pith. sign in

arxiv: 2501.12731 · v3 · submitted 2025-01-22 · 🧮 math.OC · math.PR

Constrained mean-field control with singular controls: Existence, stochastic maximum principle and constrained FBSDE

Lijun Bo , Jingfei Wang , Xiang Yu This is my paper

Pith reviewed 2026-05-23 05:00 UTC · model grok-4.3

classification 🧮 math.OC math.PR
keywords mean-field controlsingular controlsstochastic maximum principleconstrained FBSDErelaxed controlsMcKean-Vlasov processesLagrange multipliersdynamic constraints
0
0 comments X

The pith

A relaxed control formulation on canonical spaces establishes existence of optimal singular controls for mean-field problems under dynamic mixed constraints and yields a stochastic maximum principle via Lagrange multipliers on Banach space.

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

The paper studies mean-field control problems that incorporate singular controls together with general dynamic constraints linking states and control laws. It introduces a customized relaxed control formulation that accommodates singular controls while preserving the McKean-Vlasov interaction structure, then proves existence of an optimal control pair by compactification arguments in appropriate canonical spaces. The controlled process is next recast as an infinite-dimensional equality constraint, converting the original problem into a constrained optimization over Banach space; a new Lagrange-multiplier technique then produces the stochastic maximum principle and an associated class of constrained backward stochastic differential equations. The authors further establish uniqueness and stability for the solutions of this constrained forward-backward system. A reader would care because singular controls appear in applications involving cumulative or impulse actions, and the mean-field setting captures interactions among many agents.

Core claim

By treating the controlled McKean-Vlasov process as an infinite-dimensional equality constraint on a Banach space, the mean-field control problem with singular controls and dynamic mixed constraints can be recast as an optimization problem on canonical spaces; a Lagrange-multiplier method then yields the stochastic maximum principle together with a class of constrained BSDEs whose solutions are unique and stable.

What carries the argument

The customized relaxed control formulation that converts the dynamic mixed constraints into an infinite-dimensional equality constraint on Banach space, enabling the Lagrange-multiplier derivation of the stochastic maximum principle.

If this is right

  • Existence of an optimal pair of regular and singular controls is guaranteed under the relaxed formulation and compactification.
  • The optimal controls satisfy a stochastic maximum principle obtained from the Lagrange-multiplier method.
  • The associated constrained FBSDE admits unique and stable solutions.
  • The approach applies to general dynamic state-control-law constraints without requiring specific convexity assumptions.

Where Pith is reading between the lines

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

  • The constrained BSDE may serve as a starting point for numerical schemes that approximate optimal singular controls in high-dimensional mean-field settings.
  • Stability of the FBSDE solutions suggests that small changes in the constraint set produce correspondingly small changes in the optimal controls.
  • The Banach-space constraint technique could be adapted to mean-field games with singular controls by adding equilibrium conditions.

Load-bearing premise

The dynamic mixed constraints admit a relaxed control formulation that preserves the original McKean-Vlasov structure and does not lose any solutions of the unrelaxed problem.

What would settle it

An explicit example of a constrained mean-field problem with singular controls for which the proposed relaxed formulation either fails to preserve the McKean-Vlasov structure or admits no optimal control despite the compactification argument.

read the original abstract

This paper studies a class of mean-field control (MFC) problems with singular controls under general dynamic state-control-law constraints. We first propose a customized relaxed control formulation to cope with the dynamic mixed constraints and establish the existence of an optimal control using compactification argument in the proper canonical spaces to accommodate singular controls. To further characterize the optimal pair of regular and singular controls, we treat the controlled McKean-Vlasov process as an infinite-dimensional equality constraint and recast the MFC problem as an optimization problem on canonical spaces with constraints on Banach space, allowing us to derive the stochastic maximum principle (SMP) and a class of constrained BSDE using a new Lagrange multipliers method. Additionally, we investigate the uniqueness and the stability result of the solution to the constrained FBSDE associated with the constrained MFC with singular controls.

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 / 0 minor

Summary. The paper studies constrained mean-field control problems with singular controls under general dynamic state-control-law constraints. It proposes a customized relaxed control formulation to handle the dynamic mixed constraints, establishes existence of an optimal control via compactification in canonical spaces, recasts the controlled McKean-Vlasov process as an infinite-dimensional equality constraint on a Banach space to derive the stochastic maximum principle and constrained BSDEs via a new Lagrange multipliers method, and investigates uniqueness and stability of the associated constrained FBSDE.

Significance. If the constructions and derivations hold, the results would extend mean-field control theory to singular controls and general dynamic constraints by adapting relaxed-control compactification and Lagrange-multiplier techniques to the McKean-Vlasov setting. This could provide a systematic way to obtain necessary optimality conditions in the form of SMP and constrained FBSDEs, which is relevant for applications involving mean-field interactions with state and control constraints.

major comments (2)
  1. [Abstract] Abstract: The existence result is asserted via a 'customized relaxed control formulation' and 'compactification argument in the proper canonical spaces,' but no list of assumptions on the coefficients, cost, or constraints is provided, nor is any verification that the relaxed formulation preserves the original McKean-Vlasov structure and does not lose solutions; these are load-bearing for the existence claim.
  2. [Abstract] Abstract: The central step of recasting the controlled process as an 'infinite-dimensional equality constraint on Banach space' to enable the Lagrange multipliers method for the SMP is described at a high level only, with no definition of the Banach space, the precise form of the constraint, or confirmation that the original problem's solutions are retained; this step underpins the derivation of the SMP and constrained BSDE.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and the constructive comments on the abstract. The two points raised concern the level of detail provided in the abstract regarding assumptions and technical constructions. We address each below and indicate where revisions will be made.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The existence result is asserted via a 'customized relaxed control formulation' and 'compactification argument in the proper canonical spaces,' but no list of assumptions on the coefficients, cost, or constraints is provided, nor is any verification that the relaxed formulation preserves the original McKean-Vlasov structure and does not lose solutions; these are load-bearing for the existence claim.

    Authors: The assumptions on the coefficients, cost functional, and dynamic constraints are stated explicitly as (A1)–(A4) in Section 2. The preservation of the McKean-Vlasov structure under the customized relaxed formulation, together with the fact that no original solutions are lost, is established in Proposition 3.1 and the compactness arguments of Section 3. While the abstract is intentionally concise, we agree that a brief reference to the standing assumptions would improve clarity. We will revise the abstract to include one sentence indicating that the results hold under the standard Lipschitz and growth conditions listed in Section 2. revision: yes

  2. Referee: [Abstract] Abstract: The central step of recasting the controlled process as an 'infinite-dimensional equality constraint on Banach space' to enable the Lagrange multipliers method for the SMP is described at a high level only, with no definition of the Banach space, the precise form of the constraint, or confirmation that the original problem's solutions are retained; this step underpins the derivation of the SMP and constrained BSDE.

    Authors: The Banach space is introduced in Section 4 as the space of continuous paths equipped with the supremum norm; the infinite-dimensional equality constraint is the integral form of the controlled McKean-Vlasov dynamics written as an element of that space. Retention of original solutions follows from the equivalence between the original and relaxed problems shown in Section 3. The Lagrange-multiplier argument yielding the SMP and constrained BSDE appears in Theorem 4.2. We will add a short clause to the abstract directing the reader to Section 4 for the precise functional-analytic setting. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The derivation chain applies standard compactification in canonical spaces for existence and a Lagrange-multiplier treatment of an infinite-dimensional Banach-space constraint to obtain the SMP and constrained FBSDE. These steps are presented as direct applications of established techniques to the stated relaxed-control formulation and McKean-Vlasov dynamics; no equation is shown to equal its own input by definition, no fitted parameter is relabeled as a prediction, and no load-bearing premise rests on a self-citation chain. The uniqueness/stability analysis of the FBSDE is likewise treated as an independent investigation rather than a renaming or self-referential closure. The abstract and reader summary supply no concrete reduction that would trigger any of the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Paper rests on standard domain assumptions of stochastic control theory for McKean-Vlasov processes and relaxed controls; no free parameters or invented entities are introduced in the abstract.

axioms (2)
  • domain assumption Dynamic mixed constraints can be handled via a customized relaxed control formulation that preserves the original problem structure.
    Invoked to establish existence via compactification.
  • domain assumption The controlled McKean-Vlasov process can be treated as an infinite-dimensional equality constraint on a Banach space.
    Used to recast the MFC problem for application of Lagrange multipliers.

pith-pipeline@v0.9.0 · 5671 in / 1267 out tokens · 55072 ms · 2026-05-23T05:00:44.880300+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

33 extracted references · 33 canonical work pages · 2 internal anchors

  1. [1]

    Acciaio, J

    B. Acciaio, J. Backhoff-Veraguas, and R. Carmona (2017): E xtended mean field control problems: stochastic maximum principle and transport pers pective. SIAM J. Contr. Optim. 57(6): 3666-3693

    work page 2017

  2. [2]

    Andersson, and B

    D. Andersson, and B. Djehiche (2010): A maximum principl e for SDEs of mean-field type. Appl. Math. Optim. 63: 341-356

    work page 2010

  3. [3]

    Bahlali, and B

    S. Bahlali, and B. Mezerdi (2005): A general stochastic m aximum principle for singular control problems. Electron. J. Probab. 10: 988-1004

    work page 2005

  4. [4]

    Bayraktar, A

    E. Bayraktar, A. Cosso, and H. Pham (2018): Randomized dy namic programming principle and Feynman-Kac representation for optimal control of McKe an-Vlasov dynamics. Trans. AMS 370(3): 2115-60

    work page 2018

  5. [5]

    L. Bo, J. Wang, X. Wei, and X. Yu (2024): Extended mean-fiel d control problems with Poissonian common noise: Stochastic maximum principle and Hamiltonian-Jacobi-Bellman equation. Preprint avaiable at https://arxiv.org/abs/2407.05356

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  6. [6]

    L. Bo, J. Wang, and X. Yu (2024): Extended mean field contro l problems with con- straints: The generalized Fritz-John conditions and Lagra ngian method. Preprint available at: https://arxiv.org/abs/2408.06865

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  7. [7]

    Buckdahn, B

    R. Buckdahn, B. Djehiche, and J. Li (2011): A general maxi mum principle for SDEs of mean-field type. Appl. Math. Optim. 64(2): 197-216

    work page 2011

  8. [8]

    Campi, T

    L. Campi, T. De Angelis, M. Ghio, and G. Livieri (2022): Me an-field games of finite-fuel capacity expansion with singular controls. Ann. Appl. Probab. 32(5): 3674-3717

    work page 2022

  9. [9]

    H. Cao, J. Dianetti and G. Ferrari (2023): Stationary dis counted and ergodic mean field games with singular controls. Math. Oper. Res. 48(4):1871-1898

    work page 2023

  10. [10]

    Carmona, and F

    R. Carmona, and F. Delarue (2015): Forward-backward st ochastic differential equations and controlled McKean-Vlasov dynamics. Ann. Probab. 43(5): 2647-2700

    work page 2015

  11. [11]

    Chiarolla, G

    M. Chiarolla, G. Ferrari, and F. Riedel (2013): General ized Kuhn-Tucker conditions for N-firm stochastic irreversible investment under limited re sources. SIAM J. Contr. Optim. 51(5): 3863-3885

    work page 2013

  12. [12]

    Daudin (2023a): Mean-field limit for stochastic cont rol problems under state constraint

    S. Daudin (2023a): Mean-field limit for stochastic cont rol problems under state constraint. Preprint available at: https://arxiv.org/abs/2306.00949

    work page arXiv

  13. [13]

    Daudin (2023b): Optimal control of the Fokker-Planc k equation under state constraints in the Wasserstein space

    S. Daudin (2023b): Optimal control of the Fokker-Planc k equation under state constraints in the Wasserstein space. J. Math. Pures Appl. 175: 37-75

    work page

  14. [14]

    Denkert, and U

    R. Denkert, and U. Horst (2023): Extended mean-field con trol problems with multi- dimensional singular controls. Preprint available at https://arxiv.org/abs/2308.04378

    work page arXiv 2023

  15. [15]

    Djete, D

    M.F. Djete, D. Possama ¨ ı, and X. Tan (2022): McKean-Vlasov optimal control: limit theory and equivalence between different formulations. Math. Oper. Res. 47(4): 2891-930

    work page 2022

  16. [16]

    El Karoui, and S

    N. El Karoui, and S. M´ el´ eard (1990): Martingale measures and stochastic calculus. Probab. Theor. Rel. Fields 84: 83-101

    work page 1990

  17. [17]

    Fu, and U

    G. Fu, and U. Horst (2017): Mean field games with singular controls. SIAM J. Contr. Optim. 55(6): 3833-3868. 23

    work page 2017

  18. [18]

    Germain, H

    M. Germain, H. Pham, and X. Warin (2023): A level-set app roach to the control of state- constrained McKean-Vlasov equations: Applications to ren ewable energy storage and port- folio selection. Numer. Algebra Contr. Optim. 13: 555-582

    work page 2023

  19. [19]

    Guo, and P

    X. Guo, and P. Tomecek (2009): A class of singular contro l problems and the smooth fit principle. SIAM J. Contr. Optim. 47(6): 3076-3099

    work page 2009

  20. [20]

    Haussmann, and W

    U.G. Haussmann, and W. Suo (1995a): Singular optimal st ochastic controls I: Existence. SIAM J. Contr. Optim. 33: 916-936

    work page

  21. [21]

    Haussmann, and W

    U.G. Haussmann, and W. Suo (1995b): Singular optimal st ochastic controls II: Dynamic programming. SIAM J. Contr. Optim. 33(3): 937-959

    work page

  22. [22]

    Huang, R.P

    M. Huang, R.P. Malham´ e, and P.E. Caines (2006): Large p opulation stochastic dynamic games closed-loop McKean-Vlasov systems and the Nash certa inty equivalence principle. Commun. Inform. Syst. 6(3): 221-252

    work page 2006

  23. [23]

    Jacod, and J

    J. Jacod, and J. M´ emin (1981): Sur un type de convergence inter´ ediaire entre la convergence en loi et la convergence en probabilit´ e, S´ eminaire de probabilit´ es de Strasbourg15: 529-546

    work page 1981

  24. [24]

    Lacker (2015): Mean field games via controlled martin gale problems: Existence of Markovian equilibria

    D. Lacker (2015): Mean field games via controlled martin gale problems: Existence of Markovian equilibria. Stoch. Process. Appl. 125: 2856-2894

    work page 2015

  25. [25]

    Lacker (2017): Limit theory for controlled McKean-V lasov dynamics

    D. Lacker (2017): Limit theory for controlled McKean-V lasov dynamics. SIAM J. Contr. Optim. 55(3): 1641-1672

    work page 2017

  26. [26]

    Lasry, and P.L

    J.M. Lasry, and P.L. Lions (2007): Mean field games. Jpn. J. Math. 2(1): 229-260

    work page 2007

  27. [27]

    Lauri´ ere, and O

    M. Lauri´ ere, and O. Pironneau (2014): Dynamic program ming for mean-field type control. Comptes. Rendus. Math. 352(9): 707-713

    work page 2014

  28. [28]

    Leese (1978): Measurable selections and the unifo rmization of Souslin sets

    S.J. Leese (1978): Measurable selections and the unifo rmization of Souslin sets. Amer. J. Math. 100: 19-41

    work page 1978

  29. [29]

    Ma (1992): On the principle of smooth fit for a class of s ingular stochastic control problems for diffusions

    J. Ma (1992): On the principle of smooth fit for a class of s ingular stochastic control problems for diffusions. SIAM J. Contr. Optim. 30(4): 975-999

    work page 1992

  30. [30]

    Pham, and X

    H. Pham, and X. Wei (2017): Dynamic programming for opti mal control of stochastic McKean-Vlasov dynamics. SIAM J. Contr. Optim. 55(2): 1069-1101

    work page 2017

  31. [31]

    Wu, and J

    C. Wu, and J. Zhang (2020): Viscosity solutions to parab olic master equations and McKean-Vlasov SDEs with closed-loop controls. Ann. Appl. Probab. 30(2): 936-986

    work page 2020

  32. [32]

    J. Zhou, N. Touzi, and J. Zhang (2024): Viscosity soluti ons for HJB equations on the process space: Application to mean field control with common noise. Preprint available at: https://arxiv.org/abs/2401.04920

    work page arXiv 2024

  33. [33]

    Zowe, and S

    J. Zowe, and S. Kurcyusz (1979): Regularity and stabili ty for the mathematical program- ming problem in Banach spaces. Appl. Math. Optim. 5: 49-62. 24

    work page 1979