Proposition 4 — relative maximum-lateness error of BLS
ProvedBurnInBatch.ListSched.proposition_4Let jobs have positive processing times and nonnegative due dates . Fix a positive batch capacity , a positive number of identical parallel machines, and any list containing each job once. Let be the maximum lateness of BLS on that list, the minimum maximum lateness over all valid batch schedules, and the largest due date. Then
The normalization accommodates an optimal maximum lateness that can be negative. This result bounds BLS's relative maximum-lateness error using the batch capacity, machine count, and maximum due date.
Formalization Note The proof's last inequality requires , a condition implicit in the paper's deadline setting but absent from its printed Proposition 4. Without it, the proposition can fail; for one job with , , , and , its right side is negative while its left side is zero. Positive processing times and a nonempty instance ensure . The ratio has exactly the printed factor .
import Mathlib import Definitions.Def_BurnInBatch_ListSched_Model
namespace BurnInBatch.ListSched
/-- Proposition 4, p. 773: the relative maximum-lateness error of BLS.
The hypothesis `hd` (nonnegative due dates) is not printed in the proposition; the
proof's last step `d_max - d^k ≤ d_max` uses it, and without it the statement fails
(one job, `p = 4`, `d = -3`, `B = 1`, `m = 2`). -/
theorem proposition_4 {n : ℕ} (p d : Fin n → ℝ) (B m : ℕ)
(l : List (Fin n)) (hn : 0 < n) (hB : 0 < B) (hm : 0 < m)
(hp : ∀ j, 0 < p j) (hd : ∀ j, 0 ≤ d j)
(hl : ListsJobs Finset.univ l) :
(blsLateness p d B m hm l Finset.univ - LStar p d B m Finset.univ) /
(LStar p d B m Finset.univ + maxDue d hn) ≤
((B : ℝ) - 1 / (m : ℝ)) +
maxDue d hn / (LStar p d B m Finset.univ + maxDue d hn) := by sorry
end BurnInBatch.ListSched
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.