-type conditions are not discussed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Data-Driven Stochastic Control: Foundations and Guarantees." pith.science (2026). https://pith.science/paper/LOOAMGFR

@misc{pith2026250723280,
  author       = {Pith},
  title        = {Pith review of: Data-Driven Stochastic Control: Foundations and Guarantees},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LOOAMGFR}},
  note         = {Machine review of arXiv:2507.23280}
}
read the original abstract

This work establishes a step forward in advancing data-driven trajectory-based methods for stochastic systems with unknown mathematical dynamics. In contrast to scenario-based approaches that rely on independent and identically distributed (i.i.d.) trajectories, this work develops a data-driven framework where each trajectory is gathered over a finite horizon and exhibits temporal dependence, referred to as a non-i.i.d. trajectory. To ensure safety of dynamical systems using such trajectories, the current body of literature primarily considers dynamics subject to unknown-but-bounded disturbances, which facilitates robust analysis. While promising, such bounds may be violated in practice and the resulting worst-case robust analysis tends to be overly conservative. To overcome these key challenges, this paper considers stochastic systems with unknown mathematical dynamics, influenced by process noise with arbitrary distributions. In the proposed framework, data is collected from stochastic systems under multiple realizations within a finite-horizon experiment, where each realization generates a non-i.i.d. trajectory. Leveraging the concept of stochastic control barrier certificates constructed from data, this work quantifies probabilistic safety guarantees with a certified confidence level. To achieve this, the proposed conditions are formulated as a sum-of-squares (SOS) optimization problem, relying solely on empirical average of the collected trajectories and statistical features of the process noise. The efficacy of the approach has been validated on three stochastic benchmarks with unknown models and arbitrary noise distributions. In one case study, it is shown that while no safety controller exists for the robust analysis of the system under bounded disturbances, the proposed stochastic framework yields a safety controller together with quantified probabilistic safety guarantees.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

60 extracted references · 37 canonical work pages

  1. [1]

    Analysis and design of safety- critical, cyber-physical systems,

    J. McGregor, D. Gluch, and P. Feiler, “Analysis and design of safety- critical, cyber-physical systems,”ACM SIGAda Ada Letters, vol. 36, no. 2, pp. 31–38, 2017

    2017

  2. [2]

    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

  3. [3]

    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

  4. [4]

    Guarantees for data-driven control of nonlinear systems using semidefinite programming: A survey,

    T. Martin, T. B. Sch ¨on, and F. Allg ¨ower, “Guarantees for data-driven control of nonlinear systems using semidefinite programming: A survey,” Annual Reviews in Control, vol. 56, 2023

    2023

  5. [5]

    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

  6. [6]

    Model-based reinforcement learning: A survey,

    T. M. Moerland, J. Broekens, A. Plaat, C. M. Jonkeret al., “Model-based reinforcement learning: A survey,”Foundations and Trends in Machine Learning, vol. 16, no. 1, pp. 1–118, 2023

    work page 2023

  7. [7]

    Gaussian processes for machine learning,

    M. Seeger, “Gaussian processes for machine learning,”International journal of neural systems, vol. 14, no. 2, pp. 69–106, 2004

    work page 2004

  8. [8]

    Uniform error bounds for gaussian process regression with application to safe control,

    A. Lederer, J. Umlauft, and S. Hirche, “Uniform error bounds for gaussian process regression with application to safe control,”Advances in Neural Information Processing Systems, vol. 32, 2019

    work page 2019

Show all 60 references
  1. [9]

    Model-free reinforcement learning from expert demonstrations: a survey,

    J. Ram ´ırez, W. Yu, and A. Perrusqu ´ıa, “Model-free reinforcement learning from expert demonstrations: a survey,”Artificial Intelligence Review, vol. 55, no. 4, pp. 3213–3241, 2022

    2022

  2. [10]

    Iterative feedback tuning: theory and applications,

    H. Hjalmarsson, M. Gevers, S. Gunnarsson, and O. Lequin, “Iterative feedback tuning: theory and applications,”IEEE control systems maga- zine, vol. 18, no. 4, pp. 26–41, 1998

    1998

  3. [11]

    Virtual reference feed- back tuning: a direct method for the design of feedback controllers,

    M. Campi, A. Lecchini, and S. M. Savaresi, “Virtual reference feed- back tuning: a direct method for the design of feedback controllers,” Automatica, vol. 38, no. 8, pp. 1337–1346, 2002

    2002

  4. [12]

    The scenario approach to robust control design,

    G. C. Calafiore and M. C. Campi, “The scenario approach to robust control design,”IEEE Transactions on automatic control, vol. 51, no. 5, pp. 742–753, 2006

    2006

  5. [13]

    The scenario approach for DA T A-DRIVEN STOCHASTIC CONTROL: FOUNDA TIONS AND GUARANTEES 11 systems and control design,

    M. C. Campi, S. Garatti, and M. Prandini, “The scenario approach for DA T A-DRIVEN STOCHASTIC CONTROL: FOUNDA TIONS AND GUARANTEES 11 systems and control design,”Annual Reviews in Control, vol. 33, no. 2, pp. 149–157, 2009

    2009

  6. [14]

    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, 2019

    2019

  7. [15]

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

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

    2020

  8. [16]

    A stochastic fundamental lemma with reduced disturbance data requirements,

    R. Ou, G. Pan, and T. Faulwasser, “A stochastic fundamental lemma with reduced disturbance data requirements,”arXiv: 2502.09131, 2025

    2025 arXiv

  9. [17]

    Data-driven control of nonlinear systems: An on-line direct approach,

    M. Tanaskovic, L. Fagiano, C. Novara, and M. Morari, “Data-driven control of nonlinear systems: An on-line direct approach,”Automatica, vol. 75, pp. 1–10, 2017

    2017

  10. [18]

    Performance bounds for the scenario approach and an extension to a class of non-convex programs,

    P. Mohajerin Esfahani, T. Sutter, and J. Lygeros, “Performance bounds for the scenario approach and an extension to a class of non-convex programs,”IEEE Transactions on Automatic Control, vol. 60, no. 1, pp. 46–58, 2014

    2014

  11. [19]

    On the road between robust optimization and the scenario approach for chance constrained optimiza- tion problems,

    K. Margellos, P. Goulart, and J. Lygeros, “On the road between robust optimization and the scenario approach for chance constrained optimiza- tion problems,”IEEE Transactions on Automatic Control, vol. 59, no. 8, pp. 2258–2263, 2014

    2014

  12. [20]

    Data-driven control of large-scale networks with formal guarantees: A small-gain-free approach,

    B. Samari, A. Nejati, and A. Lavaei, “Data-driven control of large-scale networks with formal guarantees: A small-gain-free approach,”IEEE Transactions on Automatic Control, 2026

    2026

  13. [21]

    Data-driven safety certificates of infinite networks with unknown models and interconnection topologies,

    M. Zaker, A. Nejati, and A. Lavaei, “Data-driven safety certificates of infinite networks with unknown models and interconnection topologies,” Automatica, 2026

    2026

  14. [22]

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

    J. Berberich, J. K ¨ohler, M. 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, 2020

    2020

  15. [23]

    Meta results on data-driven control of nonlinear systems,

    N. Monshizadeh, C. De Persis, and P. Tesi, “Meta results on data-driven control of nonlinear systems,”arXiv:2405.10064, 2024

    2024 arXiv

  16. [24]

    Physics-informed data-driven control of nonlinear polynomial systems with noisy data,

    M. Ashoori, A. Aminzadeh, A. Nejati, and A. Lavaei, “Physics-informed data-driven control of nonlinear polynomial systems with noisy data,” arXiv:2508.01315, 2025

    2025 arXiv

  17. [25]

    A note on persistency of excitation,

    J. C. Willems, P. Rapisarda, I. Markovsky, and B. L. De Moor, “A note on persistency of excitation,”Systems & Control Letters, vol. 54, no. 4, pp. 325–329, 2005

    2005

  18. [26]

    Gaussian inference for data- driven state-feedback design of nonlinear systems,

    T. Martin, T. B. Sch ¨on, and F. Allg ¨ower, “Gaussian inference for data- driven state-feedback design of nonlinear systems,”IFAC-PapersOnLine, vol. 56, no. 2, pp. 4796–4803, 2023

    2023

  19. [27]

    Safety verification of hybrid systems using barrier certificates,

    S. Prajna and A. Jadbabaie, “Safety verification of hybrid systems using barrier certificates,” inInternational workshop on hybrid systems: Computation and control, 2004, pp. 477–492

    2004

  20. [28]

    A framework for worst- case and stochastic safety verification using barrier certificates,

    S. Prajna, A. Jadbabaie, and G. J. Pappas, “A framework for worst- case and stochastic safety verification using barrier certificates,”IEEE Transactions on Automatic Control, vol. 52, no. 8, pp. 1415–1428, 2007

    2007

  21. [29]

    Constructive safety using control barrier functions,

    P. Wieland and F. Allg ¨ower, “Constructive safety using control barrier functions,”IFAC Proceedings Volumes, vol. 40, no. 12, pp. 462–467, 2007

    2007

  22. [30]

    Control barrier functions: Theory and applications,

    A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada, “Control barrier functions: Theory and applications,” inEuropean control conference (ECC), 2019, pp. 3420–3431

    2019

  23. [31]

    W. Xiao, C. G. Cassandras, and C. Belta,Safe Autonomy with Control Barrier Functions: Theory and Applications. Springer, 2023

    2023

  24. [32]

    Control barrier function based quadratic programs for safety critical systems,

    A. D. Ames, X. Xu, J. W. Grizzle, and P. Tabuada, “Control barrier function based quadratic programs for safety critical systems,”IEEE Transactions on Automatic Control, vol. 62, no. 8, pp. 3861–3876, 2016

    2016

  25. [33]

    Control barrier functions for complete and incomplete in- formation stochastic systems,

    A. Clark, “Control barrier functions for complete and incomplete in- formation stochastic systems,” inAmerican Control Conference (ACC), 2019, pp. 2928–2935

    2019

  26. [34]

    A barrier function approach to finite-time stochastic system verification and control,

    C. Santoyo, M. Dutreix, and S. Coogan, “A barrier function approach to finite-time stochastic system verification and control,”Automatica, vol. 125, 2021

    2021

  27. [35]

    Multi-Robot Collision Avoidance under Uncertainty with Probabilistic Safety Barrier Certificates,

    W. Luo, W. Sun, and A. Kapoor, “Multi-Robot Collision Avoidance under Uncertainty with Probabilistic Safety Barrier Certificates,” in Advances in Neural Information Processing Systems, vol. 33, 2020, pp. 372–383

    2020

  28. [36]

    Scalable synthesis of safety barrier certifi- cates for networks of stochastic switched systems,

    A. Lavaei and E. Frazzoli, “Scalable synthesis of safety barrier certifi- cates for networks of stochastic switched systems,”IEEE Transactions on Automatic Control, vol. 11, no. 69, pp. 7294–7309, 2024

    2024

  29. [37]

    Context-triggered games for reactive synthesis over stochastic systems via control barrier certificates,

    A. Nejati, S. Prakash Nayak, and A.-K. Schmuck, “Context-triggered games for reactive synthesis over stochastic systems via control barrier certificates,” inProceedings of the 27th ACM International Conference on Hybrid Systems: Computation and Control, 2024, pp. 1–12

    2024

  30. [38]

    Compositional design of safety controllers for large-scale stochastic hybrid systems,

    M. Zaker, O. Akbarzadeh, B. Samari, and A. Lavaei, “Compositional design of safety controllers for large-scale stochastic hybrid systems,” Automatica, vol. 190, 2026

    2026

  31. [39]

    Data-driven con- troller synthesis of unknown nonlinear polynomial systems via control barrier certificates,

    A. Nejati, B. Zhong, M. Caccamo, and M. Zamani, “Data-driven con- troller synthesis of unknown nonlinear polynomial systems via control barrier certificates,” inLearning for Dynamics and Control Conference, 2022, pp. 763–776

    2022

  32. [40]

    From a single trajectory to safety controller synthesis of discrete-time nonlinear poly- nomial systems,

    B. Samari, O. Akbarzadeh, M. Zaker, and A. Lavaei, “From a single trajectory to safety controller synthesis of discrete-time nonlinear poly- nomial systems,”IEEE Control Systems Letters, vol. 8, pp. 3123–3128, 2024

    2024

  33. [41]

    From formal meth- ods to data-driven safety certificates of unknown large-scale networks,

    O. Akbarzadeh, B. Samari, A. Nejati, and A. Lavaei, “From formal meth- ods to data-driven safety certificates of unknown large-scale networks,” IEEE Transactions on Automatic Control, 2025

    2025

  34. [42]

    Controller design for robust invariance from noisy data,

    A. Bisoffi, C. De Persis, and P. Tesi, “Controller design for robust invariance from noisy data,”IEEE Transactions on Automatic Control, vol. 68, no. 1, pp. 636–643, 2022

    2022

  35. [43]

    Data-based guarantees of set invariance properties,

    A. Bisoffi, C. D. Persis, and P. Tesi, “Data-based guarantees of set invariance properties,”IFAC-PapersOnLine, vol. 53, no. 2, pp. 3953– 3958, 2020

    2020

  36. [44]

    Data-driven design of safe control for polynomial systems,

    A. Luppi, A. Bisoffi, C. De Persis, and P. Tesi, “Data-driven design of safe control for polynomial systems,”European Journal of Control, vol. 75, 2024

    2024

  37. [45]

    Learning robust safety controllers for uncertain input-affine polynomial systems,

    O. Akbarzadeh, M. Ashoori, and A. Lavaei, “Learning robust safety controllers for uncertain input-affine polynomial systems,”64th IEEE Conference on Decision and Control (CDC 2025) - arXiv:2412.03919v2, 2025

    2025 arXiv

  38. [46]

    Automated verifica- tion and synthesis of stochastic hybrid systems: A survey,

    A. Lavaei, S. Soudjani, A. Abate, and M. Zamani, “Automated verifica- tion and synthesis of stochastic hybrid systems: A survey,”Automatica, vol. 146, 2022

    2022

  39. [47]

    H. J. Kushner,Stochastic Stability and Control, ser. Mathematics in Science and Engineering. Elsevier Science, 1967

    1967

  40. [48]

    From small-gain theory to com- positional construction of barrier certificates for large-scale stochastic systems,

    M. Anand, A. Lavaei, and M. Zamani, “From small-gain theory to com- positional construction of barrier certificates for large-scale stochastic systems,”IEEE Transactions on Automatic Control, vol. 67, no. 10, pp. 5638–5645, 2022

    2022

  41. [49]

    Data-driven verification and synthesis of stochastic systems via barrier certificates,

    A. Salamati, A. Lavaei, S. Soudjani, and M. Zamani, “Data-driven verification and synthesis of stochastic systems via barrier certificates,” Automatica, vol. 159, 2024

    2024

  42. [50]

    Data-driven control of nonlinear systems from input-output data,

    X. Dai, C. De Persis, N. Monshizadeh, and P. Tesi, “Data-driven control of nonlinear systems from input-output data,” inProceedings of the 62nd IEEE Conference on Decision and Control (CDC), 2023, pp. 1613–1618

    2023

  43. [51]

    An introduction to matrix concentration inequalities,

    J. A. Troppet al., “An introduction to matrix concentration inequalities,” Foundations and Trends in Machine Learning, vol. 8, no. 1-2, pp. 1–230, 2015

    2015

  44. [52]

    SOS- TOOLS: Control applications and new developments,

    S. Prajna, A. Papachristodoulou, P. Seiler, and P. A. Parrilo, “SOS- TOOLS: Control applications and new developments,” inProceedings of IEEE International Conference on Robotics and Automation, 2004, pp. 315–320

    2004

  45. [53]

    S. H. Strogatz,Nonlinear dynamics and chaos: With applications to physics, biology, chemistry, and engineering, 2018

    2018

  46. [54]

    H. K. Khalil,Nonlinear systems. Prentice hall Upper Saddle River, NJ, 2002

    2002

  47. [55]

    Chebyshev’s approximation algorithms and applica- tions,

    M. Hern ´andez, “Chebyshev’s approximation algorithms and applica- tions,”Computers & Mathematics with Applications, vol. 41, no. 3-4, pp. 433–445, 2001

    2001

  48. [56]

    Billingsley,Probability and Measure, 3rd ed

    P. Billingsley,Probability and Measure, 3rd ed. John Wiley & Sons, 1995

    1995

  49. [57]

    The matrix cookbook,

    K. B. Petersen, M. S. Pedersenet al., “The matrix cookbook,”Technical University of Denmark, vol. 7, no. 15, 2008

    2008

  50. [58]

    LMI properties and applications in systems, stability, and control theory,

    R. J. Caverly and J. R. Forbes, “LMI properties and applications in systems, stability, and control theory,”arXiv:1903.08599, 2019

    1903 arXiv

  51. [59]

    Boole,The mathematical analysis of logic

    G. Boole,The mathematical analysis of logic. Philosophical Library, 1847

  52. [60]

    From noisy data to feedback controllers: Nonconservative design via a matrix s-lemma,

    H. J. van Waarde, M. K. Camlibel, and M. Mesbahi, “From noisy data to feedback controllers: Nonconservative design via a matrix s-lemma,” IEEE Transactions on Automatic Control, vol. 67, no. 1, pp. 162–175, 2022. VIII. APPENDIX: PROOF OFSTATEMENTS Proof of Lemma 1.Consider the...

    2022

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.