Sec. 2.2, p. 7 — an optimal constant order exists, optimal levels are bounded away from E[D], r∞ < E[D]
ProvedLostSalesExpGap.ConstOrder.best_constant_existsargminconstant-order-policylost-salesp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1
In the lost-sales model (i.i.d. nonnegative demand with finite mean and strictly positive variance, lead time , costs , empty initial state), let be the long-run average cost of the policy ordering every period, and let
Then:
- is nonempty;
- is bounded away from : there is with for all ;
- ;
- in particular .
This makes the best constant-order policy well defined and places it in the regime where formula (3) applies.
Formalization Note Costs are compared in . Strictly positive variance is encoded as "the demand law is not a point mass" (NonDeterministic). Items 3 and 4 make explicit what the paper uses when it writes and .
Preamble
import Mathlib import Definitions.Def_LostSalesExpGap_ConstOrder_ConstantOrder open MeasureTheory ProbabilityTheory open scoped ENNReal NNReal BigOperators
Formal statement
namespace LostSalesExpGap.ConstOrder
/-- Sec. 2.2, p. 7: an optimal constant-order level exists, the set of optimal levels is bounded
away from `E[D]`, and its infimum `r∞` is itself optimal, so `r∞ < E[D]`. -/
theorem best_constant_exists (P : CappedBaseStock.DemandLaw) (c : CappedBaseStock.Parameters)
(hP : LostSalesConstantOrder.LargeLeadTime.NonDeterministic P) :
(argminConst P c).Nonempty ∧
(∃ δ : ℝ, 0 < δ ∧ ∀ r ∈ argminConst P c, r ≤ CappedBaseStock.mean P - δ) ∧
rInf P c ∈ argminConst P c ∧
rInf P c < CappedBaseStock.mean P := by sorry
end LostSalesExpGap.ConstOrder
Source
Xin, Goldberg, Optimality gap of constant-order policies decays exponentially in the lead time for lost sales models, arXiv:1409.1499v2, Sec. 2.2, p. 7
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.