Existence of and — the myopic cost has a smallest minimizer, and exists
ProvedSongZipkinFluct.FixedCost.yplus_existsinventorymarkov-decision-processss-policy
Consider the fluctuating-demand inventory model under the standing hypotheses and Assumption 1 (). Recall the myopic cost
where is the expected discounted holding/backorder cost rate at the end of a lead time.
Claim. (1) For every world state , has a smallest global minimizer (the paper defines as the smallest value of minimizing , which exists since is convex and coercive, Lemmas 1–2). (2) For any such family the minimum is attained.
The paper takes both existence claims for granted when it defines these quantities; they are needed for the bounds and Lemma 6(c).
Preamble
import Mathlib import Definitions.Def_SongZipkinFluct_FixedCost_Bounds
Formal statement
namespace SongZipkinFluct.FixedCost
open Model
/-- Song and Zipkin (1993), pp. 356–357: under the standing hypotheses and Assumption 1
(`αc̄ < p`), for every world state `i` the myopic cost `G⁺(i, ·)` has a smallest global
minimizer `y⁺(i)` (it is convex and coercive, Lemmas 1–2, p. 355), and for any such choice
`y⁺_min = min_i y⁺(i)` exists. -/
theorem yplus_exists {I : Type*} [Countable I] [Nonempty I] [DecidableEq I] (M : Model I)
(hM : M.Standing) (hA1 : M.α * M.cbar < M.p) :
(∃ yplus, IsYPlus M yplus) ∧ ∀ yplus, IsYPlus M yplus → ∃ ymin, IsYPlusMin yplus ymin := by
sorry
end SongZipkinFluct.FixedCostSource
Song and Zipkin, Inventory Control in a Fluctuating Demand Environment, Oper. Res. 41(2):351–370 (1993), DOI 10.1287/opre.41.2.351, pp. 356–357 (definition of y⁺(i) p. 356 via Lemmas 1–2 p. 355; y⁺_min = min_i y⁺(i) p. 357)