Mirror bounds: and
ProvedOWMR.mirror_boundsapproximation-algorithmsinventoryinventory-theoryoperations-researchrenewal-theorystochastic-control
Let demands be i.i.d. with law on and positive mean, let , , and let with width . The mirror cost of the policy started at , the smallest long-run average warehouse-rate loss over fixed centers , where and is the post-order position, satisfies
The first bound evaluates the center at depletion , where the loss is at most on ; the second evaluates it at depletion , where the loss is times the depletion, and uses .
Formalization Note The mirror cost is mirrorCost ρ h p s S, the infimum over centers of mirrorAvg.
Preamble
import Definitions.Def_owmr_renewal open MeasureTheory open scoped ENNReal NNReal
Formal statement
namespace OWMR
theorem mirror_bounds (ρ : Measure ℝ≥0) [IsProbabilityMeasure ρ]
(hρ : 0 < ∫⁻ x, (x : ℝ≥0∞) ∂ρ) (h p s S : ℝ) (hh : 0 < h) (hp : 0 < p) (hsS : s < S) :
mirrorCost ρ h p s S ≤ ENNReal.ofReal (h * p / (h + p) * (S - s)) ∧
mirrorCost ρ h p s S ≤ ENNReal.ofReal (h * (S - s) / 2) := by sorry
end OWMRSource
M. Hu, "A 5/2 Guarantee for Echelon Stock (s,S) Policies in One-Warehouse Multi-Retailer Systems", working paper (manuscript of November 2026), Lemma 4 (Mirror Bounds and Candidate Cost), first sentence, p. 15, and Step 1 of its proof; mirror cost B_w of eq. (14).