Eq. (5.60) — the G/M/1 arrival-point law is geometric,
ProvedQueueingFundamentals.GM1.arrival_point_geometricConsider the G/M/1 queue: interarrival times are independent with law on of mean (), service times are exponential with rate , and . Let be the probabilities (5.50), the arrival-point transition matrix (5.51), and (5.55). Then there is a real number with the following properties:
- and ;
- is the only complex root of with ;
- the geometric vector
is a probability vector solving , (5.52); 4. it is the only probability vector solving (5.52).
So the number of customers an arrival finds in the system is geometric, as in the M/M/1 queue with replaced by . This is the arrival-point law , not the time-average law : the two agree only for Poisson input.
Formalization Note The root is not a hypothesis: its existence, its location in , and its uniqueness in the unit disk are part of the conclusion, and so is the uniqueness of the stationary vector, which the book's phrase "the steady-state arrival-point distribution is" asserts.
import Mathlib import Definitions.Def_QueueingFundamentals_GM1_EmbeddedChain
namespace QueueingFundamentals.GM1
open MeasureTheory
/-- Eq. (5.60) with the root result of pp.261–262: when `ρ = λ/μ < 1` there is a root `r_0 ∈ (0, 1)`
of `z = β(z)`, it is the only root of `z = β(z)` in the open unit disk, the geometric vector
`q_n = (1 - r_0) r_0^n` solves `qP = q`, `qe = 1`, and it is the only probability vector that does. -/
theorem arrival_point_geometric (A : Measure ℝ) (lam mu : ℝ) (hlam : 0 < lam) (hmu : 0 < mu)
(hA : IsInterarrivalLaw A lam) (hrho : lam / mu < 1) :
∃ r0 : ℝ, r0 ∈ Set.Ioo (0 : ℝ) 1 ∧ beta A mu (r0 : ℂ) = (r0 : ℂ) ∧
(∀ z : ℂ, ‖z‖ < 1 → beta A mu z = z → z = (r0 : ℂ)) ∧
IsArrivalPointStationary A mu (fun n => (1 - r0) * r0 ^ n) ∧
∀ q : ℕ → ℝ, IsArrivalPointStationary A mu q → q = fun n => (1 - r0) * r0 ^ n := by sorry
end QueueingFundamentals.GM1
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.