--- title: Common Noise McKean–Vlasov Problems url: https://www.emergentmind.com/topics/common-noise-mckean-vlasov-martingale-problems type: topic --- # Common Noise McKean–Vlasov Problems Common-noise McKean–Vlasov martingale problems are weak formulations of mean-field stochastic dynamics in which the coefficients depend on a **conditional law** given a shared source of randomness, typically a common Brownian motion. In contrast with the no-common-noise case, the law process is itself random, and the state equation, the associated measure-valued evolution, and the martingale characterization must all be formulated relative to the common-noise filtration. Across stochastic control, mean field games, stochastic differential games, and singular interacting systems, this framework appears through conditional McKean–Vlasov SDEs, stochastic Fokker–Planck equations, generators on spaces of probability measures, and verification principles expressed as martingale or submartingale conditions [1908.00955]. ## 1. Conditional-law formulation and basic objects A prototypical common-noise McKean–Vlasov SDE has the form \[ dX_t = b(X_t,\mu_t)\,dt + \sigma(X_t)\,dB_t + \sigma_0\,dW_t, \qquad \mu_t := \mathcal{L}(X_t\mid\mathcal{F}^W_t), \] where \(B\) is idiosyncratic noise and \(W\) is common noise. In this formulation, the interaction is through the **conditional law** rather than the unconditional distribution, so \(\mu_t\) is a random probability measure flow adapted to the common-noise filtration [2401.07665]. A more path-dependent version allows \[ X_t = \xi +\int_0^t b(s,X_{\cdot\wedge s},\mu_s)\, ds +\int_0^t \sigma(s,X_{\cdot\wedge s},\mu_s)\, dW_s +\int_0^t \rho(s,X_{\cdot\wedge s},\mu_s)\,dB_s, \quad \mu_s = \mathscr L\big(X_{\cdot\wedge s}\mid \mathcal{F}_s^{B,\mu}\big), \] which makes explicit that the mean-field argument may be a conditional law on path space [1908.00955]. The common-noise setting is equivalently described at the particle level. A standard \(N\)-particle system is \[ dX_t^i = b(t, X_t^i,\mu_t^N,u_t^i)\,dt + \sigma_{\text{ind}}(X_t^i)\,dB_t^i + \sigma_{\text{com}}(X_t^i)\,dW_t, \qquad \mu_t^N = \frac{1}{N}\sum_{j=1}^{N}\delta_{X_t^j}, \] and the large-\(N\) limit yields a random measure-valued process \(\mu_t\) rather than a deterministic flow, precisely because the empirical measure remains adapted to the common noise [1506.04594]. The same structural feature appears in controlled settings, where the coefficients may depend on the conditional law of the state and even on the conditional law of the state-control pair given the common-noise filtration [1907.08860]. A common misconception is that common noise merely replaces deterministic measure flows by random ones without altering the analytic structure. The literature shows that this is too narrow. Common noise changes the filtration, the notion of law consistency, the form of the generator, and the correct state space for dynamic programming and martingale problems; in several settings it also changes the long-time qualitative behavior, sometimes stabilizing the dynamics and sometimes inducing singular events such as blow-ups [2306.16130]. ## 2. Generator, SPDE, and martingale characterizations on measure space For a fixed measure \(\mu\), the single-particle generator is typically \[ (\mathcal{L}_\mu f)(x) = b(x,\mu) f'(x) + \tfrac12 \bigl(\sigma^2(x)+\sigma_0^2\bigr) f''(x), \] or, in the multidimensional diffusion setting with common and idiosyncratic noises, \[ \mathcal{A}_m\varphi(x) = -\big(\nabla V(x)+\nabla W*m(x)\big)\cdot\nabla\varphi(x) +\tfrac{\sigma^2+\sigma_0^2}{2}\Delta\varphi(x). \] The corresponding martingale formulation requires that for each smooth test function, \[ M_t^\varphi = \varphi(X_t)-\varphi(X_0)-\int_0^t \mathcal{A}_{\mu_s}\varphi(X_s)\,ds \] is a martingale, with \(\mu_s=\mathcal{L}(X_s\mid\mathcal{F}_s^0)\) or its path-space analogue [2401.07665; 2306.16130]. At the law level, the conditional measure flow solves a stochastic nonlinear Fokker–Planck equation. In weak form, one representative formulation is \[ d(\varphi,\mu_t) = \big(L[t,\mu_t,u_t]\varphi,\mu_t\big)\,dt + \big(\sigma_{\text{com}}(\cdot)\varphi',\mu_t\big)\,dW_t, \] with \[ L[t,\mu,u]\varphi(x) = \tfrac12\big(\sigma_{\text{ind}}^2(x)+\sigma_{\text{com}}^2(x)\big)\varphi''(x) +b(t,x,\mu,u(x))\varphi'(x), \] so the law process is itself a measure-valued diffusion driven by the common Brownian motion [1506.04594]. In density form, the same object becomes a stochastic Fokker–Planck SPDE for \(\rho_t\), and this equivalence is one of the standard bridges between SDE and martingale formulations [1908.00955]. A more intrinsic formulation acts directly on functionals \(F\) of measures. For smooth enough \(F\), the limiting generator in the presence of common noise is \[ \mathcal A_{\text{lim}}F(\mu) = \int_{\mathbb R} B_\mu\!\left(\frac{\delta F(\mu)}{\delta\mu(y)}\right)\mu(dy) + \int_{\mathbb R^2} \sigma_{\text{com}}(y)\sigma_{\text{com}}(z)\, \frac{\delta^2 F(\mu)}{\delta\mu(y)\delta\mu(z)}\,\mu(dy)\mu(dz), \] and the martingale condition is \[ M_t^F := F(\mu_t)-F(\mu_0)-\int_0^t \mathcal A_{\text{lim}}F(\mu_s)\,ds \] is a martingale [1506.04594]. The second-order term in measure space is the signature of common noise: it has no analogue in deterministic law flows. This generator-level viewpoint is central in long-time analysis as well. For nonlinear stochastic Fokker–Planck equations with common noise, invariant measures on \(\mathcal{P}(\mathcal{P}(\mathbb{R}^d))\) are characterized by \[ \langle \widebar P, \mathcal{M}F\rangle = 0, \qquad \forall F\in\mathcal{C}^2_b(\mathcal{P}(\mathbb{R}^d)), \] where \(\mathcal{M}\) is the measure-space generator containing both drift terms and the common-noise-induced second-order term in the measure variable [2306.16130]. ## 3. Weak solutions, compatibility, and controlled martingale problems In weak formulations with common noise, the dependence structure between \(X\), the common noise, and the conditional law process cannot be left implicit. A central notion is **compatibility**: \(X\) must be compatible with \((B,\mu)\), and \((X,\mu)\) must be compatible with \((B,W,\xi)\), so that the conditional law constraint \[ \mu_t = \mathscr L\bigl(X_{\cdot\wedge t}\mid\mathcal F_t^{B,\mu}\bigr) \] remains meaningful and stable under weak convergence [1908.00955]. This avoids pathological weak limits in which the process labelled as “conditional law” fails to be the actual conditional distribution of the state. Under boundedness and joint continuity of coefficients, weak existence can be proved by compactness of Euler approximations while keeping track of the full dependence structure of \((X,\mu,B,W)\); weak uniqueness can then be obtained under non-degeneracy of the idiosyncratic diffusion and drift regularity in total variation via a Monge–Kantorovich cost function and Girsanov representation [1908.00955]. In this sense, the common-noise martingale problem is well posed as a genuinely nonlinear weak problem with random coefficients. Controlled problems admit an analogous reformulation on canonical spaces with two filtrations. In the weak formulation of McKean–Vlasov optimal control with common noise, one works with a global filtration \(\mathbb F^\gamma\) and a common-noise filtration \(\mathbb G^\gamma\), and sets \[ \mu_s^\gamma = \mathcal{L}^{\mathbb P^\gamma}(X_{s\wedge\cdot}^\gamma \mid \mathcal G_s^\gamma), \qquad \bar\mu_s^\gamma = \mathcal{L}^{\mathbb P^\gamma}((X_{s\wedge\cdot}^\gamma,\alpha_s^\gamma)\mid \mathcal G_s^\gamma). \] The controlled state satisfies \[ dX_t^\alpha = b\bigl(t,X_{t\wedge\cdot}^\alpha,\mathcal{L}(X_{t\wedge\cdot}^\alpha,\alpha_t\mid\mathcal G_t),\alpha_t\bigr)\,dt + \sigma(\cdots)\,dW_t + \sigma_0(\cdots)\,dB_t, \] and the canonical martingale problem is expressed in terms of a lifted generator acting on \((X,W,B)\) together with an explicit conditional-law process \(\hat\mu\) on path space [1907.08860]. This two-filtration framework is technically significant because conditioning, concatenation, and measurable selection must all be performed relative to the common-noise filtration. The dynamic programming principle then takes the form \[ V_W(t,\nu) = \sup_{\gamma\in\Gamma_W(t,\nu)} \mathbb{E}^{\mathbb{P}^\gamma}\Big[ \int_t^{\tau^\gamma} L(\cdots)\,ds + V_W\big(\tau^\gamma,\mu_{\tau^\gamma}^\gamma\big) \Big], \] with the state variable at the stopping time given by the updated conditional law \(\mu_{\tau^\gamma}^\gamma\) [1907.08860]. A plausible implication is that, in common-noise settings, the measure-valued component is not merely a parameter in the coefficients but an essential part of the Markov state. ## 4. Weak martingale optimality in linear-quadratic and game-theoretic models Linear-quadratic common-noise McKean–Vlasov models provide a particularly explicit realization of the martingale-problem viewpoint. In the stochastic control setting, the state dynamics are linear in the state, its mean, the control, and its mean, and the cost depends quadratically on both deviations from the mean and the mean itself. In the common-noise extension, unconditional expectations are replaced by conditional expectations given the common-noise filtration, for example \[ \bar X_t^0 := \mathbb{E}[X_t\mid W^0], \qquad \bar{\boldsymbol{\alpha}}_t^0 := \mathbb{E}[\boldsymbol{\alpha}_t\mid W^0], \] and the coefficients may be \(\mathbb F^0\)-adapted [1812.00632]. The defining method is a **weak martingale optimality principle**. For a candidate control profile, one constructs value-like processes \(W_t^{\boldsymbol{\alpha},i}\) and then \[ S_t^{\boldsymbol{\alpha},i} := e^{-\rho t} W_t^{\boldsymbol{\alpha},i} + \int_0^t e^{-\rho u} f^i\big(u,X_u^{\boldsymbol{\alpha}},\bar X_u^{\boldsymbol{\alpha}},\boldsymbol{\alpha}_u,\bar{\boldsymbol{\alpha}}_u\big)\,du, \] with the property that \(\mathbb E[S_t^{\boldsymbol{\alpha},i}]\) has nonnegative drift for arbitrary controls and zero drift at equilibrium [1812.00632]. In the LQ case one uses a quadratic ansatz in \(X_t-\bar X_t\), \(\bar X_t\), and linear terms, and drift cancellation yields Riccati equations together with mean-field BSDEs. In the common-noise case, the Riccati objects become backward stochastic Riccati equations adapted to the common-noise filtration, \[ dK_t^i = -\Phi_t^{i,0}(K_t^i,\dots)\,dt + Z_t^{K^i}\,dW_t^0, \qquad d\Lambda_t^i = -\Psi_t^{i,0}(\Lambda_t^i,\dots)\,dt + Z_t^{\Lambda^i}\,dW_t^0, \] and the linear terms are governed by a mean-field BSDE driven by both idiosyncratic and common noises [1812.00632]. The martingale property is therefore imposed not on a deterministic value functional but on a random functional adapted to the common noise. The same logic appears in LQ McKean–Vlasov control. There, the weak martingale approach leads to verification via suitably chosen processes whose drift vanishes under the optimal control, with the common-noise extension replacing unconditional means by conditional means with respect to \(\mathcal F_t^0\) and adding a common-noise diffusion term [1810.10532]. This suggests a general pattern: in tractable common-noise models, equilibrium or optimality is often encoded by the vanishing of the drift of a carefully designed semimartingale functional of \((X_t,\mu_t)\). ## 5. Qualitative phenomena: blow-ups, contractivity, ergodicity, and uniqueness recovery Common noise does not have a single qualitative effect. In singular models with positive feedback it can trigger or prevent discontinuities. In the conditional McKean–Vlasov contagion model \[ \begin{cases} X_t = X_0 + \sqrt{1-\rho^2}\,B_t + \rho B_t^0 - \alpha L_t,\\ L_t = \mathbb{P}(\tau\le t \mid B^0),\\ \tau = \inf\{t\ge 0 : X_t\le 0\}, \end{cases} \] the loss process \(L_t\) may jump even though the driving noises are continuous, and the common noise can either provoke or avert such blow-ups depending on the path realization [1807.05126]. The limiting relaxed formulation uses a random probability measure \(\mathbf P\) satisfying \[ \mathbf P = \mathrm{Law}(X\mid B^0,\mathbf P), \qquad L_t = \mathbb P(\tau\le t\mid B^0,\mathbf P), \] together with a minimal jump condition \[ \Delta L_t = \inf\{x\ge 0 : \boldsymbol{\nu}_{t-}([0,\alpha x])