Eq. (9.16) implies Eq. (9.18): the relaxed optimum is a lower bound for every powers-of-two policy
ProvedInventoryControl.serial_pot_lower_boundinventory-controllot-sizingmulti-echelonoperations-researchpowers-of-two
Let minimize the serial cost (9.17) over positive batch quantities satisfying the relaxed constraints (9.18). Then every family of positive batch quantities satisfying Roundy's constraints (9.16), with , also satisfies (9.18), and its cost is at least the relaxed optimum:
This is the easy direction of Roundy's argument, and it is the sense in which the relaxed optimum is "a lower bound for the costs when using the constraints (9.16)". The book goes on to argue that the same value bounds the cost of any policy at all, including ones whose batch quantities vary over time; that stronger claim, which needs a model of time-varying policies, is not formalized in this mission.
Preamble
import Definitions.Def_InventoryControl_serial
Formal statement
namespace InventoryControl
theorem serial_pot_lower_bound {N : ℕ} (A e : Fin N → ℝ) (d : ℝ) (Qrel : Fin N → ℝ)
(hopt : ∀ Q : Fin N → ℝ, (∀ i, 0 < Q i) → SerialNested Q →
serialCost A e d Qrel ≤ serialCost A e d Q)
(Q : Fin N → ℝ) (hpos : ∀ i, 0 < Q i) (hpot : SerialPowerOfTwo Q) :
SerialNested Q ∧ serialCost A e d Qrel ≤ serialCost A e d Q := by sorry
end InventoryControl
Source
Sven Axsäter, Inventory Control, 3rd ed., Springer 2015, DOI 10.1007/978-3-319-15729-0, p. 179, Sect. 9.2.2: 'Note that (9.16) implies (9.18) while the opposite is not true. Therefore, by using (9.18) we will get a lower bound for the costs when using the constraints (9.16)'
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.