Theorem 2: the run-truncating FCFS supervisor keeps every buffer bounded under
ProvedKumarSeidman.Supervisor.supervisor_stabilizesConsider any manufacturing system with set-up times that satisfies the capacity condition
Choose supervisor parameters with (24) and thresholds . Apply the supervisor (truncation of runs at , an FCFS priority queue of buffers whose level exceeds , and processing of each queued buffer for exactly unless it clears) at every machine, on top of an arbitrary scheduling policy. Then every trajectory, from every initial state (initial levels, initial set-ups and initial queue order), has all buffer levels bounded over all time:
The capacity condition is necessary for stability of any policy. The paper shows in §III that natural distributed policies (clearing policies) can be unstable even when it holds; this theorem shows that a simple distributed supervisor makes every policy stable whenever it holds.
Formalization Note The bound may depend on the trajectory. The base policy is constrained only by the run structure of the model (contiguous runs, no idling outside runs, full processing rate on a nonempty buffer) and by the supervisor's rules; it need not be a clearing policy.
import Mathlib import Definitions.Def_KumarSeidman_Supervisor_System import Definitions.Def_KumarSeidman_Supervisor_Trajectory import Definitions.Def_KumarSeidman_Supervisor_Mechanism
namespace KumarSeidman.Supervisor
/-- Theorem 2 (Kumar–Seidman 1990, p. 297). Under the capacity condition (1), with run-length
parameters `γ` satisfying (24) and arbitrary thresholds `z ≥ 0`, every trajectory of an
arbitrary scheduling policy modified by the supervisor at all machines keeps every buffer level
bounded over `[0, ∞)`, from every initial state. -/
theorem supervisor_stabilizes {P M : ℕ} (S : System P M) (hcap : S.CapacityCondition)
(γ : Fin M → ℝ) (hγ : GammaCondition S γ) (z : Buffer S → ℝ) (hz : ∀ b, 0 ≤ z b)
(T : Trajectory S) (hT : T.Supervised γ z) :
T.IsStable := by sorry
end KumarSeidman.Supervisor
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.