Pith. sign in

REVIEW 4 major objections 5 minor 33 references

A Dual Ensemble Kalman Filter Approach to Robust Control of Nonlinear Systems: An Application to Partial Differential Equations

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read Adding a Lyapunov redesign term to a dual-EnKF learned control stabilizes nonlinear systems and PDE discretizations under bounded disturbances.

desk verdict Honest about its assumptions but the theory doesn't cover the PDE experiments; the simulations carry the paper. read the letter →

arxiv 2508.21684 v1 pith:ZVJ4B5P5 submitted 2025-08-29 math.OC cs.SYeess.SY

Anant A. Joshi , Saviz Mowlavi , Mouhacine Benosman This is my paper · ORCID
classification math.OCcs.SYeess.SY MSC 93C2093C1093E11
keywords dualensembleKalmanfilterLyapunovredesignrobustcontrolnonlinearaffinesystemspartialdifferentialequationsheatequationBurgersdata-driven
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 seeks to show that a fully data-driven controller — no model identification, just a simulator — can robustly stabilize nonlinear control-affine systems, including discretized PDEs. The robust control is the sum of an optimal stabilizing control learned by the dual ensemble Kalman filter and a Lyapunov redesign term that cancels bounded disturbances. The authors prove asymptotic stability for both linear and nonlinear systems under a full-rank condition on the input map, then demonstrate on heat and Burgers equations that the robustified control drives the L2 norm of the state orders of magnitude closer to zero than the optimal control alone. The reader would care because the approach targets high-dimensional PDE control, where standard model-based design is impractical and data-driven methods must work with simulators only.

What carries the argument

Two objects carry the argument. The first is the dual ensemble Kalman filter (dual EnKF), a controlled interacting-particle system running backward in time whose terminal empirical covariance (linear case) or empirical gradient (nonlinear case) approximates the value function of the optimal control problem without ever constructing the system matrices. The second is the Lyapunov redesign term u_d = −λ b†∇V / |∇V|, which uses the same learned value function to produce a control component that points along −∇V; under the rank condition on b (or B), b u_d equals −λ ∇V/|∇V|, exactly canceling the worst-case disturbance in the V̇ inequality.

What would settle it

For a minimal full-rank case (e.g., n=2, B=I, A with a nonzero unstable mode), compute the closed-loop V̇ exactly for a bounded disturbance with the controller of Proposition 1; if V̇ is not strictly negative for some state, the theorem is false. For the implemented method, run the heat-equation simulation with N=10000 EnKF particles and measure the steady-state L2 norm against the practical-stability radius r=0.002; norms far exceeding that radius would indicate that EnKF approximation error, not disturbance, limits performance.

Watch

Extended reading notes

Core claim

The paper's central claim is Proposition 2 (and its linear counterpart, Proposition 1): for the nonlinear affine system ẋ = a(x)+b(x)u+d, the feedback u = ū + u_d, with ū from the HJB-optimal problem and u_d = −λ b†∇V/|∇V|, renders the origin asymptotically stable for every disturbance bounded by |d| < λ. The proof shows V̇ ≤ (|d|−λ)|∇V| < 0, because the redesign term exactly cancels the disturbance once b has full rank. The dual EnKF supplies ū and V from simulator trajectories alone: backward-run interacting particles approximate the ARE solution or value-function gradient, and the control is recovered by minimizing the Hamiltonian H(x,u) = (∇V)ᵀS(x,u)+½L(x,u) using only simulator calls. E

Load-bearing premise

The asymptotic-stability proofs require the input matrix b(x) to have rank equal to the state dimension, so the Lyapunov redesign term can exactly cancel the disturbance; the PDE experiments relax this condition, so the formal guarantee does not apply there and the reported behavior rests on the simulations.

Editorial extensions

If this is right

  • Under the full-rank input-map condition, the proposed controller guarantees asymptotic stability for any disturbance bounded by a known λ — not merely practical stability.
  • The controller can be implemented with only a disturbance-free simulator: both the optimal term and the robustifying term are computed through Hamiltonian evaluations S(x,u), with no need to identify a(x) and b(x).
  • Simulations on heat and Burgers equations show the robustified control reduces the terminal L2 norm by an order of magnitude compared with the optimal control alone under the same disturbances.
  • For the Burgers equation, the robustification term also masks model mismatch: the controller built on a reduced-order linear model attains settling performance close to that of the full nonlinear model.
  • Adding a small regularizing parameter r to u_d converts the asymptotic guarantee into practical stability in a ball; the experiments confirm this behavior at r = 0.002.

Reading between the lines

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

  • Because the PDE implementations relax the full-rank assumption (n=100 with m=8 controls, n=128 with m=10), the demonstrated stability is practical, not asymptotic; a practitioner should tune λ and r explicitly rather than relying on Propositions 1–2 in underactuated regimes.
  • The Hamiltonian-minimization step is agnostic to the learning algorithm — any zero-order optimizer or policy-gradient method could supply the optimal-control component, so the Lyapunov redesign construction is a general plugin for robustifying learned controllers.
  • The framework extends naturally to stochastic disturbances (e.g., Gaussian white noise) and to output-feedback settings; the paper's filtering-based EnKF machinery gives a direct path, since value-function estimation could be carried out from sensor measurements.
  • For other dissipative PDEs such as Allen-Cahn or KdV, the same Lyapunov redesign mechanism should transfer, because the robustness term needs only a Lyapunov function and a known disturbance bound, not PDE-specific structure.
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

4 major / 5 minor

Summary. The paper proposes a data-driven robust control design for nonlinear affine-in-control systems, combining an optimal feedback law computed via the dual ensemble Kalman filter (dual EnKF) with an additive Lyapunov-redesign term. Section II treats LTI systems, Section III extends the construction to nonlinear systems, and Section IV applies the method to discretized heat and Burgers PDEs, testing constant and sinusoidal disturbances. The main theoretical claims are Propositions 1 and 2, which assert asymptotic stability under the respective rank conditions rank(B)=n and rank(b(x))=n. The numerical experiments deliberately operate outside these conditions, as acknowledged in Remarks 1 and 4.

Significance. If the gap between the assumptions and the PDE experiments can be closed, the paper is a useful demonstration that a learning-based optimal controller can be robustified by a Lyapunov-redesign term computed with only simulator access. The dual EnKF component is prior work with its own convergence analysis, and the robustification theory is standard; the manuscript's contribution is the combination and its PDE application. The simulation study is reasonably broad: two PDEs, two viscosities, two disturbance types, and 100 random initial conditions, with robust control visibly outperforming optimal control alone. However, the main theoretical guarantee does not cover the demonstrated rank-deficient regime, and the claims of practical stability in Remarks 2 are asserted without proof. These are load-bearing issues for the paper's central claim, though they appear fixable by adding a matched-disturbance analysis or by explicitly repositioning the PDE results as empirical demonstrations.

major comments (4)
  1. [Section II.B, Proposition 1 proof] The displayed Lyapunov derivative has a sign error. For the closed-loop matrix A_cl = A - B\bar K, the true derivative is \dot V = x^T \bar P A_cl x + (Bu_d+d)^T\bar P x = -\tfrac12 x^T(Q + \bar P B R^{-1}B^T\bar P)x + (Bu_d+d)^T\bar P x. The paper instead writes \dot V = -x^T\bar P(A-B\bar K)x + ... and then uses the claim that this term is nonpositive. The conclusion of Proposition 1 is restorable with the standard ARE computation, but the proof as written is not valid and should be corrected.
  2. [Section II.B/III.B vs. Section IV, App. II] Assumptions 2(i) and 4(ii) require rank(B)=n and rank(b(x))=n, but the PDE experiments violate these assumptions: heat equation has n=100, m=8/10, and Burgers full model has p=128, m=10. Remarks 1 and 4 say the conditions are relaxed, but no replacement stability bound is provided. When rank(B)<n, BB^\dagger is the orthogonal projection onto Range(B), so u_d only cancels the component of \bar P x in Range(B). For disturbances with a component in Null(B^T), the term d^T\bar P x can make \dot V positive. This is not merely a cosmetic gap: the asymptotic stability guarantee of Propositions 1 and 2 does not cover the systems on which the paper demonstrates the method.
  3. [Remark 2, Section II.B] The claim that the regularized implementation gives 'practical stability' is asserted without proof. In the rank-deficient regime relevant to the PDE experiments, the control term is -\lambda (BB^\dagger \bar P x)/|\bar P x|; if \bar P x is close to Null(B^T), this term is small while a disturbance in Null(B^T) contributes up to \lambda |\bar P x|. The derivative can then be positive for small |x|, so neither asymptotic nor practical stability follows from the given argument. A concrete matched/unmatched disturbance analysis or a modified controller is needed before the PDE claims can be supported theoretically.
  4. [Appendix I-C, Algorithm 2] In the unknown-B branch, the algorithm returns (\bar u^{(N)})_i = (R^{-1}B^T\bar P^{(N)}x)_i, which is the negative of the optimal control. Indeed, for H as defined in (4), minimizing in u gives \bar u = -R^{-1}B^T\bar P x. The B-known branch returns this negative value, so the two branches disagree. If a user implements the unknown-B branch, the learning control will have the wrong sign. This is a load-bearing implementation error for the model-free claims, although the reported PDE experiments may use the known-B branch.
minor comments (5)
  1. [Section IV.A vs. Appendix II-A] The heat equation section states m=10, while Appendix II-A states m=8. Please make these consistent.
  2. [Section IV, Eq. (8) discussion] The disturbance notation is ambiguous: U(t)=u(t)+d(t) suggests d(t) is m-dimensional, but later d(t)=d0 sin(t) and d(t)=d0 are written as scalars. Clarify whether d is a scalar or a vector in R^m.
  3. [Section II] Typo: 'stailizing' should be 'stabilizing' in the sentence introducing the stabilizing control.
  4. [Algorithm 1 and Section III.C] In the nonlinear case, the line using S(0,v) to minimize |b(x)v - \nabla V/|\nabla V|| is not generally correct, because S(0,v) = a(0)+b(0)v, not b(x)v. Please clarify how the simulator is used to evaluate b(x)v, e.g., via S(x,v)-S(x,0), or state the needed assumptions.
  5. [Section III.B, proof of Proposition 2] The proof says 'For the first equality we recall Assumption 4 and Cauchy-Schwarz'; this should be 'first inequality'. The equality/inequality labels are currently inconsistent.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: stability proofs are standard Lyapunov redesign and the dual EnKF is legitimate tool reuse.

full rationale

The derivation is self-contained. Propositions 1 and 2 are direct applications of Khalil's Lyapunov redesign under the stated rank and bounded-disturbance assumptions; the stabilizing properties of the ARE/HJB controllers are standard results (Khalil; Kwakernaak & Sivan), not conclusions that depend on the paper's own numerical method. The dual EnKF is invoked only as a numerical approximation of Pbar or ∇V, and the paper's stability statements are made conditional on having these quantities; reliance on the authors' prior [15] is legitimate tool reuse with its own published convergence analysis, not a load-bearing self-citation that defines the result. The data-driven evaluation uses the same simulator for learning and testing, but disturbances are absent during learning and present during evaluation, so the comparison is not statistically forced. The explicit relaxation of rank(B)=rank(b)=n in the PDE experiments (Remarks 1 and 4) makes Props. 1/2 inapplicable to the simulations, and Remark 2's practical-stability claim is asserted without proof; these are correctness limitations, not circular reductions. No step equates a prediction with an input by construction.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The central claim rests on four numbered assumptions (1-4), the classical ARE/DRE theory, and the imported nonlinear dual EnKF gradient formula. The free parameters are user choices (cost weights, lambda, regularization, particle count, reduced dimension) rather than fitted constants, but they influence the quantitative results. No new physical entities are postulated.

free parameters (5)
  • Robust gain lambda = swept values; 0.2 in trajectory plots
    Gain multiplying the Lyapunov redesign term; chosen by the user to approximate the disturbance bound. The paper sweeps lambda and reports the robust control works for all values, but the displayed trajectory plots use lambda = 0.2.
  • Regularization parameter r = 0.002
    Added to |P̄x| in Algorithm 1 to avoid division by zero; turns asymptotic stability into practical stability (Remark 2).
  • Number of dual EnKF particles N = 10000 (heat), 1000 (Burgers)
    Hyperparameter of the dual EnKF; affects accuracy of P̄ and ∇φ estimates; no sensitivity study is provided.
  • DMDc reduced state dimension n = 10
    Choice of reduced order for the Burgers linear simulator; affects model mismatch and the comparison between reduced and full nonlinear control.
  • Cost weights Q, R, G = Q=I, G=I, R=I (heat), R=0.1I (Burgers)
    User-selected weights in the optimal control cost; influence the learned controller.
assumptions (6)
  • standard math Standard linear optimal control theory: the DRE has a positive definite solution converging to the stabilizing ARE solution under controllability and observability (Section II.A).
    Used to define the optimal gain K̄ and Lyapunov function V = 0.5 xᵀP̄x.
  • domain assumption Assumption 1: (A, B) controllable and (A, C) observable, R, G positive definite (Section II.A).
    Ensures the ARE solution exists and gives a stabilizing controller for the nominal linear system.
  • domain assumption Assumption 2: rank(B) = n and a known bound λ with |d(t,x)| < λ(t,x) (Section II.B).
    Needed for the proof of Proposition 1 so that B B† P̄x = P̄x and the redesign term cancels the disturbance.
  • domain assumption Assumption 3: access to a disturbance-free simulator that can be run forward and backward, and knowledge of Q, R, G (Section II.C).
    Underlies the data-driven implementation; without it the dual EnKF cannot be run.
  • domain assumption Assumption 4: existence of a stabilizing ū and Lyapunov function V, rank(b(x)) = n for all x, and a known bound λ (Section III.B).
    Needed for the proof of Proposition 2; the rank condition is relaxed in the PDE implementations.
  • ad hoc to paper The nonlinear dual EnKF produces the value function gradient as ∇φ^(N)(x) = (S_0^(N))^{-1} x (Appendix I-B).
    This linear-in-state gradient formula is exact only for linear dynamics; for the nonlinear extension in Section III it is imported from [15] without restated assumptions, yet it is the basis of the Hamiltonian in (6) and hence of the learned optimal control.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Dual Ensemble Kalman Filter Approach to Robust Control of Nonlinear Systems: An Application to Partial Differential Equations." pith.science (2026). https://pith.science/paper/ZVJ4B5P5

@misc{pith2026250821684,
  author       = {Pith},
  title        = {Pith review of: A Dual Ensemble Kalman Filter Approach to Robust Control of Nonlinear Systems: An Application to Partial Differential Equations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZVJ4B5P5}},
  note         = {Machine review of arXiv:2508.21684}
}
read the original abstract

This paper considers the problem of data-driven robust control design for nonlinear systems, for instance, obtained when discretizing nonlinear partial differential equations (PDEs). A robust learning control approach is developed for nonlinear affine in control systems based on Lyapunov redesign technique. The robust control is developed as a sum of an optimal learning control which stabilizes the system in absence of disturbances, and an additive Lyapunov-based robustification term which handles the effects of disturbances. The dual ensemble Kalman filter (dual EnKF) algorithm is utilized in the optimal control design methodology. A simulation study is done on the heat equation and Burgers partial differential equation.

Figures

Figures reproduced from arXiv: 2508.21684 by the authors.

Figure 1
Figure 1. Results for control of heat equation. (a) Mean of ∥()∥ [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Results for control of Burgers equation using reduced order DMDc model. [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Results for control of Burgers equation using full nonlinear model. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 33 canonical work pages

  1. [15]

    A. A. J OSHI , A. T AGHVAEI , P. G. M EHTA, AND S. P. M EYN, Con- trolled interacting particle algorithms for simulation-based reinforce- ment learning, Systems & Control Letters, 170 (2022), p. 105392

    work page 2022

  2. [1]

    B ACH AND V

    F. B ACH AND V. PERCHET , Highly-smooth zero-th order online op- timization, in 29th Annual Conference on Learning Theory, V . Feld- man, A. Rakhlin, and O. Shamir, eds., vol. 49 of Proceedings of Machine Learning Research, Columbia University, New York, New York, USA, 23–26 Jun 2016, PMLR, pp. 257–283

    work page 2016

  3. [2]

    B ARBAGALLO , D

    A. B ARBAGALLO , D. S IPP, AND P. J. S CHMID , Closed-loop control of an open cavity flow using reduced-order models , Journal of Fluid Mechanics, 641 (2009), pp. 1–50

    work page 2009

  4. [3]

    B HATTACHARYA , B

    K. B HATTACHARYA , B. H OSSEINI , N. B. K OVACHKI , AND A. M. STUART , Model reduction and neural networks for parametric PDEs, The SMAI Journal of Computational Mathematics, 7 (2021), pp. 121– 157

    work page 2021

  5. [4]

    A. N. B ISHOP AND P. D EL MORAL , On the mathematical theory of ensemble (linear-gaussian) kalman–bucy filtering , Mathematics of Control, Signals, and Systems, 35 (2023), pp. 835–903

    work page 2023

  6. [5]

    A. B. B LANCHARD , G. Y. C ORNEJO MACEDA , D. F AN, Y. L I, Y. ZHOU , B. R. N OACK , AND T. P. S APSIS , Bayesian optimization for active flow control , Acta Mechanica Sinica, (2021), pp. 1–13

    work page 2021

  7. [6]

    D URIEZ , S

    T. D URIEZ , S. L. B RUNTON , AND B. R. N OACK, Machine Learn- ing Control-Taming Nonlinear Dynamics and Turbulence , vol. 116, Springer, 2017

    work page 2017

  8. [7]

    E VENSEN , Data Assimilation

    G. E VENSEN , Data Assimilation. The Ensemble Kalman Filter , Springer-Verlag, New York, 2006

    work page 2006

Show all 33 references
  1. [8]

    D. F AN, L. Y ANG , Z. W ANG , M. S. T RIANTAFYLLOU , AND G. E. KARNIADAKIS , Reinforcement learning for bluff body active flow control in experiments and simulations , Proceedings of the National Academy of Sciences, 117 (2020), pp. 26091–26098

    2020

  2. [9]

    F AZEL , R

    M. F AZEL , R. G E, S. K AKADE , AND M. M ESBAHI , Global conver- gence of policy gradient methods for the linear quadratic regulator , in Proceedings of the 35th International Conference on Machine Learning, J. Dy and A. Krause, eds., vol. 80 of Proceedings of Machine Learnin...

    2018

  3. [10]

    W. H. F LEMING AND S. K. M ITTER , Optimal Control and Nonlin- ear Filtering for Nondegenerate Diffusion Processes , Stochastics, 8 (1982), pp. 63–77

    1982

  4. [11]

    F RESCA , L

    S. F RESCA , L. D EDE ’, AND A. M ANZONI , A comprehensive deep learning-based approach to reduced order modeling of nonlinear time-dependent parametrized PDEs, Journal of Scientific Computing, 87 (2021), pp. 1–36

    2021

  5. [12]

    G ARNIER , J

    P. G ARNIER , J. V IQUERAT , J. R ABAULT, A. L ARCHER , A. K UHNLE , AND E. H ACHEM , A review on deep reinforcement learning for fluid mechanics , Computers & Fluids, 225 (2021), p. 104973

    2021

  6. [13]

    H OFFMANN AND P

    C. H OFFMANN AND P. ROSTALSKI , Linear optimal control on factor graphs — a message passing perspective — , IFAC-PapersOnLine, 50 (2017), pp. 6314–6319. 20th IFAC World Congress

    2017

  7. [14]

    H OVLAND , J

    S. H OVLAND , J. T. G RAVDAHL , AND K. E. W ILLCOX , Explicit model predictive control for large-scale systems via model reduction , Journal of guidance, control, and dynamics, 31 (2008), pp. 918–926

    2008

  8. [16]

    K AISER , J

    E. K AISER , J. N. K UTZ , AND S. L. B RUNTON , Data-driven dis- covery of Koopman eigenfunctions for control , Machine Learning: Science and Technology, 2 (2021), p. 035023

    2021

  9. [17]

    H. J. K APPEN , Linear theory for control of nonlinear stochastic systems, Phys. Rev. Lett., 95 (2005), p. 200201

    2005

  10. [18]

    H. K. K HALIL , Nonlinear systems, Macmillan Pub. Co., New York, 1992

    1992

  11. [19]

    K WAKERNAAK AND R

    H. K WAKERNAAK AND R. S IVAN, Linear optimal control systems , Wiley Interscience, New York, 1972

    1972

  12. [20]

    L EIBFRITZ AND S

    F. L EIBFRITZ AND S. V OLKWEIN , Numerical feedback controller design for PDE systems using model reduction: Techniques and case studies, in Real-Time PDE-Constrained Optimization, SIAM, 2007, pp. 53–72

    2007

  13. [21]

    L EVINE , Reinforcement learning and control as probabilistic inference: Tutorial and review , 2018

    S. L EVINE , Reinforcement learning and control as probabilistic inference: Tutorial and review , 2018

    2018

  14. [22]

    L IBERZON , Calculus of variations and optimal control theory , Princeton University Press, Princeton, NJ, 2012

    D. L IBERZON , Calculus of variations and optimal control theory , Princeton University Press, Princeton, NJ, 2012

    2012

  15. [23]

    M OHAMMADI , A

    H. M OHAMMADI , A. Z ARE , M. S OLTANOLKOTABI , AND M. R. JOVANOVI ´C, Convergence and sample complexity of gradient meth- ods for the model-free linear–quadratic regulator problem , IEEE Transactions on Automatic Control, 67 (2022), pp. 2435–2450

    2022

  16. [24]

    J. L. P ROCTOR , S. L. B RUNTON , AND J. N. K UTZ, Dynamic mode decomposition with control , SIAM Journal on Applied Dynamical Systems, 15 (2016), pp. 142–161

    2016

  17. [25]

    S IPP AND P

    D. S IPP AND P. J. S CHMID , Linear closed-loop control of fluid instabilities and noise-induced perturbations: A review of approaches and tools, Applied Mechanics Reviews, 68 (2016), p. 020801

    2016

  18. [26]

    T AGHVAEI , J

    A. T AGHVAEI , J. DE WILJES , P. G. MEHTA, AND S. R EICH , Kalman filter and its modern extensions for the continuous-time nonlinear filtering problem , Journal of Dynamic Systems, Measurement, and Control, 140 (2017), p. 030904

    2017

  19. [27]

    T ODOROV , Linearly-solvable markov decision problems , in Ad- vances in Neural Information Processing Systems, B

    E. T ODOROV , Linearly-solvable markov decision problems , in Ad- vances in Neural Information Processing Systems, B. Sch ¨olkopf, J. Platt, and T. Hoffman, eds., vol. 19, MIT Press, 2007

    2007

  20. [28]

    T SOLOVIKOS , E

    A. T SOLOVIKOS , E. B AKOLAS , S. S URYANARAYANAN , AND D. G OLDSTEIN , Estimation and control of fluid flows using sparsity- promoting dynamic mode decomposition , IEEE Control Systems Letters, 5 (2020), pp. 1145–1150

    2020

  21. [29]

    V IJAYAKUMAR , K

    S. V IJAYAKUMAR , K. R AWLIK , AND M. T OUSSAINT , On stochastic optimal control and reinforcement learning by approximate inference, in Robotics: Science and Systems VIII, N. Roy, P. Newman, and S. Srinivasa, eds., MIT Press, 2013, pp. 353–360

    2013

  22. [30]

    Y ANG , R

    T. Y ANG , R. S. L AUGESEN , P. G. M EHTA, AND S. P. M EYN, Multivariable feedback particle filter, Automatica, 71 (2016), pp. 10– 23

    2016

  23. [31]

    Z HANG , W

    X. Z HANG , W. M AO, S. M OWLAVI , M. B ENOSMAN , AND T. BAS¸ AR, Controlgym: Large-scale safety-critical control environ- ments for benchmarking reinforcement learning algorithms , arXiv preprint arXiv:2311.18736, (2024)

    2024 arXiv

  24. [32]

    Z HANG , S

    X. Z HANG , S. M OWLAVI , M. B ENOSMAN , AND T. BAS¸ AR, Policy optimization for pde control with a warm start , arXiv preprint arXiv 2403.01005, (2024). APPENDIX I DUAL ENKF ALGORITHM In this algorithm, we simulate over the time horizon [0, T] an ensemble of N particles {Y i ...

    2024 arXiv

  25. [33]

    is used for PDE simulation. B. Burgers equation Simulation parameters are as follows. The simulation time T = 3 with simulation time step = 0.001. The number of states of the discretized PDE (8) is p = 128. The number of control basis functions is m = 10 , and they are χj is t...

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.