Proof of Theorem 8 — E[K̃] ≤ ⌈(m/ρ′)²⌉ + E[K(ρ′, 2√2 m²/ρ′)]
ProvedAIMInventory.UnknownRho.restart_boundLet be i.i.d. nonnegative demands with finite positive mean, run the non-perishable AIM algorithm with scale , upper bound and costs , and put . Let be the queue , . For let
and let be the first time the random walk started at with increments enters . Then
Up to the deterministic time , the queue is fed at least as fast as the walk with drift ; after it is fed more slowly. The inequality bounds the expected time until the queue first empties after , which is the length of the initial stretch of periods in which the cost gap is bounded only by the trivial bound .
Formalization Note. Both expectations are Lebesgue integrals in , so the inequality is meaningful (and trivially true) when the right side is infinite; and take the value on paths where the defining set is empty. The hypothesis is implicit in the paper (the bound divides by ). Under , .
import Mathlib import Definitions.Def_AIMInventory_UnknownRho_Model noncomputable section open MeasureTheory open scoped ENNReal NNReal
namespace AIMInventory.UnknownRho
/-- The displayed `E[K̃]` inequality in the proof of Theorem 8, p. 14.
Expectations retain `∞` on paths with no return. -/
theorem restart_bound (P : DemandLaw) (c : Parameters) (ρ' : ℝ)
(hρ : 0 < ρ') (hsmall : ρ' < m c) :
(∫⁻ d, restartTime c ρ' d ∂CappedBaseStock.demandPathLaw P) ≤
(Nat.ceil ((m c / ρ') ^ 2) : ℝ≥0∞) + expectedK P ρ' (initialHeight c ρ') := by sorry
end AIMInventory.UnknownRho
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.