§5 and (6), p. 1259 — f*^{L−1}(x^{L−2}) ≤ q^{L−1}(x^{L−1}) + f*^L(x^{L−1}) and f*^L(x^{L−1}) ≥ γE[f*^{L−1}(x_+^{L−2})]
ProvedLostSalesOldNew.LeadTime.lead_time_monotoneConsider two lost-sales systems as in Zipkin (2008), §2, identical except for the lead time: system and system , with . Both have cost data , discount factor , and i.i.d. demand with law on with finite mean. Let and be their infinite-horizon optimal discounted costs.
A state of system is ; removing its most recent order gives , a state of system . Let be the one-period cost of §2, and let be the first components of the next state of system under demand . Then:
- the cost of system is at most the cost of system plus the one extra period cost that system incurs:
- inequality (6) holds:
The first inequality makes formal, in the discounted case, the paper's claim that the optimal cost is nondecreasing in the lead time: the two systems have state spaces of different dimension, and the comparison is at matching states. The second inequality is the lower bound on system computed from the solution of system , used for the numerical bounds of the paper.
Formalization Note. In Lean the lead times are and , avoiding natural-number subtraction; is x : Fin (n + 1) → ℝ, is Fin.init x, and is shift x d. Optimal costs take values in , so the first inequality is kept in the paper's additive form rather than the "turned around" form with subtraction. The optimal cost is the limit of the finite-horizon costs of (2). The cost signs, (both part of the paper's setting), finite mean demand and are hypotheses.
import Mathlib import Definitions.Def_LostSalesOldNew_LeadTime_Model
namespace LostSalesOldNew.LeadTime
open MeasureTheory
theorem lead_time_monotone (K : Costs) (D : Measure ℝ) [IsProbabilityMeasure D]
(hD : LostSalesOldNew.StateReduction.IsDemand D) (hmean : Integrable (fun d : ℝ => d) D)
(n : ℕ) (hn : 1 ≤ n) (x : Fin (n + 1) → ℝ) (hx : ∀ i, 0 ≤ x i) :
fStar K D n (Fin.init x) ≤ ENNReal.ofReal (qL K D n x) + fStar K D (n + 1) x ∧
ENNReal.ofReal K.γ * ∫⁻ d, fStar K D n (LostSalesOldNew.StateReduction.shift x d) ∂D ≤ fStar K D (n + 1) x := by sorry
end LostSalesOldNew.LeadTime
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.