§2, (6), p. 656 — c(s − 1, S) is a weighted average of c(s, S) and G(s)
ProvedZhengFedergruenSS.Algorithm.eq6_weighted_averageinventoryp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1renewal-theorys-s-policy
In the model of Zheng and Federgruen (demand law with , fixed cost , unimodal, the growth condition), let . For integers and ,
So lowering the reorder level by one moves the cost toward . This identity underlies Lemma 0 and through it every comparison the algorithm makes.
Formalization Note. is possible: it happens when , e.g. for demand concentrated on . The upper bound is therefore weak, as printed.
Preamble
import Mathlib import Definitions.Def_VeinottWagnerSS_RenewalCost_Demand import Definitions.Def_ZhengFedergruenSS_Algorithm_Model
Formal statement
namespace ZhengFedergruenSS.Algorithm
/-- Display (6), §2, p. 656 (Zheng and Federgruen 1991): with `α_n = M(n)/M(n + 1)`,
`0 < α_n ≤ 1` and, for `s < S` and `n = S − s`,
`c(s − 1, S) = α_n c(s, S) + (1 − α_n) G(s)`.
Formalization Note: `α_n` may equal `1` (when `m(n) = 0`, e.g. for demand concentrated on
`{0, 2}`), so the upper bound is weak, as printed. -/
theorem eq6_weighted_average (D : VeinottWagnerSS.RenewalCost.DemandDist) (hp0 : D.φ 0 < 1)
(K : ℝ) (hK : 0 < K) (G : ℤ → ℝ) (hG : NegUnimodal G) (hgrowth : GrowthCond K G)
(s S : ℤ) (hsS : s < S) :
0 < alpha D.φ (S - s).toNat ∧ alpha D.φ (S - s).toNat ≤ 1 ∧
c D.φ K G (s - 1) S =
alpha D.φ (S - s).toNat * c D.φ K G s S + (1 - alpha D.φ (S - s).toNat) * G s := by sorry
end ZhengFedergruenSS.Algorithm
Source
Zheng and Federgruen, Finding Optimal (s, S) Policies Is About As Simple As Evaluating a Single Policy, Oper. Res. 39(4):654–665 (1991), DOI 10.1287/opre.39.4.654, p. 656, §2, (6)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.