Proof of Theorem 4(i), p. 578 — the total cost increase of the one-more-unit policy
ProvedRoslingAssembly.Unbounded.thm4i_proof_total_cost_increaseConsider the assembly model under its standing hypotheses (a tree indexed so that ; i.i.d. demands with a density and a finite positive mean; ; well-formed initial data ). Let be a non-end item, , let , and let be a measurable policy, feasible from , whose cost (2) is a real number . Let . Then the policy that orders more units of every item in period and holds them forever has the real cost
The fraction is the paper's "total cost increase", capitalized to period ; the factor discounts it to period 1, where (2) is valued. Multiplying the condition of Theorem 4(i) by shows that the increase is negative exactly under that condition.
Formalization Note. The paper's argument is for one unit (); the cost is linear in . The statement is restricted to : for the end item the extra units also change and hence the backlog term of (2), which the printed display does not account for. The goal theorem still covers . The cost is an extended real; the hypothesis that the cost of is the real number excludes the value.
import Mathlib import Definitions.Def_RoslingAssembly_Unbounded_Model import Definitions.Def_RoslingAssembly_Unbounded_Perturb
namespace RoslingAssembly.Unbounded
open MeasureTheory
/-- Proof of Theorem 4(i), p. 578, "the total cost increase": for a non-end item `i ≥ 2` and a
measurable feasible policy `π` with real cost `J`, the policy with `δ ≥ 0` extra units of the
subsystem `{i} ∪ B(i)` has real cost
`J + δ α^{M_m − M_i} α^{l_i} (h_i + ∑_{k ∈ B(i)} h_k α^{−(M_{s(k)} − M_{s(i)})}) / (1 − α)`,
where `M_m = Mmax i`; the factor `α^{M_m − M_i}` discounts the paper's increase, capitalized to
period `M_m − M_i + 1`, back to period 1. -/
theorem thm4i_proof_total_cost_increase (S : Model) (x0 : ℕ → ℕ → ℝ)
(hT : S.IsTree) (hI : S.IsIndexed) [IsProbabilityMeasure S.ν] (hD : S.DemandStanding)
(hα0 : 0 < S.α) (hα1 : S.α < 1) (hx0 : S.WellFormed x0)
(i : ℕ) (hi : i ∈ Finset.Icc 2 S.N) (δ : ℝ) (hδ : 0 ≤ δ)
(π : RoslingAssembly.SeriesEquiv.Policy) (hπm : IsMeasurablePolicy π) (hπ : S.Feasible x0 π)
(J : ℝ) (hJ : S.cost π = (J : EReal)) :
S.cost (S.perturb π i δ) =
((J + δ * S.α ^ (S.Mmax i - S.M i) * S.α ^ S.l i *
(S.h i + ∑ k ∈ S.B i,
S.h k * S.α ^ (-((S.M (S.succ k) : ℤ) - (S.M (S.succ i) : ℤ)))) / (1 - S.α) : ℝ) :
EReal) := by sorry
end RoslingAssembly.Unbounded
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.