§7.1, (110) — has the law of
ProvedMartingaleHT.FiniteWaiting.eq_110Let , let be independent standard Brownian motions on a probability space , and let be a standard Brownian motion on a probability space . Write for the identity map and for the zero function. Then, as processes indexed by ,
that is, the three processes , and have the same law on path space.
This identifies the noise in the heavy-traffic limit: the arrival and service noise combine into a Brownian motion with variance , and the abandonment noise vanishes because its time change converges to zero.
Formalization Note The display (110) appears in the proof of Theorem 7.1 (§7.1); the proof of Theorem 1.2 invokes it through "The proof can be much the same as in §7.1" (p. 250). Equality in law is IdentDistrib of the paths indexed by (product σ-algebra). is realized as in . Brownian motion is ErlangA.Diffusion.IsStandardBM.
import Mathlib import Definitions.Def_ErlangA_Diffusion_SDE open MeasureTheory ProbabilityTheory Filter Topology open scoped ENNReal NNReal
namespace MartingaleHT.FiniteWaiting
open ErlangA.Diffusion
/-- **§7.1, (110)** (p. 250). Let `µ > 0`, let `B₁`, `B₂`, `B₃` be independent standard Brownian
motions on `(Ω, P)`, and let `B` be a standard Brownian motion on `(Ω'', P'')`. With `e(t) = t`
and `η(t) ≡ 0`,
`B₁ ∘ µe − B₂ ∘ µe − B₃ ∘ η =ᵈ B₁ ∘ µe − B₂ ∘ µe − η =ᵈ √(2µ) B`
as processes indexed by `t ≥ 0` (equality of laws on path space). -/
theorem eq_110 (μ : ℝ) (hμ : 0 < μ) {Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω)
[IsProbabilityMeasure P] (B₁ B₂ B₃ : ℝ≥0 → Ω → ℝ)
(h₁ : IsStandardBM P B₁) (h₂ : IsStandardBM P B₂) (h₃ : IsStandardBM P B₃)
(hind : iIndepFun (![fun ω (t : ℝ≥0) => B₁ t ω, fun ω (t : ℝ≥0) => B₂ t ω,
fun ω (t : ℝ≥0) => B₃ t ω] : Fin 3 → Ω → ℝ≥0 → ℝ) P)
{Ω'' : Type*} [MeasurableSpace Ω''] (P'' : Measure Ω'') [IsProbabilityMeasure P'']
(B : ℝ≥0 → Ω'' → ℝ) (hB : IsStandardBM P'' B) :
IdentDistrib (fun ω (t : ℝ≥0) => B₁ (μ.toNNReal * t) ω - B₂ (μ.toNNReal * t) ω - B₃ 0 ω)
(fun ω (t : ℝ≥0) => B₁ (μ.toNNReal * t) ω - B₂ (μ.toNNReal * t) ω - 0) P P ∧
IdentDistrib (fun ω (t : ℝ≥0) => B₁ (μ.toNNReal * t) ω - B₂ (μ.toNNReal * t) ω - 0)
(fun ω (t : ℝ≥0) => Real.sqrt (2 * μ) * B t ω) P P'' := by sorry
end MartingaleHT.FiniteWaiting
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.