BCR regrouping with local linear potential certificates
OpenKServer.bcr_lemma15_potential_certificateFix α > 0, an integer β ≥ 2, and a level w with α(w+1)² > 1. Assume the mission's BCRInductiveSubchunks predicate at level w. On the same next-level space, there is a chunk construction with constant initial information and size window
required expected total, escape price, and minimum chunk count
The original probability, filtration, adaptedness, request-feasibility, terminal-request, offline-cost and horizon constraints are retained. For every chunk i and filtration atom A containing ω₀ there is a local potential v with a finite-depth certificate for the atom weights and that chunk's request law, satisfying
at every position x compatible with the known preceding request history. The local certificate consists of escape, movement, branch-sum and terminal linear inequalities, with recursively certified continuation potentials.
This remains an open construction problem. Its cost criterion is an exact local-potential formulation of the Bellman lower-value criterion, rather than an additional regularity assumption on the input subchunk system. The distribution and requests still have to be constructed; linearity concerns the potential variables after those data are fixed.
import Definitions.Def_KServer_bcr_induction import Definitions.Def_KServer_chunk_potentials open KServer
theorem KServer.bcr_lemma15_potential_certificate (α : ℝ) (hα : 0 < α) (β : ℕ) (hβ : 0 < β)
(hβ2 : 2 ≤ β) (w : ℕ) (hw : 1 < α * ((w + 1 : ℕ) : ℝ) ^ 2)
(h : KServer.BCRInductiveSubchunks α β hβ w) :
∃ C : @PotentialChunkSystemB (bcrLevel2 β hβ (w + 1)).carrier
(bcrLevel2 β hβ (w + 1)).metric
(bcrLevel2 β hβ (w + 1)).fin
(bcrLevel2 β hβ (w + 1)).s (bcrLevel2 β hβ (w + 1)).t
((3 : ℝ) ^ (w + 1) / 2) (3 * (3 : ℝ) ^ (w + 1) / 2)
((α * β * ((w + 1 : ℕ) : ℝ) ^ 2) * 3 ^ (w + 1))
(2 * β * (3 : ℝ) ^ (w + 1)) ⌈α * β * ((w + 1 : ℕ) : ℝ) ^ 2⌉₊,
∀ ω₁ ω₂ : C.Ω, C.hist 0 ω₁ = C.hist 0 ω₂ := by sorry