§A.2.1 — capacity consumed at a node is the sum of the inventories consumed there
ProvedPicardFO.Fulfill.sec_A_2_1_conservationConsider the FO problem with initial state , an arbitrary policy , a sequence of orders and an assignment of products to processes. Let be the states of the sequential run, and let be the states of any process of the Picard iteration simulating with any cache (cached actions replayed when feasible, otherwise). Then for every time and node ,
In words, every unit of capacity used at a node is used by fulfilling one unit of some product's inventory there. Consequently, a process holding at least the sequential inventory of every product at node also holds at least the sequential capacity at , which reduces Lemma A.1's capacity inequality (2) to its inventory inequality (1).
Formalization Note No assumption on is used, and the identity holds for every cache , not only those produced by Algorithm 1. Times are 0-based (procState … α m t and seqState … t are the states before Lean order t, i.e. after orders).
Differences from the source:
- reading: "the capacity consumed at a node is the sum of the inventories consumed at that node. Therefore … x^{k,m} ≥ x^{seq} for all i implies c^{k,m} ≥ c^{seq}" → the identity c = c₁ − Σᵢ (x₁ − x) for process runs and for the sequential run; reason: this is the fact the sentence asserts, and the implication follows from it by monotonicity of the sum.
- generalization: process states of Algorithm 1's iterations → process states for an arbitrary cache α and an arbitrary policy; reason: the dynamics alone give the identity, and Algorithm 1's caches are particular α.
- encoding: 1-based times t ∈ [T] → 0-based times t ∈ ℕ (state after t orders); reason: Lean convention, same instances.
import Mathlib import Definitions.Def_PicardFO_Fulfill_Model
namespace PicardFO.Fulfill
/-- §A.2.1 (pp. 12–13): along every run, the capacity consumed at a node equals the sum over
products of the inventory consumed there, both for every process of Algorithm 1 (any cache `α`)
and for the sequential run. No assumption on `π` is needed. -/
theorem sec_A_2_1_conservation {I J : ℕ} {Ω ι : Type*} (prod : Ω → Fin I)
(π : State I J → Ω → Option (Fin J)) (s₁ : State I J) (ω : ℕ → Ω) (own : Fin I → ι) :
(∀ (α : ℕ → Option (Fin J)) (m : ι) (t : ℕ) (j : Fin J),
(procState prod π s₁ ω own α m t).c j =
s₁.c j - ∑ i, (s₁.x i j - (procState prod π s₁ ω own α m t).x i j)) ∧
(∀ (t : ℕ) (j : Fin J),
(seqState prod π s₁ ω t).c j =
s₁.c j - ∑ i, (s₁.x i j - (seqState prod π s₁ ω t).x i j)) := by sorry
end PicardFO.Fulfill
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.