Classical rules with an outcome-dependent upstream remainder
DefinitionTwoStageRnQ_RuleRandTwo classes of shipment policies on the model TwoStageRnQ_Model, for fixed integer reorder points and lot parameters . Write , , and for the physical stock held at Stage 2.
FollowsRnQRand S π R₁ R₂ Q n: there are an upstream remainder , measurable with respect to the information at time (so it may vary with the outcome), and an integer process such that, almost surely, for every :
- ;
- ;
- divides ;
- ;
- ;
- .
With a constant this is Rule.FollowsRnQ for the controls .
PaperRnQAll S π R₁ R₂ Q n transcribes the paper's compatible initialization and operating rule (Appendix A.2) with no condition on the remainder: there is a process of pending downstream requests such that, almost surely, the initial Stage-2 pipeline consists of complete lots ( divides for ) and, for every , , divides , divides , , and implies .
Formalization Note The remainder of a PaperRnQAll policy may be any time-0 random variable. The companion target TwoStageRnQ.paper_rule_follows_rand proves PaperRnQAll → FollowsRnQRand and FollowsRnQ → FollowsRnQRand. Conditions of the paper that this inclusion does not need (finite moments of the initial state, complete Stage-1 pipeline lots, FIFO order) are left out of PaperRnQAll, which only enlarges it. The earlier definition TwoStageRnQ_Rule covers a fixed remainder only: its FollowsRnQ holds for compatible initializations whose remainder equals the control .
import Definitions.Def_TwoStageRnQ_Rule
/-!
# Classical `(R,nQ)` rules from compatible initializations with an outcome-dependent remainder
Definitions accompanying "A 5/3 Guarantee for Echelon Stock (R,nQ)
Policies in Two-Stage Stochastic Serial Systems" (manuscript `rq_167.tex`), Section 4.1 and
Appendices A.2-A.3, on the model `TwoStageRnQ_Model`.
The controls `R₁, R₂ ∈ ℤ` and `Q, n ≥ 1` are fixed numbers. With `IP_i = x_i + U_i - N`,
`IL₂(t) = x₂ + U₂(t - L₂) - N(t)` and the Stage-2 physical stock `W = IL₂ - IP₁`:
* `FollowsRnQRand S π R₁ R₂ Q n`: there are an upstream remainder `ρ(ω) ∈ {0, …, Q-1}`, known at time
`0` and allowed to vary with the outcome, and an integer nominal downstream position `Î₁` such
that, almost surely, for all `t ≥ 0`: `IP₂(0) - ρ ≡ Î₁(0) (mod Q)`;
`Î₁(t) + N(t) ≡ Î₁(0) + N(0) (mod Q)`; `nQ` divides `U₂(t) - U₂(0)`; `R₂ < IP₂(t) ≤ R₂ + nQ`;
`R₁ < Î₁(t) ≤ R₁ + Q`; and `IP₁(t) = min(Î₁(t), IL₂(t) - ρ)`. For a constant `ρ` this is
`Rule.FollowsRnQ` with controls `(R₁, R₂, Q, n, ρ)`.
* `PaperRnQAll S π R₁ R₂ Q n`: a literal transcription of the paper's compatible initialization and
operating rule with no condition on the remainder: there is a process `β ≥ 0` of pending
downstream requests such that, almost surely, the initial Stage-2 pipeline consists of complete
lots (`nQ ∣ U₂(0) - U₂(s)` for `s ∈ [-L₂, 0]`) and, for all `t ≥ 0`, the nominal position
`IP₁ + Qβ` lies in `{R₁+1, …, R₁+Q}`, downstream dispatches are complete lots
(`Q ∣ U₁(t) - U₁(0)`), upstream orders are lots of `nQ` with `IP₂ ∈ {R₂+1, …, R₂+nQ}`, and a
pending request implies `W < Q`. Its remainder `W(0) mod Q` may be any time-`0` random variable.
-/
open MeasureTheory
noncomputable section
namespace TwoStageRnQ
namespace RuleRand
/-- `π` follows the classical `(R, nQ)` rule with fixed controls `(R₁, R₂, Q, n)` from a compatible
initialization whose upstream remainder `ρ`, known at time `0`, may vary with the outcome. -/
def FollowsRnQRand {P : Params} (S : DemandSpace P) (π : ShipmentPolicy P S) (R1 R2 : ℤ)
(Q n : ℕ+) : Prop :=
∃ ρ : S.Ω → ℕ, Measurable[S.𝓕 0] ρ ∧ (∀ ω, ρ ω < (Q : ℕ)) ∧
∃ Ihat : ℝ → S.Ω → ℤ, ∀ᵐ ω ∂S.ν,
((Q : ℕ) : ℤ) ∣ (π.x2 ω + π.U2 0 ω - S.N 0 ω) - (ρ ω : ℤ) - Ihat 0 ω ∧
∀ t : ℝ, 0 ≤ t →
((Q : ℕ) : ℤ) ∣ (Ihat t ω + S.N t ω) - (Ihat 0 ω + S.N 0 ω) ∧
((n : ℕ) : ℤ) * ((Q : ℕ) : ℤ) ∣ π.U2 t ω - π.U2 0 ω ∧
(R2 < π.x2 ω + π.U2 t ω - S.N t ω ∧
π.x2 ω + π.U2 t ω - S.N t ω ≤ R2 + ((n : ℕ) : ℤ) * ((Q : ℕ) : ℤ)) ∧
(R1 < Ihat t ω ∧ Ihat t ω ≤ R1 + ((Q : ℕ) : ℤ)) ∧
π.x1 ω + π.U1 t ω - S.N t ω
= min (Ihat t ω) (π.x2 ω + π.U2 (t - P.L2) ω - S.N t ω - (ρ ω : ℤ))
/-- Literal transcription of the paper's classical complete-lot rule with controls
`(R₁, R₂, Q, n)` from a compatible initialization, with no condition on the upstream remainder. -/
def PaperRnQAll {P : Params} (S : DemandSpace P) (π : ShipmentPolicy P S) (R1 R2 : ℤ)
(Q n : ℕ+) : Prop :=
∃ β : ℝ → S.Ω → ℤ, ∀ᵐ ω ∂S.ν,
(∀ s : ℝ, -P.L2 ≤ s → s ≤ 0 → ((n : ℕ) : ℤ) * ((Q : ℕ) : ℤ) ∣ π.U2 0 ω - π.U2 s ω) ∧
∀ t : ℝ, 0 ≤ t →
0 ≤ β t ω ∧
(R1 < π.x1 ω + π.U1 t ω - S.N t ω + ((Q : ℕ) : ℤ) * β t ω ∧
π.x1 ω + π.U1 t ω - S.N t ω + ((Q : ℕ) : ℤ) * β t ω ≤ R1 + ((Q : ℕ) : ℤ)) ∧
((Q : ℕ) : ℤ) ∣ π.U1 t ω - π.U1 0 ω ∧
((n : ℕ) : ℤ) * ((Q : ℕ) : ℤ) ∣ π.U2 t ω - π.U2 0 ω ∧
(R2 < π.x2 ω + π.U2 t ω - S.N t ω ∧
π.x2 ω + π.U2 t ω - S.N t ω ≤ R2 + ((n : ℕ) : ℤ) * ((Q : ℕ) : ℤ)) ∧
(0 < β t ω →
(π.x2 ω + π.U2 (t - P.L2) ω - S.N t ω) - (π.x1 ω + π.U1 t ω - S.N t ω) < ((Q : ℕ) : ℤ))
end RuleRand
end TwoStageRnQ