Theorem 2, p. 216 — the limit SDE has a unique strong solution
ProvedErlangA.Diffusion.sde_well_poseddiffusion-approximationp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1stochastic-differential-equations
Let , and , and let
Let be a probability space carrying a standard Brownian motion (every path continuous) and a real random variable independent of . Then:
- Existence. There is a process with , continuous paths on , each measurable with respect to , such that almost surely, for all ,
- Pathwise uniqueness. If and both solve this equation driven by the same and almost surely, then almost surely for all .
This makes " is the unique solution of the stochastic differential equation" in Theorem 2 meaningful; the paper omits the discussion and refers to Dupuis and Ishii (1993) and Mandelbaum and Pats (1995).
Formalization Note Because the noise is additive and is Lipschitz, the equation is written pathwise with a Lebesgue integral; a solution is a strong solution. The hypotheses , are those of Theorem 2.
Preamble
import Mathlib import Definitions.Def_ErlangA_Diffusion_SDE open MeasureTheory ProbabilityTheory Filter Topology open scoped ENNReal NNReal
Formal statement
namespace ErlangA.Diffusion
/-- Well-posedness of the limit SDE of Theorem 2 (p. 216; p. 224). Let `μ > 0`, `θ > 0` and
`β ∈ ℝ`. On a probability space carrying a standard Brownian motion `b` (every path continuous)
and a real random variable `X₀` independent of `b`:
1. (existence) there is a process `q` with `q(0) = X₀`, adapted to the filtration generated by
`X₀` and `b`, solving `dq(t) = f(q) dt + √(2μ) db(t)`;
2. (pathwise uniqueness) any two solutions driven by the same `b` with almost surely equal
initial values are indistinguishable on `[0, ∞)`. -/
theorem sde_well_posed (μ β θ : ℝ) (hμ : 0 < μ) (hθ : 0 < θ)
{Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω) [IsProbabilityMeasure P]
(b : ℝ≥0 → Ω → ℝ) (hb : IsStandardBM P b)
(X0 : Ω → ℝ) (hX0 : Measurable X0)
(hind : IndepFun X0 (fun ω (t : ℝ≥0) => b t ω) P) :
(∃ q : Ω → ℝ → ℝ, (∀ ω, q ω 0 = X0 ω) ∧ IsAdaptedToInitialAndBM b X0 q ∧
SolvesDiffusionSDE P μ β θ b q) ∧
∀ q q' : Ω → ℝ → ℝ, SolvesDiffusionSDE P μ β θ b q → SolvesDiffusionSDE P μ β θ b q' →
(∀ᵐ ω ∂P, q ω 0 = q' ω 0) → ∀ᵐ ω ∂P, ∀ t : ℝ, 0 ≤ t → q ω t = q' ω t := by sorry
end ErlangA.Diffusion
Source
Garnett, Mandelbaum & Reiman, Designing a Call Center with Impatient Customers, M&SOM 4(3), 2002, p. 216, Theorem 2 ("q is the unique solution"); p. 224, App. C (uniqueness remark)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.