Lemma 5.5 — stochastic boundedness for integral representations
ProvedMartingaleHT.InfiniteServer.lemma_5_5integral-equationsp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1stochastic-boundedness
Let be Lipschitz as in (62) (there is with ), fix , and for each let and be real processes with paths in and measurable coordinates such that, almost surely,
If is stochastically bounded in and is stochastically bounded in for each , then is stochastically bounded in .
Applied to the representation (32) with , it reduces stochastic boundedness of to that of the initial values and of the martingales.
Formalization Note is not assumed, as in the paper. The representation is required almost surely.
Preamble
import Mathlib import Definitions.Def_BellWilliams2001_ThresholdPolicy_Paths import Definitions.Def_ManyServerQED_Scheduling_Model import Definitions.Def_ErlangA_Diffusion_SDE import Definitions.Def_ErlangA_Diffusion_Queue import Definitions.Def_MartingaleHT_InfiniteServer_Model import Definitions.Def_MartingaleHT_InfiniteServer_Toolkit open MeasureTheory ProbabilityTheory Filter Topology open scoped ENNReal NNReal BoundedContinuousFunction
Formal statement
namespace MartingaleHT.InfiniteServer
open BellWilliams2001.ThresholdPolicy ManyServerQED.Scheduling ErlangA.Diffusion
/-- **Lemma 5.5** (stochastic boundedness for integral representations, p. 236). Let `h` be
Lipschitz as in (62) and, for each `n ≥ 1`, let `Xₙ`, `Yₙ,₁, …, Yₙ,ₖ` be processes with paths in
`D` and measurable coordinates such that, almost surely, for all `t ≥ 0`,
`Xₙ(t) = Xₙ(0) + Yₙ,₁(t) + ⋯ + Yₙ,ₖ(t) + ∫₀ᵗ h(Xₙ(s)) ds`. If `(Xₙ(0))` is stochastically bounded
in `ℝ` and each `(Yₙ,ᵢ)` is stochastically bounded in `D`, then `(Xₙ)` is stochastically bounded
in `D`. -/
theorem lemma_5_5 {Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω) [IsProbabilityMeasure P]
{k : ℕ} (h : ℝ → ℝ) (c : ℝ) (hc : 0 < c)
(hLip : ∀ s₁ s₂ : ℝ, |h s₁ - h s₂| ≤ c * |s₁ - s₂|)
(X : ℕ → Ω → ℝ → ℝ) (Y : ℕ → Fin k → Ω → ℝ → ℝ)
(hXcad : ∀ n : ℕ, 1 ≤ n → ∀ ω, IsCadlag (X n ω))
(hYcad : ∀ n : ℕ, 1 ≤ n → ∀ i ω, IsCadlag (Y n i ω))
(hXmeas : ∀ n : ℕ, 1 ≤ n → ∀ t, Measurable (fun ω => X n ω t))
(hYmeas : ∀ n : ℕ, 1 ≤ n → ∀ i t, Measurable (fun ω => Y n i ω t))
(hrep : ∀ n : ℕ, 1 ≤ n → ∀ᵐ ω ∂P, ∀ t : ℝ, 0 ≤ t →
X n ω t = X n ω 0 + ∑ i, Y n i ω t + ∫ s in (0 : ℝ)..t, h (X n ω s))
(hX0 : IsSBReal P (fun n ω => X n ω 0))
(hY : ∀ i, IsSBD P (fun n ω t (_ : Fin 1) => Y n i ω t)) :
IsSBD P (fun n ω t (_ : Fin 1) => X n ω t) := by sorry
end MartingaleHT.InfiniteServer
Source
Pang, Talreja & Whitt, Martingale Proofs of Many-Server Heavy-Traffic Limits for Markovian Queues, arXiv:0712.4211v1 (reprint of Probab. Surveys 4 (2007) 193–267), p. 236, Lemma 5.5
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.