Pith. sign in

REVIEW 5 major objections 5 minor 123 references

Three time scales make real-time control of river flow tractable.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-03 16:08 UTC pith:5IOACNAW

load-bearing objection A candid, well-organized roadmap for real-time mixed-integer control of shallow-water flows, but the central sufficiency claim and the lone existence theorem are both under-supported; useful as a research agenda, not a validated method. the 5 major comments →

arxiv 2512.14387 v4 pith:5IOACNAW submitted 2025-12-16 math.OC

Towards Real Time Control of Water Engineering with Nonlinear Hyperbolic Partial Differential Equations

Fabio DiFonzo , Michael Holst , Morteza Kimiaei , Vyacheslav Kungurtsev , Songqiang Qiu This is my paper
classification math.OC MSC 49J2035L6565K1090C1193C20
keywords real-time controlSaint-Venant equationsshallow water equationsmixed-integer PDE-constrained optimizationapproximate dynamic programmingmodel predictive controlfeasibility machine learninghydropower cascade
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper argues that real-time optimal control of a hydropower river system—continuous turbine settings, binary dam on/off switches, nonlinear shallow-water dynamics, and forecast uncertainty—cannot be solved by any single optimization method, but can be made tractable by splitting computation into three time scales. The offline scale uses large-scale computing and machine learning to build a catalog of feasible, high-quality control trajectories and to learn where the PDE solver is reliable. The meso scale evolves a small population of candidate integer schedules using approximate dynamic programming and second-order trust-region optimization. The real-time scale runs fast Newton steps on reduced, piecewise-linear models, warm-started by the meso layer. The authors frame the three-tier architecture as simultaneously necessary and probably sufficient for satisfactory hydrological control, and they give an existence result for the Saint-Venant system under boundedness assumptions.

Core claim

The paper's central claim is that a three-time-scale decomposition—offline catalog generation, meso-scale approximate dynamic programming over particles, and real-time Newton-based reduced-order control—constitutes a simultaneously necessary and probably sufficient set of operations for real-time control of river flow in hydrological engineering. The claim is presented as a framework, not a finished algorithm: the exact stochastic dynamic programming problem is acknowledged to be intractable, so the paper organizes a chain of approximations that transfers information between scales. It also asserts existence of a generalized solution to the Saint-Venant system when states are bounded away fr

What carries the argument

The load-bearing object is the three-tier architecture plus the learned feasibility map ψ_ω, a classifier that predicts whether a control–uncertainty triple yields a stable PDE solve. The offline layer produces a catalog of validated solutions using branch-and-bound, heuristics, derivative-free optimization, and reinforcement learning; the meso layer evolves particles with fixed binary schedules and relaxed final-step decisions, solved by a trust-region filter sequential quadratically constrained quadratic program; the real-time layer uses real-time iteration and piecewise-linear semi-implicit discretizations of the shallow-water equations, with active-set/wetting-drying structure encoded by

Load-bearing premise

The whole framework stands on the assumption that a machine-learned classifier can reliably identify the region of control–uncertainty inputs where the Saint-Venant solver converges; if that feasibility map is wrong or fails under new conditions, the offline catalog, the meso particles, and the real-time warm starts all lose their validity.

What would settle it

Run the offline phase on a modest 1D Saint-Venant cascade, train the feasibility classifier, then deploy it on a shifted test set with different inflow hydrographs or bathymetry. If any control chosen from the trusted region causes the high-fidelity solver to fail, or if the classifier's false-positive rate rises sharply under the shift, the central claim is falsified. A second check is to test the double-loop fixed-point iterations in Theorem 3.2 on bounded initial data and see whether the iterates actually converge to a solution.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • If the framework is right, mixed-integer decisions and uncertainty no longer have to be excluded from real-time PDE-constrained control; they are handled at different clock speeds.
  • The machine-learned feasibility map becomes a safety-critical component: every downstream optimization is restricted to regions where the PDE solver is known to converge.
  • Piecewise-linear semi-implicit shallow-water models make the real-time subproblems convex or mixed-integer linear, enabling Newton-like speed on embedded hardware.
  • The existence theorem, if it withstands scrutiny, gives a functional-analytic basis for treating the control-to-state map as well-defined on bounded state sets.
  • The same three-scale pattern can be applied to groundwater irrigation control and other nonlinear evolutionary PDE systems with discrete actuators.

Where Pith is reading between the lines

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

  • A natural test of the 'probably sufficient' claim is to implement the three tiers on a small cascade with synthetic inflow forecasts and measure whether meso particles can supply valid warm starts fast enough; the paper itself does not report such an experiment.
  • The framework's weakest link is distribution shift: if the feasibility classifier is trained on one set of hydrological regimes and deployed on another, the catalog and particle evolution inherit its blind spots. A targeted stress test varying bathymetry or inflow statistics would reveal this.
  • The paper leaves closed-loop stability and risk consistency as open. Those are precisely the properties that would need to be proven before the architecture could be certified for operational dam control.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper proposes a three-time-scale computational architecture for real-time mixed-integer optimal control of systems governed by nonlinear hyperbolic PDEs, with the Saint–Venant shallow-water equations as the motivating application. The three layers are an offline HPC catalog of high-quality reference trajectories, a meso-scale approximate-dynamic-programming/particle-evolution layer, and a real-time Newton/MPC layer with model-order reduction. The paper surveys relevant theory in PDE-constrained optimization, mixed-integer control, stochastic programming, and approximate dynamic programming; presents a purported existence theorem for generalized solutions of the Saint–Venant system in Colombeau algebras; and identifies the learning of a feasibility map for the control-to-state operator as the critical component. The central claim, stated in Section 4, is that the proposed three-time-scale set of operations is 'simultaneously necessary and probably sufficient' for the target control problem, but this claim is explicitly presented as a belief and is not supported by numerical validation or a proof of sufficiency.

Significance. If the framework were validated, it would provide a useful organizing structure for a genuinely difficult class of problems at the intersection of PDE-constrained optimization, mixed-integer decisions, uncertainty, and real-time execution. The paper is honest about its aspirational nature, and its comprehensive literature review and explicit enumeration of open problems (e.g., duality for Colombeau algebras, learnability of feasibility maps) could help focus future research. However, the current manuscript offers no computational evidence, the main existence theorem is proved in a single hand-waving sentence, and the key feasibility-map premise is untested. As it stands, the work is a research agenda rather than a demonstrated methodology, and its contribution is accordingly prospective.

major comments (5)
  1. [Section 3.1, Theorem 3.2] The proof of Theorem 3.2 consists of one sentence: 'The bound follows immediately by integration with respect to t and x and general compactness and boundedness arguments.' This does not verify the hypotheses of Schauder's fixed-point theorem (continuity and compactness of the iteration map, invariance of the domain), nor does it establish the bound (10) from the assumptions (9). Moreover, equation (9) uses the same constants K_l, K_u for both Q/A and gA/2, which is dimensionally inconsistent (Q/A has units 1/time, gA has units length/time^2). As written, the theorem is not a convincing mathematical result. The authors should either provide a complete, rigorous proof or clearly state this as a conjecture and remove the theorem-like presentation.
  2. [Section 5.4 and 5.3.1] The paper correctly identifies the feasibility classifier ψ_ω as 'critical' (§5.4), since the offline catalog, the ADP terminal value, and the real-time warm starts all inherit its errors. However, no evidence is provided that such a classifier can be learned reliably for the Saint–Venant system: there are no experiments, no generalization bounds, and no analysis of the geometry of the feasible control-to-state set. The feasible set is likely fragmented (wetting/drying, shocks, switching), and the existence theorem (Thm 3.2) says nothing about its learnability. A concrete test is needed: e.g., train ψ_ω on a set of inflow scenarios and report precision/recall on out-of-distribution scenarios, or prove that a discretized piecewise-linear model yields a semi-algebraic feasible set with bounded complexity. Without such support, the central 'probably sufficient' claim rests on an untested em
  3. [Section 4] The paper claims that the three-time-scale architecture is 'simultaneously necessary and probably sufficient' for the control task. The sufficiency part is stated as a belief, but the necessity part is not argued at all. No complexity argument or counterexample shows that a two-scale approach is impossible. Since the abstract also disclaims proposing a complete solution, the claim should be weakened to 'a plausible and potentially sufficient framework' unless the authors can provide concrete evidence, e.g., a simplified but nontrivial instance where the offline-meso-real-time separation is required to meet a stated latency budget.
  4. [Section 6.2, Eqs. (16)–(17)] The text states 'Due to convexity, there is one unique solution' for the QCQP subproblems. However, the tangential subproblem (17) contains quadratic inequality constraints of the form -η(...) ≤ e' t + (1/2) t^T e'' t ≤ η(...), which are not convex in general unless e'' is positive semidefinite with appropriate structure. Additionally, the assertion that 'the optimality conditions define a linear PDE' is not justified for a QCQP with quadratic constraints; the KKT system generally involves complementarity conditions and may be nonlinear. The authors should either specify the convexity/regularity assumptions on the Hessians of the constraints and objective or revise the statement. This issue affects the theoretical grounding of the proposed SQCQP algorithm.
  5. [Overall (no single section)] The paper presents no numerical experiments or simulations, despite proposing a specific algorithmic architecture and claiming sufficiency for a challenging real-time control task. A proof-of-concept on a simplified but non-trivial Saint–Venant control problem (e.g., a one-dimensional dam with a small catalog, a few uncertainty scenarios, and a simulated real-time loop) would greatly strengthen the credibility of the framework. Without any such validation, the paper remains speculative. This is particularly important because the feasibility-map learning (major comment 2) and the algorithmic claims (major comment 4) can only be assessed empirically.
minor comments (5)
  1. [Section 3.1, Eq. (10)] Equation (10) lists ∥A∥_L∞ and ∥A^{-1}∥_L∞ twice; presumably one pair should be ∥Q∥_L∞ and ∥Q^{-1}∥_L∞. Please correct the typo.
  2. [Section 5.3.1] The classifier ψ_ω is described as predicting 'probability of a stable PDE solution' but the labels are binary solver convergence flags. Solver failure may be due to numerical discretization artifacts rather than non-existence of a PDE solution. This distinction should be clarified.
  3. [Section 6.2.2] The filter acceptance criterion uses 'F_trial < F_j or ϕ_trial < ϕ_j'; a strict inequality in both components would be more standard in filter methods. The current 'or' permits comparable points; please check the intended definition.
  4. [Section 10] The phrase 'the MGOpt framework' appears near the end of Section 9 but is not defined or cited. If this refers to a specific software tool, please add a reference.
  5. [Section 7] The non-parametric ADP terminal value V_ADP(x) ≈ min_c J_c + ω(∥x - x_c∥) is a reasonable nearest-neighbor-type approximation, but the choice of the stabilization weight ω is not discussed. A sentence on how ω is selected or learned would improve reproducibility.

Circularity Check

0 steps flagged

No circular derivation: the paper is an explicitly aspirational framework assembled from external component tools, with only minor non-load-bearing self-citations.

full rationale

The paper does not claim to derive a numerical prediction from fitted constants. Its central object is a proposed three-time-scale architecture (offline catalog, meso ADP/MPC, real-time Newton/MOR) that is presented as a research framework rather than as a proved result; the wording 'we believe constitutes ... probably sufficient' is explicitly an opinion, not a derivation. The offline catalog is assembled from established mixed-integer methods (B&B, DFO, RL), and the meso/real-time layers are assembled from standard MPC, ADP, trust-region SQCQP, and RTI components. No step in the derivation chain reduces to its own input by construction. The feasibility classifier ψ_ω is trained from CFD convergence labels and used to restrict the catalog; this is a surrogate model, not a fitted parameter disguised as a prediction of the paper's core claim. The ADP terminal value is explicitly a data-driven approximation from the catalog, not an independent first-principles prediction. Self-citations appear (Kimiaei–Neumaier DFO and integer-sequence mutation; Kungurtsev multi-periodic noise; Berardi–Difonzo irrigation), but these are isolated component tools and none is load-bearing for the central framework. Theorem 3.2's proof is very terse and may be underjustified, but the issue is mathematical rigor rather than circularity: the boundedness conclusion is not shown to be merely a restatement of the hypotheses in a way that would make the theorem vacuous. Overall, the manuscript is self-contained as a research roadmap and contains no circular derivation; the score reflects only the presence of minor, non-load-bearing self-citations.

Axiom & Free-Parameter Ledger

3 free parameters · 5 axioms · 0 invented entities

The paper introduces no new physical entities; its framework is composed of existing algorithmic components. The load-bearing assumptions are the suitability of Colombeau algebras for non-conservative shallow-water equations, the reliance on prior semilinear existence results, and the learnability of the feasible control-to-state map. The only hand-chosen parameters are the state bounds in Theorem 3.2 and standard algorithmic tuning constants.

free parameters (3)
  • State bounds l_A, u_A, l_Q, u_Q
    Assumed pointwise bounds on the solution (Eq. 8) that make the Schauder fixed-point argument in Theorem 3.2 plausible. They are chosen by hand and are not derived; if these bounds are not satisfied, the theorem's hypotheses fail.
  • Barrier parameter β_B
    Introduced in Section 2.5.3 to enforce state constraints approximately. The rate at which β_B should be decreased is not specified; it affects the convergence and feasibility of the meso-layer optimization.
  • Trust-region and filter parameters (Δ_min, Δ_max, γ_inc, γ_dec, ρ_good, ρ_bad, η)
    Standard algorithmic parameters for the SQCQP method in Algorithm 4. They are chosen by hand and influence convergence, but the paper does not propose any tuning procedure or concrete values.
axioms (5)
  • standard math Schauder's fixed point theorem
    Used in Theorem 3.2 to claim existence of a fixed point for the double-loop iteration in Colombeau algebras. Standard result, but its application requires the constructed map to be compact and continuous on a convex set, which is not established in the proof.
  • domain assumption Colombeau algebras are the appropriate solution framework for non-conservative shallow-water equations
    Section 3.1 asserts that non-conservative form makes Young measures and statistical solutions inapplicable, and that Colombeau algebras are 'the most appropriate functional analysis toolbox.' This is a modeling choice, not a theorem, and it underpins the existence argument.
  • domain assumption Existence of generalized solutions for semilinear hyperbolic systems (Oberguggenberger [86])
    The double-loop fixed-point scheme is built on the existence theory in [86], which applies to semilinear forms. The authors argue their nonlinear system can be handled by freezing coefficients, but this is not fully justified.
  • domain assumption The numerical PDE solver used offline converges for catalog generation
    The entire offline layer (Section 5) treats the CFD solver as a reliable oracle. The paper acknowledges that some initial/boundary conditions lead to solver failure, and the ML feasibility classifier ψ_ω is proposed to filter them, but no guarantees are given for the classifier's accuracy.
  • domain assumption The control-to-state map has learnable 'domains of attraction' that can be identified via ML
    Sections 5.3 and 5.4 assume that feasible control configurations form connected regions that a supervised classifier can learn. This is plausible but unproven, and it is essential for the catalog-based warm-starting strategy.

pith-pipeline@v1.3.0-alltime-deepseek · 39310 in / 9030 out tokens · 464759 ms · 2026-08-03T16:08:30.620407+00:00 · methodology

0 comments
read the original abstract

This paper examines aspirational requirements for software addressing mixed-integer optimization problems constrained by the nonlinear Shallow Water partial differential equations (PDEs), motivated by applications such as river-flow management in hydropower cascades. Realistic deployment of such software would require the simultaneous treatment of nonlinear and potentially non-smooth PDE dynamics, limited theoretical guarantees on the existence and regularity of control-to-state mappings under varying boundary conditions, and computational performance compatible with operational decision-making. In addition, practical settings motivate consideration of uncertainty arising from forecasts of demand, inflows, and environmental conditions. At present, the theoretical foundations, numerical optimization methods, and large-scale scientific computing tools required to address these challenges in a unified and tractable manner remain the subject of ongoing research across the associated research communities. Rather than proposing a complete solution, this work uses the problem as a case study to identify and organize the mathematical, algorithmic, and computational components that would be necessary for its realization. The resulting framework highlights open challenges and intermediate research directions, and may inform both more circumscribed related problems and the design of future large-scale collaborative efforts aimed at addressing such objectives.

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

123 extracted references · 3 linked inside Pith

  1. [1]

    A new approach for designing well-balanced schemes for the shallow water.Appl

    Remi Abgrall and Yongle Liu. A new approach for designing well-balanced schemes for the shallow water.Appl. Math. Comput, 5:370–402, 2023

    2023

  2. [2]

    A. Alla, M. Berardi, and L. Saluzzi. State Dependent Riccati for dynamic boundary control to optimize irrigation in Richards’ equation framework. Mathematics and Computers in Simulation, 232:261–275, 2025

    2025

  3. [3]

    Birkh¨ auser, 2012

    Frank Allg¨ ower and Alex Zheng.Nonlinear model predictive control, vol- ume 26. Birkh¨ auser, 2012

    2012

  4. [4]

    Quasilinear Elliptic-Parabolic Differential Equations.Mathematische Zeitschrift, 183(3):311–341, 1983

    Hans Wilhelm Alt and Stefan Luckhaus. Quasilinear Elliptic-Parabolic Differential Equations.Mathematische Zeitschrift, 183(3):311–341, 1983

    1983

  5. [5]

    Charles Audet and J. E. Dennis. Mesh adaptive direct search algo- rithms for constrained optimization.SIAM Journal on Optimization, 17(1):188–217, January 2006

    2006

  6. [6]

    A Fast and Stable Well-Balanced Scheme 59 with Hydrostatic Reconstruction for Shallow Water Flows.SIAM Journal on Scientific Computing, 25(6):2050–2065, 2004

    Emmanuel Audusse, Fran¸ cois Bouchut, Marie-Odile Bristeau, Rupert Klein, and Beno ˆ ıt Perthame. A Fast and Stable Well-Balanced Scheme 59 with Hydrostatic Reconstruction for Shallow Water Flows.SIAM Journal on Scientific Computing, 25(6):2050–2065, 2004

    2050

  7. [7]

    Con- vergence of trust-region methods based on probabilistic models.SIAM Journal on Optimization, 24(3):1238–1264, 2014

    Afonso S Bandeira, Katya Scheinberg, and Luis Nunes Vicente. Con- vergence of trust-region methods based on probabilistic models.SIAM Journal on Optimization, 24(3):1238–1264, 2014

    2014

  8. [8]

    Elsevier, New York, 1972

    Jacob Bear.Dynamics of Fluids in Porous Media. Elsevier, New York, 1972

    1972

  9. [9]

    Jacob Bear and Arnold Verruijt.Modeling Groundwater Flow and Pollu- tion. D. Reidel Publishing Company, 1987

    1987

  10. [10]

    Real-time pde constrained optimal control of a periodic multicomponent separation process.Trends in PDE Constrained Optimization, pages 521–537, 2014

    Malte Behrens, Hans Georg Bock, Sebastian Engell, Phawitphorn Khobkhun, and Andreas Potschka. Real-time pde constrained optimal control of a periodic multicomponent separation process.Trends in PDE Constrained Optimization, pages 521–537, 2014

    2014

  11. [11]

    SIAM, 2017

    Peter Benner, Mario Ohlberger, Albert Cohen, and Karen Willcox.Model reduction and approximation: theory and algorithms. SIAM, 2017

    2017

  12. [12]

    Interpretation of runoff processes in hydrological modelling—experience from the hbv approach.Hydrologi- cal Processes, 29(16):3535–3545, 2015

    Sten Bergstr¨ om and G¨ oran Lindstr¨ om. Interpretation of runoff processes in hydrological modelling—experience from the hbv approach.Hydrologi- cal Processes, 29(16):3535–3545, 2015

    2015

  13. [13]

    SIAM, 2007

    Lorenz T Biegler, Omar Ghattas, Matthias Heinkenschloss, David Keyes, and Bart van Bloemen Waanders.Real-time PDE-constrained Optimiza- tion. SIAM, 2007

    2007

  14. [14]

    Schl¨ oder.1

    Hans Georg Bock, Moritz Diehl, Ekaterina Kostina, and Johannes P. Schl¨ oder.1. Constrained Optimal Feedback Control of Systems Governed by Large Differential Algebraic Equations, page 3–24. Society for Indus- trial and Applied Mathematics, January 2007

    2007

  15. [15]

    Borz ` ı and G

    A. Borz ` ı and G. von Winckel. Multigrid methods and sparse-grid colloca- tion techniques for parabolic optimal control problems with random coeffi- cients.SIAM Journal on Scientific Computing, 31(3):2172–2192, January 2009

    2009

  16. [16]

    Iterative Solution of Piecewise Linear Systems.SIAM J

    Luigi Brugnano and Vincenzo Casulli. Iterative Solution of Piecewise Linear Systems.SIAM J. Scientific Computing, 30:463–472, 01 2008

    2008

  17. [17]

    Iterative Solution of Piecewise Linear Systems and Applications to Flows in Porous Media.SIAM Journal on Scientific Computing, 31(3):1858–1873, 2009

    Luigi Brugnano and Vincenzo Casulli. Iterative Solution of Piecewise Linear Systems and Applications to Flows in Porous Media.SIAM Journal on Scientific Computing, 31(3):1858–1873, 2009

    2009

  18. [18]

    Ordinary differential equation methods for markov decision processes and application to kullback–leibler control cost

    Ana Buˇ si´ c and Sean Meyn. Ordinary differential equation methods for markov decision processes and application to kullback–leibler control cost. SIAM Journal on Control and Optimization, 56(1):343–366, January 2018. 60

    2018

  19. [19]

    Yanzhao Cao, Somak Das, and Hans-Werner van Wyk. Adaptive stochas- tic gradient descent for optimal control of parabolic equations with ran- dom parameters.Numerical Methods for Partial Differential Equations, 38(6):2104–2122, February 2022

    2022

  20. [20]

    Inverse Problem in Hydrogeology.Water Resources Research, 41:W09424, 2005

    Jesus Carrera, Alvaro Alcolea, Agust ´ ı Medina, Julian Hidalgo, and Lucas Slooten. Inverse Problem in Hydrogeology.Water Resources Research, 41:W09424, 2005

    2005

  21. [21]

    Springer, 2019

    Oscar Castro-Orgaz, Willi H Hager, et al.Shallow water hydraulics. Springer, 2019

    2019

  22. [22]

    High resolution methods for multi- dimensional advection–diffusion problems in free-surface hydrodynamics

    Vincenzo Casulli and Paola Zanolli. High resolution methods for multi- dimensional advection–diffusion problems in free-surface hydrodynamics. Ocean Modelling, 10(1):137–151, 2005. The Second International Work- shop on Unstructured Mesh Numerical Modelling of Coastal, Shelf and Ocean Flows

    2005

  23. [23]

    A Nested Newton-Type Algorithm for Finite Volume Methods Solving Richards’ Equation in Mixed Form

    Vincenzo Casulli and Paola Zanolli. A Nested Newton-Type Algorithm for Finite Volume Methods Solving Richards’ Equation in Mixed Form. SIAM Journal on Scientific Computing, 32(4):2255–2273, 2010

    2010

  24. [24]

    Celia, Edward T

    Michael A. Celia, Edward T. Bouloutas, and Rebecca L. Zarba. A General Mass-Conservative Numerical Solution for the Unsaturated Flow Equa- tion.Water Resources Research, 26(7):1483–1496, 1990

    1990

  25. [25]

    Tidal, residual, intertidal mudflat (trim) model and its applications to san francisco bay, california.Estuarine, Coastal and Shelf Science, 36(3):235–280, 1993

    Ralph T Cheng, Vincenzo Casulli, and Jeffrey W Gartner. Tidal, residual, intertidal mudflat (trim) model and its applications to san francisco bay, california.Estuarine, Coastal and Shelf Science, 36(3):235–280, 1993

    1993

  26. [26]

    Elsevier, 2011

    Jean Fran¸ cois Colombeau.Elementary introduction to new generalized functions, volume 113. Elsevier, 2011

    2011

  27. [27]

    Springer, 2010

    Antonio J Conejo, Miguel Carri´ on, Juan M Morales, et al.Decision mak- ing under uncertainty in electricity markets, volume 1. Springer, 2010

    2010

  28. [28]

    Conn, Katya Scheinberg, and Luis N

    Andrew R. Conn, Katya Scheinberg, and Luis N. Vicente.Introduction to Derivative-Free Optimization. Society for Industrial and Applied Mathe- matics, January 2009

    2009

  29. [29]

    Real time iterations for mixed-integer model predictive control

    Massimo De Mauri, Wim Van Roy, Joris Gillis, Jan Swevers, and Goele Pipeleers. Real time iterations for mixed-integer model predictive control. In2020 European Control Conference (ECC), pages 699–705. IEEE, 2020

    2020

  30. [30]

    Marie Delaunay and Florin A. Radu. On the Stability and Convergence of Linearization Schemes for Richards Equation.SIAM Journal on Nu- merical Analysis, 55(2):540–565, 2017

    2017

  31. [31]

    Springer Science & Business Media, 2011

    Peter Deuflhard.Newton methods for nonlinear problems: affine invari- ance and adaptive algorithms, volume 35. Springer Science & Business Media, 2011. 61

    2011

  32. [32]

    Castro D ´ ıaz, Alexander Kurganov, and Tom´ as Morales de Luna

    Mar ´ ıa J. Castro D ´ ıaz, Alexander Kurganov, and Tom´ as Morales de Luna. Path-Conservative Central-Upwind Schemes for Nonconservative Hyper- bolic Systems.ESAIM: Mathematical Modelling and Numerical Analysis, 53(3):959–985, 2019

    2019

  33. [33]

    A real-time iteration scheme for nonlinear optimization in optimal feedback control

    Moritz Diehl, Hans Georg Bock, and Johannes P Schl¨ oder. A real-time iteration scheme for nonlinear optimization in optimal feedback control. SIAM Journal on control and optimization, 43(5):1714–1736, 2005

    2005

  34. [34]

    Sakine Esmaili, M. R. Eslahchi, and Delfim F. M. Torres. Optimal control for a nonlinear stochastic pde model of cancer growth.Optimization, 73(9):2745–2789, July 2023

    2023

  35. [35]

    Binary control pulse optimization for quantum systems.Quantum, 7:892, 2023

    Xinyu Fei, Lucas T Brady, Jeffrey Larson, Sven Leyffer, and Siqian Shen. Binary control pulse optimization for quantum systems.Quantum, 7:892, 2023

    2023

  36. [36]

    A brief history of filter methods.Preprint ANL/MCS-P1372-0906, Argonne National Laboratory, Mathematics and Computer Science Division, 36, 2006

    Roger Fletcher, Sven Leyffer, Ph L Toint, et al. A brief history of filter methods.Preprint ANL/MCS-P1372-0906, Argonne National Laboratory, Mathematics and Computer Science Division, 36, 2006

    2006

  37. [37]

    Calculus of variations and optimal control for gener- alized functions.Nonlinear Analysis, 216:112718, 2022

    Gastao SF Frederico, Paolo Giordano, Alexandr A Bryzgalov, and Matheus J Lazo. Calculus of variations and optimal control for gener- alized functions.Nonlinear Analysis, 216:112718, 2022

    2022

  38. [38]

    Topological structures in colombeau algebras: investiga- tion of the duals ofG c(Ω),G(Ω) andG S (Rn).Monatshefte f¨ ur Mathematik, 146(3):203–226, 2005

    Claudia Garetto. Topological structures in colombeau algebras: investiga- tion of the duals ofG c(Ω),G(Ω) andG S (Rn).Monatshefte f¨ ur Mathematik, 146(3):203–226, 2005

    2005

  39. [39]

    Improved penalty algorithm for mixed integer PDE constrained op- timization problems.Computers & Mathematics with Applications, 116:2– 14, 2022

    Dominik Garmatter, Margherita Porcelli, Francesco Rinaldi, and Martin Stoll. Improved penalty algorithm for mixed integer PDE constrained op- timization problems.Computers & Mathematics with Applications, 116:2– 14, 2022

    2022

  40. [40]

    Surowiec, and Michael Ulbrich

    Sebastian Garreis, Thomas M. Surowiec, and Michael Ulbrich. An Interior-Point Approach for Solving Risk-Averse PDE-Constrained Op- timization Problems with Coherent Risk Measures.SIAM Journal on Optimization, 31(1):1–29, 2021

    2021

  41. [41]

    Optimality conditions for convex stochastic optimization problems in banach spaces with almost sure state constraints.SIAM Journal on Optimization, 31(4):2455–2480, 2021

    Caroline Geiersbach and Winnifried Wollner. Optimality conditions for convex stochastic optimization problems in banach spaces with almost sure state constraints.SIAM Journal on Optimization, 31(4):2455–2480, 2021

    2021

  42. [42]

    Future paths for integer programming and links to artificial intelligence.Computers & Operations Research, 13(5):533–549, 1986

    Fred Glover. Future paths for integer programming and links to artificial intelligence.Computers & Operations Research, 13(5):533–549, 1986

    1986

  43. [43]

    Tabu search—part i.ORSA Journal on Computing, 1(3):190– 206, 1989

    Fred Glover. Tabu search—part i.ORSA Journal on Computing, 1(3):190– 206, 1989. 62

    1989

  44. [44]

    Tabu search—part ii.ORSA Journal on Computing, 2(1):4– 32, 1990

    Fred Glover. Tabu search—part ii.ORSA Journal on Computing, 2(1):4– 32, 1990

    1990

  45. [45]

    Quadratic programming with one quadratic constraint in hilbert spaces.Annals of Functional Analysis, 16(2):31, 2025

    Santiago Gonzalez Zerbo, Alejandra Maestripieri, and Francisco Mart ´ ınez Per ´ ıa. Quadratic programming with one quadratic constraint in hilbert spaces.Annals of Functional Analysis, 16(2):31, 2025

    2025

  46. [46]

    Springer, 1990

    J¨ urgen Guddat, F Guerra Vazquez, and Hubertus Th Jongen.Parametric optimization: singularities, pathfollowing and jumps. Springer, 1990

    1990

  47. [47]

    Turnpike properties of optimal bound- ary control problems with random linear hyperbolic systems.ESAIM: Control, Optimisation and Calculus of Variations, 29:55, 2023

    Martin Gugat and Michael Herty. Turnpike properties of optimal bound- ary control problems with random linear hyperbolic systems.ESAIM: Control, Optimisation and Calculus of Variations, 29:55, 2023

    2023

  48. [48]

    Guth, Vesa Kaarnioja, Frances Y

    Philipp A. Guth, Vesa Kaarnioja, Frances Y. Kuo, Claudia Schillings, and Ian H. Sloan. Parabolic PDE-constrained optimal control under un- certainty with entropic risk measure using quasi-Monte Carlo integration. Numerische Mathematik, 156(2):565–608, March 2024

    2024

  49. [49]

    Decomposition and approximation for PDE-constrained mixed-integer optimal control

    Mirko Hahn, Christian Kirches, Paul Manns, Sebastian Sager, and Clemens Zeile. Decomposition and approximation for PDE-constrained mixed-integer optimal control. InNon-Smooth and Complementarity- Based Distributed Parameter Systems: Simulation and Hierarchical Opti- mization, pages 283–305. Springer, 2021

    2021

  50. [50]

    Mixed-integer PDE- constrained optimal control of gas networks.Mathematics and Computer Science, 113, 2017

    Mirko Hahn, Sven Leyffer, and Victor M Zavala. Mixed-integer PDE- constrained optimal control of gas networks.Mathematics and Computer Science, 113, 2017

    2017

  51. [51]

    Coordinated predictive control of a hydropower cascade

    Andrew Hamann. Coordinated predictive control of a hydropower cascade. Pittsburgh, Pennsylvania: Carnegie Mellon University, 2015

    2015

  52. [52]

    Springer Berlin Heidelberg

    Nikolaus Hansen.The CMA Evolution Strategy: A Comparing Review, pages 75–102. Springer Berlin Heidelberg

  53. [53]

    A matrix-free trust-region sqp method for equality constrained optimization.SIAM Journal on Op- timization, 24(3):1507–1541, 2014

    Matthias Heinkenschloss and Denis Ridzal. A matrix-free trust-region sqp method for equality constrained optimization.SIAM Journal on Op- timization, 24(3):1507–1541, 2014

    2014

  54. [54]

    Springer, Berlin, 1997

    Rainer Helmig.Multiphase Flow and Transport Processes in the Subsur- face: A Contribution to the Modeling of Hydrosystems. Springer, Berlin, 1997

    1997

  55. [55]

    Pde-constrained optimization subject to pointwise constraints on the control, the state, and its deriva- tive.SIAM Journal on Optimization, 20(3):1133–1156, 2010

    Michael Hinterm¨ uller and Karl Kunisch. Pde-constrained optimization subject to pointwise constraints on the control, the state, and its deriva- tive.SIAM Journal on Optimization, 20(3):1133–1156, 2010

    2010

  56. [56]

    Optimization methods in banach spaces.Optimization with PDE Constraints, pages 97–156, 2009

    Michael Hinze, Rene Pinnau, Michael Ulbrich, Stefan Ulbrich, and Michael Ulbrich. Optimization methods in banach spaces.Optimization with PDE Constraints, pages 97–156, 2009. 63

    2009

  57. [57]

    Holland.Adaptation in Natural and Artificial Systems: An In- troductory Analysis with Applications to Biology, Control, and Artificial Intelligence

    John H. Holland.Adaptation in Natural and Artificial Systems: An In- troductory Analysis with Applications to Biology, Control, and Artificial Intelligence. The MIT Press, April 1992

    1992

  58. [58]

    Advanced step non- linear model predictive control for air separation units.Journal of Process Control, 19(4):678–685, 2009

    Rui Huang, Victor M Zavala, and Lorenz T Biegler. Advanced step non- linear model predictive control for air separation units.Journal of Process Control, 19(4):678–685, 2009

    2009

  59. [59]

    Ergodic optimization in dynamical systems.Ergodic Theory and Dynamical Systems, 39(10):2593–2618, 2019

    Oliver Jenkinson. Ergodic optimization in dynamical systems.Ergodic Theory and Dynamical Systems, 39(10):2593–2618, 2019

    2019

  60. [60]

    Johnson, Christos H

    David S. Johnson, Christos H. Papadimitriou, and Mihalis Yannakakis. How easy is local search?Journal of Computer and System Sciences, 37(1):79–100, 1988

    1988

  61. [61]

    Springer US, 2011

    James Kennedy.Particle Swarm Optimization, page 760–766. Springer US, 2011

    2011

  62. [62]

    Efficient unconstrained black box optimization.Mathematical Programming Computation, 14(2):365–414, February 2022

    Morteza Kimiaei and Arnold Neumaier. Efficient unconstrained black box optimization.Mathematical Programming Computation, 14(2):365–414, February 2022

    2022

  63. [63]

    Effective matrix adaptation strat- egy for noisy derivative-free optimization.Mathematical Programming Computation, 16(3):459–501, July 2024

    Morteza Kimiaei and Arnold Neumaier. Effective matrix adaptation strat- egy for noisy derivative-free optimization.Mathematical Programming Computation, 16(3):459–501, July 2024

    2024

  64. [64]

    MATRS: heuristic methods for noisy derivative-free bound-constrained mixed-integer optimization

    Morteza Kimiaei and Arnold Neumaier. MATRS: heuristic methods for noisy derivative-free bound-constrained mixed-integer optimization. Mathematical Programming Computation, 17(3):505–546, May 2025

    2025

  65. [65]

    Springer, 2011

    Christian Kirches.Fast numerical methods for mixed-integer nonlinear model-predictive control. Springer, 2011

    2011

  66. [66]

    Foun- dations of the calculus of variations in generalized function algebras.Acta Applicandae Mathematicae, 103(2):169–199, 2008

    Sanja Konjik, Michael Kunzinger, and Michael Oberguggenberger. Foun- dations of the calculus of variations in generalized function algebras.Acta Applicandae Mathematicae, 103(2):169–199, 2008

    2008

  67. [67]

    A trust-region algorithm with adaptive stochastic collocation for pde optimization under uncertainty.SIAM Journal on Scientific Computing, 35(4):A1847–A1879, 2013

    Drew P Kouri, Matthias Heinkenschloss, Denis Ridzal, and Bart G van Bloemen Waanders. A trust-region algorithm with adaptive stochastic collocation for pde optimization under uncertainty.SIAM Journal on Scientific Computing, 35(4):A1847–A1879, 2013

    2013

  68. [68]

    Inexact objective function evaluations in a trust- region algorithm for PDE-constrained optimization under uncertainty

    Drew P Kouri, Matthias Heinkenschloss, Denis Ridzal, and Bart G van Bloemen Waanders. Inexact objective function evaluations in a trust- region algorithm for PDE-constrained optimization under uncertainty. SIAM Journal on Scientific Computing, 36(6):A3011–A3029, 2014

    2014

  69. [69]

    Kouri, Mathias Staudigl, and Thomas M

    Drew P. Kouri, Mathias Staudigl, and Thomas M. Surowiec. A relaxation- based probabilistic approach for PDE-constrained optimization under un- certainty with pointwise state constraints.Computational Optimization and Applications, 85(2):441–478, February 2023. 64

    2023

  70. [70]

    Risk-averse PDE-constrained optimization using the conditional value-at-risk.SIAM Journal on Opti- mization, 26(1):365–396, 2016

    Drew P Kouri and Thomas M Surowiec. Risk-averse PDE-constrained optimization using the conditional value-at-risk.SIAM Journal on Opti- mization, 26(1):365–396, 2016

    2016

  71. [71]

    Quasi-ergodic control of multi-periodic au- toregressive processes: Formulation and examples.arXiv preprint arXiv:2509.24729, 2025

    Vyacheslav Kungurtsev. Quasi-ergodic control of multi-periodic au- toregressive processes: Formulation and examples.arXiv preprint arXiv:2509.24729, 2025

    arXiv 2025

  72. [72]

    Sequential quadratic program- ming methods for parametric nonlinear optimization.Computational Op- timization and Applications, 59(3):475–509, 2014

    Vyacheslav Kungurtsev and Moritz Diehl. Sequential quadratic program- ming methods for parametric nonlinear optimization.Computational Op- timization and Applications, 59(3):475–509, 2014

    2014

  73. [73]

    A predictor-corrector path-following algorithm for dual-degenerate parametric optimization problems.SIAM Journal on Optimization, 27(1):538–564, 2017

    Vyacheslav Kungurtsev and Johannes Jaschke. A predictor-corrector path-following algorithm for dual-degenerate parametric optimization problems.SIAM Journal on Optimization, 27(1):538–564, 2017

    2017

  74. [74]

    A Second-Order Well- Balanced Positivity-Preserving Central-Upwind Scheme for the Saint– Venant System.Communications in Mathematical Sciences, 5(1):133–160, 2007

    Alexander Kurganov and Guergana Petrova. A Second-Order Well- Balanced Positivity-Preserving Central-Upwind Scheme for the Saint– Venant System.Communications in Mathematical Sciences, 5(1):133–160, 2007

    2007

  75. [75]

    A. H. Land and A. G. Doig. An automatic method of solving discrete programming problems.Econometrica, 28(3):497, July 1960

    1960

  76. [76]

    LeVeque.Finite Volume Methods for Hyperbolic Problems

    Randall J. LeVeque.Finite Volume Methods for Hyperbolic Problems. Cambridge Texts in Applied Mathematics. Cambridge University Press, 2002

    2002

  77. [77]

    Sequential linear integer programming for integer optimal control with total variation regularization.ESAIM: Control, Optimisation and Calculus of Variations, 28:66, 2022

    Sven Leyffer and Paul Manns. Sequential linear integer programming for integer optimal control with total variation regularization.ESAIM: Control, Optimisation and Calculus of Variations, 28:66, 2022

    2022

  78. [78]

    Existence and uniqueness of solutions to nonlinear evolution equations with locally monotone operators.Nonlinear Analysis: Theory, Methods & Applications, 74(18):7543–7561, 2011

    Wei Liu. Existence and uniqueness of solutions to nonlinear evolution equations with locally monotone operators.Nonlinear Analysis: Theory, Methods & Applications, 74(18):7543–7561, 2011

    2011

  79. [79]

    A derivative-free algorithm for bound constrained optimization.Computational Optimization and Appli- cations, 21(2):119–142, February 2002

    Stefano Lucidi and Marco Sciandrone. A derivative-free algorithm for bound constrained optimization.Computational Optimization and Appli- cations, 21(2):119–142, February 2002

    2002

  80. [80]

    V., and Guglielmi R

    Berardi M., Difonzo F. V., and Guglielmi R. A preliminary model for optimal control of moisture content in unsaturated soils.Computational Geosciences, 27:1133–1144, 2023

    2023

Showing first 80 references.