Scribd
Upload
4 views

The_Linear_Birth_Death_Process__An_Inferential_Retrospective

Uploaded by

haidangel29
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
4 views

The_Linear_Birth_Death_Process__An_Inferential_Retrospective

Uploaded by

haidangel29
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 9

The Linear Birth–Death Process:

An Inferential Retrospective

Author: Simon Tavaré


Breakdown by Dang H. Dang

2024-12-10
Queuing system corresponding to the Markov process
Model Definition
The linear birth-death process is a continuous-time Markov process used
to model population dynamics, which translates well into queuing theory.
The process describes the following transitions:
▶ State: The number of individuals (or ”customers”) in the system is
Z (t) at time t, with state space S = {0, 1, 2, . . .}
▶ Rates:
▶ Births: An arrival (birth) occurs at a rate proportional to the current
population: m → m + 1 as rate mλ, where λ > 0 is the per-individual birth
rate.
▶ Deaths: A departure (death) occurs at a rate proportional to the current
population: m → m − 1 as rate mµ, where µ < 0 is the per-individual death
rate.
Queuing system corresponding to the Markov process
Linear B/D Process does not fit perfectly into a specific queuing model:
λ 2λ 3λ
0 1 2 3 ···
µ 2µ 3µ 4µ
However, if we were to compare, it is closest to M/M/∞, with difference
being that arrival rates & service rates change linearly with current state:
λ λ λ
0 1 2 3 ···
µ 2µ 3µ 4µ
Proceeding the long-term behavior of Markov process

λ 2λ 3λ
0 1 2 3 ···
µ 2µ 3µ 4µ
Transition rates: For state i, transition rates are: λi = i · λ, µi = i · µ
Global balance equations: The stationary distribution πi must satisfy:
 i
λi −1 λ 1
λi−1 πi−1 = µi πi ⇒ πi = πi−1 ⇒ πi = π0
µ i µ i!
Proceeding the long-term behavior of Markov process
λ 2λ 3λ
0 1 2 3 ···
µ 2µ 3µ 4µ

1
πi = π0 (λ/µ)i
i!
∞ ∞  i
λ 1
= 1 ⇒ π0 e λ/µ = 1
P P
Normalization: πi = 1 ⇒ π0 µ i!
i=0 i=0
i
e −λ/µ (λ/µ)
Stationary distribution=Poisson distribution: πi = i! .
Revisit mean dynamics of L-B/D: a simpler framework

▶ Willy Feller (1939) showed that if the population starts from a single
individual, the mean population size at time t is E [Z (t)] = e (λ−µ)t .
▶ But we can also use Mean population dynamics (the result of
studying Kolmogorov forward equation) to obtain:
∞ ∞
dM (t) X X
= B rate − D rate = λn Pn (t) − µn Pn (t) = (λ − µ) M (t)
dt n=0 n=0
It is known that the process eventually dies out or tends to ∞, and that:
 µ
P (population dies out) = min 1,
λ
Distribution of Z (t)
Defining p1m (t) = P (Z (t) = m | Z (0) = 1)
p1m (t) = (1 − α (t)) (1 − β (t)) β (t)m−1 , m = 1, 2, 3, . . .
p10 (t) = α (t)
where, for λ ̸= µ,

µ e (µ−λ)t − 1 λ
α (t) = , β (t) = α (t)
λe (µ−λ)t − µ µ
while, for λ = µ,
λt
α (t) = = β (t)
1 + λt
We can completely calculate pnm using the alpha and beta terms. As an
example of historical inference in the process, we can find the distribution
of the number of families at time t: P (F (t) = j | Z (t) = m, Z (0) = n).
Simulating the process

The earlier results can be exploited for simulating the process at discrete
time points 0 = t0 < t1 < t2 < · · · < tm = t.
Write Zi = Z (ti ) , ∆i = ti − ti−1 . Starting from Z0 = Z (0) = n, simulate

Fl ∼ Bin (Zl−1 , 1 − α (∆l )) , Zl ∼ NB (Fl , β (∆l ))


Application to vehicles parking management system
This section models a parking lot with capacity K , using the generalized
linear birth-death process. Key assumptions include:
▶ Vehicles arriving depend on available spaces K − n, and vehicles
departing depend on current occupancy n.
▶ N (t) represents the number of parked vehicles at time t, to govern it:
Transition rates:
▶ Arrival rate: λ (n) = (K − n) λ
▶ Departure rate: µ (n) = nµ
λ
K 1 − e −(λ+µ)t

Expected occupancy : E [N (t)] = λ+µ
State probabilities: Please visit here: On the Generalized Birth–Death
Process and Its Linear Versions (Vishwakarma, P., and K. K. Kataria).

You might also like