pith. machine review for the scientific record. sign in

arxiv: 2604.12114 · v1 · submitted 2026-04-13 · 🧮 math.OC · cs.SY· eess.SY· math.PR· q-fin.MF

Recognition: unknown

A Decomposition Method for LQ Conditional McKean-Vlasov Control Problems with Random Coefficients

On\'esime Hounkpe , Dena Firoozi , Shuang Gao
Authors on Pith no claims yet

Pith reviewed 2026-05-10 15:01 UTC · model grok-4.3

classification 🧮 math.OC cs.SYeess.SYmath.PRq-fin.MF
keywords decomposition methodMcKean-Vlasov controllinear-quadraticconditional expectationsrandom coefficientsforward-backward SDEsstochastic optimal control
0
0 comments X

The pith

A decomposition reduces LQ conditional McKean-Vlasov control problems with random coefficients to two decoupled auxiliary problems whose optimal controls sum to the original optimum.

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

The paper develops a decomposition method for linear-quadratic McKean-Vlasov control problems that feature conditional expectations and random coefficients, with dynamics driven by two independent Wiener processes. It splits the original problem into two simpler stochastic optimal control problems that can be solved with classical techniques, one of them subject to a control constraint. Equivalence is shown between the well-posedness and solvability of the auxiliary problems and the original problem, and the sum of the auxiliary optimal controls recovers the original optimum. A variational argument then yields a characterization through two separate sets of linear forward-backward stochastic differential equations, and standard dynamic programming applies directly to the auxiliaries.

Core claim

The original conditional McKean-Vlasov control problem is equivalent in well-posedness and solvability to two decoupled auxiliary stochastic optimal control problems, one with a constrained admissible control set. The optimal control of the original problem is exactly the sum of the optimal controls obtained from the two auxiliaries. The solution is further characterized by two decoupled sets of non-McKean-Vlasov linear forward-backward stochastic differential equations, each associated with one auxiliary problem, and standard dynamic programming methods suffice to solve the auxiliaries.

What carries the argument

Decomposition of the conditional McKean-Vlasov dynamics into two independent components, each generating a separate linear-quadratic stochastic control problem that can be solved classically.

If this is right

  • The optimal control for the McKean-Vlasov problem equals the sum of the two auxiliary optimal controls.
  • Well-posedness of the original problem is equivalent to well-posedness of the two auxiliary problems.
  • The optimal solution is characterized by two independent sets of linear forward-backward SDEs without McKean-Vlasov terms.
  • Classical dynamic programming applied to each auxiliary problem yields the solution to the original problem.

Where Pith is reading between the lines

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

  • The method may reduce computational cost by allowing separate, smaller-dimensional solves whose results are simply added.
  • Similar decoupling ideas could be tested on non-LQ McKean-Vlasov problems if an analogous splitting of the conditional expectation can be found.
  • The constrained auxiliary problem suggests a way to incorporate state or control restrictions without altering the original formulation.

Load-bearing premise

The claimed equivalence and summation property hold only when the problem has linear-quadratic structure, dynamics driven by two independent Wiener processes, and one auxiliary problem carries an explicit control constraint.

What would settle it

A concrete counter-example in which the sum of the two auxiliary optimal controls fails to satisfy the original dynamics or to minimize the original cost would show the equivalence does not hold.

read the original abstract

We propose a decomposition method for solving a general class of linear-quadratic (LQ) McKean-Vlasov control problems involving conditional expectations and random coefficients, where the system dynamics are driven by two independent Wiener processes. Unlike existing approaches in the literature for these problems, such as the extended stochastic maximum principle and the extended dynamic programming methods, which often involve additional technical complexities and sometimes impose restrictive conditions on control inputs, our approach decomposes the original McKean-Vlasov control problem into two decoupled stochastic optimal control problems, one of which has a constrained admissible control set. These auxiliary problems can be solved using classical methods. We establish an equivalence between the well-posedness and solvability of the auxiliary problems and those of the original problem, and show that the sum of the optimal controls of the auxiliary problems yields the optimal control of the original problem. Moreover, by applying a variational method, we characterize the optimal solution to the McKean-Vlasov control problem via two decoupled sets of (non-McKean-Vlasov) linear forward-backward stochastic differential equations, each corresponding to one of the auxiliary problems. Finally, we show that standard dynamic programming can also be applied to solve the resulting auxiliary problems.

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 proposes a decomposition method for a class of linear-quadratic conditional McKean-Vlasov control problems with random coefficients driven by two independent Wiener processes. The original problem is reduced to two auxiliary LQ problems (one featuring a constrained admissible control set), with equivalence of well-posedness and solvability established; the sum of the auxiliary optima is shown to be optimal for the original problem. A variational argument yields a characterization via two decoupled linear (non-McKean-Vlasov) FBSDEs, and dynamic programming is noted as applicable to the auxiliaries.

Significance. If the claimed equivalence and decoupling hold, the method offers a concrete reduction of conditional McKean-Vlasov LQ problems to classical LQ problems solvable by standard FBSDE or Riccati techniques, bypassing some technical overhead of extended stochastic maximum principles. This is potentially useful for applications with common noise or partial information, and the paper correctly highlights the role of the two-Wiener-process structure in enabling clean separation.

major comments (2)
  1. [§3.2, Theorem 3.1] §3.2, Theorem 3.1: the equivalence between well-posedness of the original problem and the pair of auxiliary problems rests on the handling of the constrained admissible set in the second auxiliary problem; the proof sketch does not explicitly verify that the projection onto the constraint preserves the linear structure of the resulting FBSDE or that the variational inequality remains equivalent under random coefficients.
  2. [§4] §4, the variational characterization leading to the two decoupled FBSDEs: while the independence of the Wiener processes is invoked to eliminate cross terms, the argument does not address whether the conditional expectation in the original dynamics introduces additional martingale terms that survive the decomposition when coefficients are random and non-Markovian.
minor comments (2)
  1. [§2] The definition of the constrained control set in the second auxiliary problem (early in §2) should include an explicit statement of how the constraint set is chosen so that it does not inadvertently restrict the original feasible controls.
  2. Notation for conditional expectations and the two Wiener processes could be standardized across sections to avoid minor ambiguity when switching between the original and auxiliary formulations.

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, providing clarifications where needed and indicating planned revisions to the manuscript.

read point-by-point responses

  1. Referee: [§3.2, Theorem 3.1] §3.2, Theorem 3.1: the equivalence between well-posedness of the original problem and the pair of auxiliary problems rests on the handling of the constrained admissible set in the second auxiliary problem; the proof sketch does not explicitly verify that the projection onto the constraint preserves the linear structure of the resulting FBSDE or that the variational inequality remains equivalent under random coefficients.

    Authors: We thank the referee for highlighting this aspect of the proof. The admissible set for the second auxiliary problem is constructed so that the projection is a measurable operation compatible with the linear dynamics and quadratic cost; because the constraint is imposed pathwise on the control and the coefficients are random but adapted to the appropriate filtration, the projected process remains in the linear class and the resulting FBSDE retains its linear structure. The variational inequality equivalence follows directly from the same adaptation and the independence of the driving Wiener processes, which separates the conditional expectations without introducing cross terms. To address the referee's concern explicitly, we will expand the proof of Theorem 3.1 in the revised manuscript with an additional lemma that verifies preservation of linearity under projection and equivalence of the variational inequalities for random coefficients. revision: yes

  2. Referee: [§4] §4, the variational characterization leading to the two decoupled FBSDEs: while the independence of the Wiener processes is invoked to eliminate cross terms, the argument does not address whether the conditional expectation in the original dynamics introduces additional martingale terms that survive the decomposition when coefficients are random and non-Markovian.

    Authors: We appreciate the referee's observation on potential martingale terms. The conditional expectation appearing in the original McKean-Vlasov dynamics is taken with respect to the filtration generated by one of the two independent Wiener processes. Independence implies that the associated martingale increments are orthogonal to the control variations of the second auxiliary problem; consequently, after applying the tower property, no residual martingale terms survive in the decoupled variational equations. The random, non-Markovian coefficients are absorbed into the linear coefficients of the auxiliary FBSDEs and do not generate extra terms precisely because of this orthogonality. In the revised version we will insert a short clarifying paragraph (or remark) in Section 4 that makes this orthogonality argument explicit, using the independence of the Wiener processes and the tower property. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper decomposes the conditional McKean-Vlasov LQ problem into two auxiliary LQ problems (one constrained) whose optima sum to the original, with equivalence shown via variational characterization yielding two decoupled linear FBSDEs. This rests on standard FBSDE theory and classical LQ methods applied to the given dynamics with independent Wiener processes and random coefficients; the steps are self-contained and do not reduce by definition, fitted inputs, or self-citation chains to the target result itself. No self-definitional, ansatz-smuggling, or renaming patterns appear.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The approach rests on standard assumptions from stochastic analysis; no free parameters or new entities are introduced in the abstract.

axioms (2)
  • standard math The controlled SDEs driven by two independent Wiener processes admit unique strong solutions
    Required for the dynamics and for the equivalence statements to make sense.
  • domain assumption The linear-quadratic structure permits characterization of optimality via linear FBSDEs
    Central to the variational characterization of the auxiliary problems.

pith-pipeline@v0.9.0 · 5543 in / 1432 out tokens · 97135 ms · 2026-05-10T15:01:33.516569+00:00 · methodology

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

38 extracted references · 4 canonical work pages

  1. [1]

    Andersson, B

    D. Andersson, B. Djehiche, A maximum principle for SDEs of mean-field type, Appl Math Optim 63 (2011) 341–356

    2011

  2. [2]

    Buckdahn, B

    R. Buckdahn, B. Djehiche, J. Li, A general stochastic maximum principle for SDEs of mean-field-type, Appl Math Optim 64 (2011) 197–216

    2011

  3. [3]

    Yong, A linear-quadratic optimal control problem for mean-field stochastic differential equations, SIAM J

    J. Yong, A linear-quadratic optimal control problem for mean-field stochastic differential equations, SIAM J. Control Optim 51 (2013) 2809–2838

    2013

  4. [4]

    Carmona, F

    R. Carmona, F. Delarue, A. Lachapelle, Control of McKean–Vlasov dynamics versus mean field games, Math Finan Econ 7 (2013) 131–166

    2013

  5. [5]

    Laurière, O

    M. Laurière, O. Pironneau, Dynamic programming for mean-field type control, C. R.Acad.Sci.Paris,Ser.I 352 (2014) 707–713

    2014

  6. [6]

    H. Pham, X. Wei, Dynamic programming for optimal control of stochastic McKean–Vlasov dynamics, SIAM J. Control Optim 55 (2017) 2809–2838

    2017

  7. [7]

    H. Pham, Linear quadratic optimal control of conditional McKean-Vlasov equation with random coefficients and applications, Probability, Uncertainty and Quantitative Risk 1 (2016) 2809–2838

    2016

  8. [8]

    Acciaio, J

    B. Acciaio, J. Backhoff-Veraguas, R. Carmona, Extended mean field control problems: Stochastic maximum principle and transport perspective, SIAM J. Control Optim 57 (2019) 3666–3693

    2019

  9. [9]

    Bensoussan, J

    A. Bensoussan, J. Kim, S. C. P. Yam, Extended mean field type control theory: A master equation approach with some applications, Journal of Optimization Theory and Applications 207 (2025)

    2025

  10. [10]

    Carmona, L

    R. Carmona, L. Tangpi, K. Zhang, Conditional McKean-Vlasov control,https://arxiv.org/abs/2510. 06543v1, 2025

    2025

  11. [11]

    Optimal control of McKean-Vlasov systems under partial observation and hidden Markov switching.arXiv:2601.09311, 2026

    M. Fuhrman, H. Pham, S. Rudà, Optimal control of McKean-Vlasov systems under partial observation and hidden markov switching,https://arxiv.org/abs/2601.09311v1, 2026

    work page arXiv 2026

  12. [12]

    P. J. Graber, Linear quadratic mean field type control and mean field games with common noise, with application to production of an exhaustible resource, Appl Math Optim 74 (2016) 459–486

    2016

  13. [13]

    Peng, A general stochastic maximum principle for optimal control problems, SIAM J

    S. Peng, A general stochastic maximum principle for optimal control problems, SIAM J. Control Optim 28 (1990) 966–979. 14

    1990

  14. [14]

    J. Yong, X. Y. Zhou, Stochastic controls: Hamiltonian systems and HJB equations, volume 43, Springer Science & Business Media, 2012

    2012

  15. [15]

    Pham, Continuous-time Stochastic Control and Optimization with Financial Applications, volume 61, Springer Berlin, Heidelberg, 2009.https://doi.org/10.1007/978-3-540-89500-8

    H. Pham, Continuous-time Stochastic Control and Optimization with Financial Applications, volume 61, Springer Berlin, Heidelberg, 2009.https://doi.org/10.1007/978-3-540-89500-8

    work page doi:10.1007/978-3-540-89500-8 2009

  16. [16]

    Carmona, F

    R. Carmona, F. Delarue, Forward-backward stochastic differential equations and controlled McKean-Vlasov dynamics, The Annals of Probability 43 (2015) 2647–2700

    2015

  17. [17]

    M. Tang, Q. Meng, Linear-quadratic optimal control problems for mean-field stochastic differential equations with jumps, Asian J. Control 21 (2019) 809–823

    2019

  18. [18]

    Bensoussan, J

    A. Bensoussan, J. Frehse, S. C. P. Yam, The master equation in mean field theory, J. Math. Pures Appl 103 (2015) 1441–1474

    2015

  19. [19]

    H. Pham, X. Wei, Bellman equation and viscosity solutions for mean-field stochastic control problem, ESAIM: COCV 24 (2018) 437–461

    2018

  20. [20]

    Sun, Mean-field stochastic linear quadratic optimal control problems: Open-loop solvabilities, ESAIM: COCV 23 (2017) 1099–1127

    J. Sun, Mean-field stochastic linear quadratic optimal control problems: Open-loop solvabilities, ESAIM: COCV 23 (2017) 1099–1127

    2017

  21. [21]

    Y. Yang, M. Tang, Q. Meng, A mean-field stochastic linear-quadratic optimal control problem with jumps under partial information, ESAIM: COCV 28 (2022)

    2022

  22. [22]

    Xiong, W

    J. Xiong, W. Xu, Mean-field stochastic linear quadratic control problem with random coefficients, SIAM J. Control Optim 63 (2025) 3042–3060

    2025

  23. [23]

    Arabneydi, A

    J. Arabneydi, A. Mahajan, Team-optimal solution of finite number of mean-field coupled LQG subsystems, 2015 54th IEEE Conference on Decision and Control (CDC) (2015) 5308–5313

    2015

  24. [24]

    Huang, Large-population LQG games involving a major player: the Nash certainty equivalence principle, SIAM J

    M. Huang, Large-population LQG games involving a major player: the Nash certainty equivalence principle, SIAM J. Control Optim 48 (2010) 3318–3353

    2010

  25. [25]

    Firoozi, S

    D. Firoozi, S. Jaimungal, P. E. Caines, Convex analysis for LQG systems with applications to major–minor LQG mean–field game systems, Systems & Control Letters 142 (2020) 104734

    2020

  26. [26]

    Carmona, X

    R. Carmona, X. Zhu, A probabilistic approach to mean field games with major and minor players, The Annals of Applied Probability 26 (2016) 1535–1580

    2016

  27. [27]

    Nourian, P

    M. Nourian, P. E. Caines,ϵ-Nash mean field game theory for nonlinear stochastic dynamical systems with major and minor agents, SIAM J. Control Optim 51 (2013) 3302–3331

    2013

  28. [28]

    Firoozi, LQG mean field games with a major agent: Nash certainty equivalence versus probabilistic approach, Automatica 146 (2022) 110559

    D. Firoozi, LQG mean field games with a major agent: Nash certainty equivalence versus probabilistic approach, Automatica 146 (2022) 110559

    2022

  29. [29]

    Huang, Linear-quadratic mean field games with a major player: Nash certainty equivalence versus master equations, Communications in Information and Systems, vol

    M. Huang, Linear-quadratic mean field games with a major player: Nash certainty equivalence versus master equations, Communications in Information and Systems, vol. 21, no. 3 (2021)

    2021

  30. [30]

    Carmona, P

    R. Carmona, P. Wang, An alternative approach to mean field game with major and minor players, and applications to herders impacts, Appl Math Optim 76 (2017) 5–27

    2017

  31. [31]

    J. M. Bismut, Linear quadratic optimal stochastic control with random coefficients, SIAM J. Control Optim 14 (1976) 419–444

    1976

  32. [32]

    Tang, General linear quadratic optimal stochastic control problems with random coefficients: Linear stochastic Hamilton systems and backward stochastic Riccati equations, SIAM J

    S. Tang, General linear quadratic optimal stochastic control problems with random coefficients: Linear stochastic Hamilton systems and backward stochastic Riccati equations, SIAM J. Control Optim 42 (2003) 53–75

    2003

  33. [33]

    J. Sun, J. Xiong, J. Yong, Indefinite stochastic linear-quadratic optimal control problems with random coefficients: Closed-loop representation of open-loop optimal controls, The Annals of Applied Probability 31 (2021) 460–499. 15

    2021

  34. [34]

    J. Ma, P. Protter, J. Yong, Solving forward–backward stochastic differential equations explicitly—a four step scheme, Probability Theory and Related Fields 98 (1994) 339–359. doi:10.1007/BF01192258

    work page doi:10.1007/bf01192258 1994

  35. [35]

    Peng, Stochastic Hamilton-Jacobi-Bellman equations, SIAM J

    S. Peng, Stochastic Hamilton-Jacobi-Bellman equations, SIAM J. Control Optim 30 (1992) 284–304

    1992

  36. [36]

    Qiu, Viscosity solution of stochastic Hamilton-Jacobi-Bellman equations, SIAM J

    J. Qiu, Viscosity solution of stochastic Hamilton-Jacobi-Bellman equations, SIAM J. Control Optim 56 (2018) 3708–3730

    2018

  37. [37]

    Moon, Stochastic optimal control with random coefficients and associated stochastic Hamil- ton–Jacobi–Bellman equations, Adv Cont Discr Mod 2022 (2022)

    J. Moon, Stochastic optimal control with random coefficients and associated stochastic Hamil- ton–Jacobi–Bellman equations, Adv Cont Discr Mod 2022 (2022). https://doi.org/10.1186/ s13662-021-03674-5

    2022

  38. [38]

    Tang, Dynamic programming for general linear quadratic optimal stochastic control with random coefficients, SIAM Journal on Control and Optimization 53 (2015) 1082–1106

    S. Tang, Dynamic programming for general linear quadratic optimal stochastic control with random coefficients, SIAM Journal on Control and Optimization 53 (2015) 1082–1106. doi:10.1137/130947656. Appendix A. Proof of Lemma 1 We prove properties (iv) and (v) in the following sections. Appendix A.1. Proof of Property (iv) We aim to show thatE[(xt −H¯xt)⊺ St...

    work page doi:10.1137/130947656 2015