pith. sign in

arxiv: 2605.12751 · v1 · pith:XG335CY3new · submitted 2026-05-12 · 🧮 math.OC

Asymptotic KKT Conditions for Continuous-Time Nonlinear Programming

Mois\'es R. C. do Monte , Rodrigo B. Moreira , Valeriano A. de Oliveira This is my paper

Pith reviewed 2026-05-14 20:12 UTC · model grok-4.3

classification 🧮 math.OC
keywords continuous-time nonlinear programmingasymptotic KKT conditionssequential optimality conditionsaugmented Lagrangian methodnecessary optimality conditionsconvex sufficiency
0
0 comments X

The pith

Continuous-time nonlinear programs have asymptotic KKT conditions satisfied along a convergent sequence of approximate solutions.

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

The paper develops sequential necessary optimality conditions for continuous-time nonlinear programming problems that include both equality and inequality constraints. It proves that a sequence of points converging to an optimal solution exists such that the Karush-Kuhn-Tucker conditions hold in the limit. These asymptotic conditions are also shown to be sufficient for optimality when the problem is convex. The authors introduce an augmented Lagrangian method for solving these problems and establish its convergence with respect to both feasibility and optimality. Such conditions supply practical stopping criteria for numerical algorithms applied to continuous-time problems.

Core claim

For continuous-time nonlinear programs with equality and inequality constraints, there exists a sequence of solutions converging to the optimal solution such that Karush-Kuhn-Tucker-type conditions are satisfied asymptotically; these sequential conditions are also sufficient for optimality under convexity assumptions.

What carries the argument

Asymptotic KKT conditions, defined as the limiting form of stationarity, complementarity, and feasibility conditions along a sequence of approximate solutions that converges to the true optimum.

If this is right

  • The asymptotic conditions supply a termination test for numerical solvers of continuous-time problems.
  • An augmented Lagrangian method is shown to generate sequences that converge to points satisfying the asymptotic conditions.
  • Under convexity the asymptotic conditions certify global optimality without further verification.
  • The sequential form remains valid for problems with both equality and inequality constraints.

Where Pith is reading between the lines

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

  • The same sequential pattern could be tested on other infinite-dimensional problems such as optimal control with state constraints.
  • Discretization schemes for continuous-time problems might be analyzed by checking whether the discrete KKT residuals approach zero along the refinement sequence.
  • The framework may link to existing sequential optimality results in finite-dimensional nonlinear programming by taking the continuous-time limit of the discretized problem.

Load-bearing premise

A sequence of approximate solutions exists that converges to the true optimum while satisfying the asymptotic KKT conditions in the limit.

What would settle it

An explicit continuous-time nonlinear program in which every sequence of feasible points converging to the optimum fails to satisfy at least one of the limiting stationarity, complementarity, or primal-feasibility conditions.

Figures

Figures reproduced from arXiv: 2605.12751 by Mois\'es R. C. do Monte, Rodrigo B. Moreira, Valeriano A. de Oliveira.

Figure 1
Figure 1. Figure 1: Numerical results for (x 0 1 (t), x0 2 (t)) ≡ (1, 1), N = 1050, and (˜v 1 1 (t), v˜ 1 2 (t)) ≡ (1, 1). The method performed 2 iterations [PITH_FULL_IMAGE:figures/full_fig_p019_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Numerical results for (x 0 1 (t), x0 2 (t)) ≡ (0.5, 0.5), N = 1050, and (˜v 1 1 (t), v˜ 1 2 (t), v˜ 1 3 (t)) ≡ (1, 1, 1), after 78 iterations. 19 [PITH_FULL_IMAGE:figures/full_fig_p019_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Numerical results for (x 0 1 (t), x0 2 (t), x0 3 (t)) ≡ (−100, −100, −100), M = N = 1050, (˜u 1 1 (t), v˜ 1 1 (t), v˜ 1 2 (t)) ≡ (1, 1, 1), and 20 iterations. Example 5.4 (de Oliveira [10, Example 1]). Consider the problem given as minimize Z 2 0 c(t) T x(t) dt subject to A(t)x(t) − b(t) ≤ 0 a.e. t ∈ [0, 2], x ∈ L∞([0, 2]; R 2 ), where, for almost every t ∈ [0, 2], c(t) =  (t − 1) sgn(1 − t) −1  , A(t) =… view at source ↗
Figure 4
Figure 4. Figure 4: shows the numerical results obtained by the proposed augmented La￾grangian method and the optimal solution [PITH_FULL_IMAGE:figures/full_fig_p021_4.png] view at source ↗
read the original abstract

This paper addresses the class of continuous-time nonlinear programming problems with equality and inequality constraints. The paper presents necessary optimality conditions of the sequential form. To be more precise, a sequence of solutions converging to the optimal solution is demonstrated to exist, and such that Karush-Kuhn-Tucker-type conditions are satisfied asymptotically. It is shown that these sequential Karush-Kuhn-Tucker-type conditions also become sufficient for optimality under convexity assumptions. Sequential optimality conditions are a valuable tool for determining when to terminate a numerical method of solution. In this regard, an augmented Lagrangian-type method is proposed for numerically solving continuous-time programming problems. A convergence analysis concerning viability and optimality is presented. The performance of the method is evaluated by applying it to solve instances of continuous-time problems found in the literature.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 2 minor

Summary. The paper claims that for continuous-time nonlinear programming problems with equality and inequality constraints, there exists a sequence of feasible points converging (in a suitable topology) to a local optimum such that asymptotic KKT-type conditions hold, i.e., the stationarity residual and complementarity gap vanish in the limit. These sequential conditions are shown to be sufficient for optimality under convexity assumptions. The paper further proposes an augmented-Lagrangian algorithm for solving such problems and provides a convergence analysis establishing both feasibility and optimality of the generated sequence.

Significance. If the central claims hold, the work supplies a theoretically grounded termination criterion for numerical methods in infinite-dimensional optimization and extends the finite-dimensional theory of asymptotic KKT conditions to a continuous-time setting. The accompanying augmented-Lagrangian scheme with explicit viability and optimality guarantees is a concrete algorithmic contribution that could be useful for applications in optimal control and dynamic optimization.

major comments (2)
  1. [§3] §3 (Necessity theorem): the existence of the approximating sequence x_k → x* satisfying the asymptotic stationarity and complementarity conditions is asserted without an explicit construction or invocation of a constraint qualification in the underlying function space (e.g., L^2 or Sobolev). The argument appears to rely on the later augmented-Lagrangian iterates rather than deriving the sequence directly from optimality of x* alone; this is load-bearing for the necessity claim.
  2. [Theorem 4.2] Theorem 4.2 (sufficiency): the proof that the asymptotic KKT conditions imply optimality under convexity assumes the multipliers are measurable and the limiting process preserves the inequality constraints pointwise, but no measurability or domination argument is supplied. In the continuous-time setting this step requires additional justification beyond the finite-dimensional case.
minor comments (2)
  1. [§2] Notation for the time-dependent multipliers and the precise topology of convergence (e.g., strong L^2 vs. weak*) should be introduced once and used consistently throughout §§2–4.
  2. [§5] The numerical examples in §5 would benefit from a table reporting the final KKT residual norms and CPU times for each instance to allow direct comparison with existing continuous-time solvers.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive comments. We address the major comments point by point below. Where revisions are needed to improve clarity or add justifications, we will incorporate them in the revised manuscript.

read point-by-point responses

  1. Referee: [§3] §3 (Necessity theorem): the existence of the approximating sequence x_k → x* satisfying the asymptotic stationarity and complementarity conditions is asserted without an explicit construction or invocation of a constraint qualification in the underlying function space (e.g., L^2 or Sobolev). The argument appears to rely on the later augmented-Lagrangian iterates rather than deriving the sequence directly from optimality of x* alone; this is load-bearing for the necessity claim.

    Authors: We thank the referee for highlighting this point. The necessity result in Section 3 is derived directly from local optimality of x* via a penalization construction in the underlying function space and does not depend on the augmented-Lagrangian algorithm introduced later in the paper. To eliminate any ambiguity, we will revise Section 3 to include an explicit construction of the sequence {x_k} together with a suitable constraint qualification (e.g., a Slater-type condition adapted to L^2). This will render the necessity theorem independent of the algorithmic section. revision: partial

  2. Referee: [Theorem 4.2] Theorem 4.2 (sufficiency): the proof that the asymptotic KKT conditions imply optimality under convexity assumes the multipliers are measurable and the limiting process preserves the inequality constraints pointwise, but no measurability or domination argument is supplied. In the continuous-time setting this step requires additional justification beyond the finite-dimensional case.

    Authors: We agree that the continuous-time setting requires additional technical justification. In the revised manuscript we will insert a supporting lemma that establishes measurability of the limiting multipliers (via weak compactness arguments in L^2) and supplies a domination argument, leveraging convexity, to guarantee that the inequality constraints are preserved pointwise almost everywhere in the limit. This will complete the proof of Theorem 4.2. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper derives asymptotic KKT conditions as necessary optimality conditions for continuous-time NLP by establishing existence of a converging sequence of approximate solutions satisfying the sequential stationarity and complementarity conditions. Sufficiency is shown separately under convexity. The augmented Lagrangian method is introduced afterward as a numerical algorithm with its own viability and optimality convergence analysis. No load-bearing step reduces the central existence claim to a fitted parameter, self-definition, or self-citation chain; the derivation remains independent of the solver and does not rename or smuggle in prior results by construction. The result is self-contained against standard sequential optimality theory.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The claims rest on standard smoothness and convexity assumptions typical for KKT theory in function spaces, plus the existence of a convergent approximating sequence whose construction is not detailed in the abstract.

axioms (2)
  • domain assumption The objective and constraint functions are sufficiently smooth (at least continuously differentiable) in the appropriate function space.
    Required for the definition of KKT multipliers and gradients in continuous time.
  • domain assumption A constraint qualification holds at the limit point so that multipliers exist.
    Standard for KKT necessity; not stated explicitly but implicit in the claim.

pith-pipeline@v0.9.0 · 5434 in / 1330 out tokens · 24761 ms · 2026-05-14T20:12:05.842932+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

40 extracted references · 40 canonical work pages

  1. [1]

    Andreani, N

    R. Andreani, N. S. Fazzio, M. L. Schuverdt, and L. D. Secchin. A sequen- tial optimality condition related to the quasi-normality constraint qualifi- cation and its algorithmic consequences.SIAM Journal on Optimization, 29(1):743–766, 2019

    work page 2019

  2. [2]

    Andreani, W

    R. Andreani, W. G´ omez, G. Haeser, L. M. Mito, and A. Ramos. On optimality conditions for nonlinear conic programming.Mathematics of Operations Research, 47(3):2160–2185, 2022

    work page 2022

  3. [3]

    Andreani, P.S

    R. Andreani, P.S. Gon¸ calves, and G. N. Silva. Discrete approximations for strict convex continuous time problems and duality.Comp. Appl. Math., 23:81–105, 2004

    work page 2004

  4. [4]

    Andreani, G

    R. Andreani, G. Haeser, and J. M. Mart´ ınez. On sequential optimality conditions for smooth constrained optimization.Optimization, 60(5):627– 641, 2011

    work page 2011

  5. [5]

    Andreani, G

    R. Andreani, G. Haeser, A. Ramos, and P. J. S. Silva. A second-order se- quential optimality condition associated to the convergence of optimization algorithms.IMA Journal of Numerical Analysis, 37(4):1902–1929, 2017. 22

    work page 1902

  6. [6]

    Andreani, J

    R. Andreani, J. M. Mart´ ınez, and B. F. Svaiter. A new sequential optimal- ity condition for constrained optimization and algorithmic consequences. SIAM Journal on Optimization, 20(6):3533–3554, 2010

    work page 2010

  7. [7]

    R. Bellman. Bottleneck problems and dynamic programming. InPro- ceedings of the National Academy of Sciences, volume 39, pages 947–951. National Acad Sciences, 1953

    work page 1953

  8. [8]

    B¨ orgens, C

    E. B¨ orgens, C. Kanzow, P. Mehlitz, and G. Wachsmuth. New constraint qualifications for optimization problems in Banach spaces based on asymp- totic KKT conditions.SIAM Journal on Optimization, 30(4):2956–2982, 2020

    work page 2020

  9. [9]

    Brezis.Functional analysis, Sobolev spaces and partial differential equa- tions

    H. Brezis.Functional analysis, Sobolev spaces and partial differential equa- tions. Universitext. Springer, New York, 2011

    work page 2011

  10. [10]

    V. A. de Oliveira. Optimality conditions and duality for a class of continuous-time linear optimization problems.Comput. Appl. Math., 43(5):Paper No. 278, 29, 2024

    work page 2024

  11. [11]

    M. d. R. de Pinho, P. Loewen, and G. N. Silva. A weak maximum principle for optimal control problems with nonsmooth mixed constraints.Set-Valued and Variational Analysis, 17(2):203–221, 2009

    work page 2009

  12. [12]

    M. R. C. do Monte.Qualifica¸ c˜ oes de Restri¸ c˜ oes em Otimiza¸ c˜ ao N˜ ao Linear com Tempo Cont´ ınuo. PhD thesis, Universidade Estadual Paulista, S˜ ao Jos´ e do Rio Preto, S˜ ao Paulo, Brazil, 2018. In portuguese

    work page 2018

  13. [13]

    M. R. C. do Monte and V. A. de Oliveira. A full rank condition for continuous-time optimization problems with equality and inequality con- straints.TEMA (S˜ ao Carlos), 20(1):15–35, 2019

    work page 2019

  14. [14]

    M. R. C. do Monte and V. A. de Oliveira. Necessary conditions for continuous-time optimization under the Mangasarian–Fromovitz constraint qualification.Optimization, 69(4):777–798, 2020

    work page 2020

  15. [15]

    M. R. C. do Monte and V. A. de Oliveira. A constant rank constraint qualification in continuous-time nonlinear programming.Set-Valued and Variational Analysis, 29(1):61–81, 2021

    work page 2021

  16. [16]

    I. Ekeland. On the variational principle.Journal of Mathematical Analysis and Applications, 47(2):324–353, 1974

    work page 1974

  17. [17]

    W. H. Farr and M. A. Hanson. Continuous time programming with non- linear constraints.Journal of Mathematical Analysis and Applications, 45(1):96–115, 1974

    work page 1974

  18. [18]

    R. A. Gordon.The integrals of Lebesgue, Denjoy, Perron, and Henstock, volume 4. American Mathematical Society, Rhode Island, 1994

    work page 1994

  19. [19]

    R. C. Grinold. Continuous programming part one: linear objectives.Jour- nal of Mathematical Analysis and Applications, 28(1):32–51, 1969. 23

    work page 1969

  20. [20]

    Haeser and M

    G. Haeser and M. L. Schuverdt. On approximate KKT condition and its extension to continuous variational inequalities.Journal of Optimization Theory and Applications, 149(3):528–539, 2011

    work page 2011

  21. [21]

    E. S. Helou, S. A. Santos, and L. E. A. Simo˜ es. A new sequential optimal- ity condition for constrained nonsmooth optimization.SIAM Journal on Optimization, 30(2):1610–1637, 2020

    work page 2020

  22. [22]

    Jovi´ c and B

    A. Jovi´ c and B. Marinkovi´ c. New optimality criteria for convex continuous- time problems of vector optimization.Optimization, 71(15):4555–4570, 2022

    work page 2022

  23. [23]

    Kanzow, A

    C. Kanzow, A. B Raharja, and A. Schwartz. Sequential optimality condi- tions for cardinality-constrained optimization problems with applications. Computational Optimization and Applications, 80:185–211, 2021

    work page 2021

  24. [24]

    Kanzow, D

    C. Kanzow, D. Steck, and D. Wachsmuth. An augmented Lagrangian method for optimization problems in Banach spaces.SIAM J. Control Optim., 56(1):272–291, 2018

    work page 2018

  25. [25]

    Lauren¸ cot

    P. Lauren¸ cot. Weak compactness techniques and coagulation equations. In Evolutionary equations with applications in natural sciences, volume 2126 ofLecture Notes in Math., pages 199–253. Springer, Cham, 2015

    work page 2015

  26. [26]

    Levinson

    N. Levinson. A class of continuous linear programming problems.Journal of Mathematical Analysis and Applications, 16(1):73–83, 1966

    work page 1966

  27. [27]

    The MathWorks Inc., Natick, Massachusetts, 2022

    MATLAB.Version 9.12.0.2039608 (R2022a) Update 5. The MathWorks Inc., Natick, Massachusetts, 2022

    work page 2022

  28. [28]

    R. B. Moreira and V. A. de Oliveira. An asymptotic weak maximum prin- ciple.SIAM Journal on Control and Optimization, 62(5):2807–2833, 2024

    work page 2024

  29. [29]

    M. C. Pullan. An algorithm for a class of continuous linear programs.SIAM Journal on Control and Optimization, 31(6):1558–1577, 1993

    work page 1993

  30. [30]

    T. W. Reiland. Optimality conditions and duality in continuous program- ming I. convex programs and a theorem of the alternative.Journal of Mathematical Analysis and Applications, 77(1):297–325, 1980

    work page 1980

  31. [31]

    T. W. Reiland. Optimality conditions and duality in continuous program- ming II. the linear problem revisited.Journal of Mathematical Analysis and Applications, 77(2):329–343, 1980

    work page 1980

  32. [32]

    T. W. Reiland and M. A. Hanson. Generalized Kuhn-Tucker conditions and duality for continuous nonlinear programming problems.Journal of Mathematical Analysis and Applications, 74(2):578–598, 1980

    work page 1980

  33. [33]

    Vicanovi´ c and B

    J. Vicanovi´ c and B. Marinkovi´ c. Necessary optimality conditions for convex continuous-time optimization problems.J. Convex Anal., 30(1):5–16, 2023

    work page 2023

  34. [34]

    R. B. Vinter.Optimal control, volume 2. Springer, 2010

    work page 2010

  35. [35]

    G. Weiss. A simplex based algorithm to solve separated continuous linear programs.Mathematical Programming, 115(1):151–198, 2008. 24

    work page 2008

  36. [36]

    Wen, Y.-Y

    C.-F. Wen, Y.-Y. Lur, and H.-C. Lai. Approximate solutions and error bounds for a class of continuous-time linear programming problems.Opti- mization, 61(2):163–185, 2012

    work page 2012

  37. [37]

    H.-C. Wu. Continuous-time linear programming problems revisited: a per- turbation approach.Optimization, 62(1):33–70, 2013

    work page 2013

  38. [38]

    H.-C. Wu. Solving continuous-time linear programming problems based on the piecewise continuous functions.Numerical Functional Analysis and Optimization, 37(9):1168–1201, 2016

    work page 2016

  39. [39]

    G. J. Zalmai. Optimality conditions and Lagrangian duality in continuous- time nonlinear programming.Journal of mathematical analysis and appli- cations, 109(2):426–452, 1985

    work page 1985

  40. [40]

    G. J. Zalmai. Sufficient optimality conditions in continuous-time nonlin- ear programming.Journal of Mathematical Analysis and Applications, 111(1):130–147, 1985. 25

    work page 1985