Вы находитесь на странице: 1из 49

6.

The Bernoulli and Poisson


Processes

Broadband Communication, NCTU, Taiwan

Outline
Introduction

The Bernoulli Process


The Poisson Process

Broadband Communication, NCTU, Taiwan

Introduction
A stochastic process is simply a (finite

or infinite) sequence of random


variables.
Example: The sequence of

Daily prices of a stock;


Scores in a football game;
Failure times of a machine;
Hourly traffic loads at a node of a
communication network.

Broadband Communication, NCTU, Taiwan

Emphasis
Dependencies
For example, how do future prices of a
stock depend on past values?
Long-term averages
For example, what is the fraction of time
that a machine is idle?
Boundary events
For example, what is the frequency with
which some buffer in a computer network
overflows with data?

Broadband Communication, NCTU, Taiwan

Two Major Categories of Stochastic


Processes
Arrival-type processes
Bernoulli process (Sec. 6.1)
Poisson process (Sec. 6.2)
Markov processes (Ch. 7)

Broadband Communication, NCTU, Taiwan

6.1 Bernoulli Process


The Bernoulli process is a sequence X1,

X2, of independent Bernoulli random


variables Xi with
P(Xi=1) = P(success at the ith trial) = p,
P(Xi=0) = P(failure at the ith trial) = 1-p,
for each i.

Broadband Communication, NCTU, Taiwan

Random Variables Associated with


the Bernoulli Process
S: number of successes in n

independent trials ~ binomial(p, n).

T: number of trials up to (and including)

the first success ~ geometric(p).

Broadband Communication, NCTU, Taiwan

Properties of the Bernoulli Process


Fresh-Start
For any given time n, the sequence of RV
Xn+1, Xn+2, is also a Bernoulli process, and
is independent from X1, , Xn.

Memorylessness
n: given time

: the time of the first success after time n

~ geometric(p), and is independent


of the RV X1, , Xn, i.e.,

Broadband Communication, NCTU, Taiwan

Example 6.1: Independence


(a) Let U be the number of successes in trials 1

to 5. Let V be the number of successes in trials


6 to 10. Then, U and V are independent. This is
because U = X1 + + X5, V = X6+ + X10,
and the two collections {X1, . . . , X5}, {X6, . . . ,
X10} have no common elements.

(b) Let U (respectively, V ) be the first odd

(respectively, even) time i in which we have a


success. Then, U is determined by the odd-time
sequence X1, X3, . . ., whereas V is determined
by the even-time sequence X2, X4, . . .. Since
these two sequences have no common
elements, U and V are independent.
Broadband Communication, NCTU, Taiwan

Example 6.2 (1/3)


A computer executes two types of tasks,

priority and nonpriority, and operates in


discrete time unit (slots).
Priority task : probability=p, slot
independent.

Broadband Communication, NCTU, Taiwan

10

Example 6.2 (2/3)


T = the time index of the first idle slot

B = the length (number of slots) of the

first busy period


I = the length of the first idle period
Z = the number of slots after the first
slot of the first busy period up to and
including the first subsequent idle slot
Z=B
pT (k ) p k 1 (1 p),k 1, 2,....
Broadband Communication, NCTU, Taiwan

11

Example 6.2 (3/3)


T is geometrically distributed random

variable with parameter 1p, and its


PMF is
pT (k ) p k 1 (1 p),k 1, 2,....

1
p
,var(T )
.
2
1 p
(1 p )
1
1 p
pI (k ) (1 p ) k 1 p,k 1, 2,....,E[ I ] ,var( I ) 2
p
p

E[T ]

Broadband Communication, NCTU, Taiwan

12

Example 6.3: Fresh-Start at a


Random Time
N: first i for which Xi-1 = Xi = 1.

P(XN+1 = XN+2 = 0) = ?
Sol:

Broadband Communication, NCTU, Taiwan

13

Interarrival Times
Yk: the time of the kth success (or

arrival)
Tk: kth interarrival time
T1 = Y1, Tk = Yk Yk1, k = 2, 3, .
Yk = T1 + T2 + + Tk.

Broadband Communication, NCTU, Taiwan

14

Alternative Description of the


Bernoulli Process
Start with a sequence of independent

geometric RV T1, T2, , with common


parameter p.
Let Tis stand for the interarrival times.
Record a success (or arrival) at times T1,
T1 + T2, T1 + T2 + T3, etc.

Broadband Communication, NCTU, Taiwan

15

Example 6.4: Rainy Days


After a rainy day, the number of days

until it rains again ~ geometric(p),


independent of the past.
P(It rains on both the 5th and 8th day
of the month.) = ?
Sol: p2

Broadband Communication, NCTU, Taiwan

16

Properties of the kth Arrival Time Yk


Yk = T1 + T2 + + Tk.

Ti ~ i.i.d. geometric(p).
E[Yk] = E[T1] + + E[Tk] = k/p.
var(Yk) = var(T1) + + var(Tk)

= k(1-p)/p2.
The PMF of Yk is
and is known as the Pascal PMF of order
k.
Broadband Communication, NCTU, Taiwan

17

Example 6.5
single foul with probability p

No foul with probability 1p


The number of fouls in different minutes

are assumed independent.


Sixth foul : foul out
She will play 30 minutes if she does not
foul out.
What is the PMF of Alicias playing time?

Broadband Communication, NCTU, Taiwan

18

Example 6.5 - Solution


Bernoulli process with parameter p

Z: Alicias playing time


Y6: the time until the sixth foul
Z= min{Y6,30}

Broadband Communication, NCTU, Taiwan

19

Splitting of a Bernoulli Process


Bernoulli(pq)

Bernoulli(p)

Bernoulli(p(1-q))

Broadband Communication, NCTU, Taiwan

20

Merging of Independent Bernoulli


Processes

Broadband Communication, NCTU, Taiwan

21

Poisson Approximation to the


Binomial
Z ~ Poisson().
For any fixed nonnegative integer k, the

binomial probability

converges to pZ(k) as n and p = /n,


while keeping constant.
The Poisson PMF is a good
approximation to the binomial as long
as = np for large n and small p.
Broadband Communication, NCTU, Taiwan

22

Example 6.6: Poisson/Binomial


approximation

is valid to several decimal places if n


100, p 0.01, and = np.

Broadband Communication, NCTU, Taiwan

23

Example 6.7: Packet Transmission


An n-symbol packet, transmitted over

noisy channel.
Symbol error rate p = 10-4.
Errors of symbols are independent.
How small should n be in order for P(at
least one symbol in error) < 10-3?
Sol:

Broadband Communication, NCTU, Taiwan

24

6.2 Poisson Process - Terminology


Consider an continuous-time arrival

process.
Any real time t is a possible arrival time.
Define P(k, ) = P(there are exactly k
arrivals during an interval of length ).
Positive parameter : arrival rate or
intensity of the process.

Broadband Communication, NCTU, Taiwan

25

Definition of the Poisson Process


(1/2)
An arrival process is called a Poisson

process with rate if it has the


following properties:

Time-homogeneity: The probability P(k, )


of k arrivals is the same for all intervals of
the same length .
Independence: The number of arrivals
during a particular interval is independent
of the history of arrivals outside this
interval.

Broadband Communication, NCTU, Taiwan

26

Definition of the Poisson Process


(2/2)

Small interval probabilities: The


probabilities P(k, ) satisfy
P(0, ) = 1 + o(),
P(1, ) = + o1(),
P(k, ) = ok(), for k=2, 3,
Here, o() and ok() are functions of that
satisfy

Broadband Communication, NCTU, Taiwan

27

Interpretation of the Definition


Time-homogeneity
Arrivals are equally likely at all times.
Independence
Analogous to the independence of trials in a
Bernoulli process.
Small interval probabilities
For small ,
P(0 arrival) : 1 ;
P(1 arrival) : ;
P(2 or more arrivals) : 0.
Broadband Communication, NCTU, Taiwan

28

Bernoulli Approximation of the


Poisson Process

Broadband Communication, NCTU, Taiwan

29

Random Variables Associated with


the Poisson Process
N: Number of arrivals in a Poisson

process with rate , over an interval of


length ~ Poisson().

T: The time until the first arrival ~

exponential().

Broadband Communication, NCTU, Taiwan

30

Poisson v.s. Bernoulli


POISSON

BERNOULLI

Times of
Arrival

Continuous

Discrete

PMF of # of
Arrivals

Poisson

Binomial

Interarrival
Time CDF

Exponential

Geometric

Arrival Rate

/ unit time

p / per trial

Broadband Communication, NCTU, Taiwan

31

Example 6.8: Check Email


Email arrival: Poisson process, = 0.2

messages/hour.
You check email every hour.
P(0 new message) = ?
P(1 new message) = ?
If you have not checked email for a
whole day, P(no new message) = ?
Sol:

Broadband Communication, NCTU, Taiwan

32

Example 6.9
The Sum of Independent Poisson

Random Variables is Poisson.


Arrival of customers are modeled by a
Poisson process with a rate of = 10
customers per minute.
Let M be the number of customers
arriving between 9:00 and 9:10. Also,
let N be the number of customers
arriving between 9:30 and 9:35. What
is the distribution of M + N?
Broadband Communication, NCTU, Taiwan

33

Example 6.9 - Solution


M is Poisson with parameter = 1010 = 100 and

N is Poisson with parameter = 105 = 50. M


and N are independent.

Derive it in a more direct and intuitive manner:

Let be the number of customers that arrive


between 9:10 and 9:15. Note that has the
same distribution as N (Poisson with parameter
50). Furthermore, is also independent of N.
Thus, the distribution of M + N is the same as
the distribution of M + . But M + is the
number of arrivals during an interval of length 15,
and has therefore a Poisson distribution with
parameter 10 15 = 150.
Broadband Communication, NCTU, Taiwan

34

Properties of the Poisson Process


Fresh-Start
For any given time t > 0, the history of the
process after time t is also a Poisson
process, and is independent from the
history of the process until time t.
Memorylessness
t: given time

: the time of the first arrival after time t

~ exponential(), and is independent


of the history of the process until time t.

Broadband Communication, NCTU, Taiwan

35

Example 6.10
You and your partner go to a tennis

court, and have to wait until the players


occupying the court finish playing.
Assume (somewhat unrealistically) that
their playing time has an exponential
PDF. Then the PDF of your waiting time
(equivalently, their remaining playing
time) also has the same exponential
PDF, regardless of when they started
playing.

Broadband Communication, NCTU, Taiwan

36

Example 6.11: Queuing in the Bank


You enter the bank.
You see 3 tellers are serving other

customers.
No other customers in queue.
Service time ~ i.i.d. exp. RV.
P(You will be the last to leave.) = ?
Sol:

Focus at the moment when you start service


with one of the tellers. Then, the remaining
time of each of the other two customers being
served, as well as your own remaining time,
have the same PDF. Therefore, you and the
other two customers have equal probability
1/3 of being the last to leave.
Broadband Communication, NCTU, Taiwan

37

Alternative Description of the


Poisson Process
Start with a sequence of independent

exponential RV T1, T2, , with common


parameter .
Let Tis represent the interarrival times.
Record an arrival at times T1, T1 + T2,
T1 + T2 + T3, etc.

Broadband Communication, NCTU, Taiwan

38

Properties of the kth Arrival Time Yk


Yk = T1 + T2 + + Tk.

Ti ~ i.i.d. exp().
E[Yk] = E[T1] + + E[Tk] = k/.
var(Yk) = var(T1) + + var(Tk)

= k/2.
The PDF of Yk is
and is known as the Erlang PDF of order
k.
Broadband Communication, NCTU, Taiwan

39

Example 6.12
You call the IRS hotline and you are

told that you are the 56th person in line,


excluding the person currently being
served. Callers depart according to a
Poisson process with a rate of = 2 per
minute. How long will you have to wait
on the average until your service starts,
and what is the probability you will
have to wait for more than an hour?

Broadband Communication, NCTU, Taiwan

40

Splitting a Poisson Process


: Poisson process with rate

Split

For each point toss a coin (with bias

p):

p the point goes to A1(t)


With probability 1-p the point goes to A2(t)

With probability

A1(t) and A2(t) are two independent


Poisson processes with rates

Broadband Communication, NCTU, Taiwan

41

Example 6.13: Splitting of Poisson


Process
Packet is destined for that node with

probability p, or else it must be relayed


to another node with probability 1p
Packets arrive according to a Poisson
process with rate . Packets are
independent to each other.
The process of local packet arrivals is
Poisson with rate p.

Broadband Communication, NCTU, Taiwan

42

Merging Two Poisson Processes


merge

Merging two independent Poisson

processes with rates 1 and 2 creates


a Poisson process with rate 1+2.

Broadband Communication, NCTU, Taiwan

43

Example 6.14: Merging of Poisson


Process
People with letters to mail arrive at the

post office according to a Poisson


process with rate 1, while people with
packages to mail arrive according to an
independent Poisson process with rate
2.

Broadband Communication, NCTU, Taiwan

44

Example 6.15: Competing Exponentials


Two light bulbs.

Independent exp. lifetimes Ta and Tb.


Parameters a and b, respectively.

What is the distribution of the first time

when a bulb burns out?


Sol:

FZ z Pmin Ta , Tb z 1 Pmin Ta , Tb z
1 PTa z , Tb z 1 PTa z PTb z
1 e a z e b z 1 e a b z
Broadband Communication, NCTU, Taiwan

45

Example 6.16: More on Competing


Exponentials (1/2)
Three light bulbs have independent

exponentially distributed lifetimes with


a common parameter . What is the
expected value of the time until the last
bulb burns out?

Sol:

Let T1, T2, T3 be the time to the first,


second and third burnout, respectively.

Broadband Communication, NCTU, Taiwan

46

Example 6.16: More on Competing


Exponentials (2/2)
T1: exponential with parameter 3

T2: exponential with parameter 2


T3: exponential with parameter

Broadband Communication, NCTU, Taiwan

47

The Random Incidence Paradox


L: interarrival time of a Poisson process.

L ~ exp? Or L ~ Erlang of order two?

Broadband Communication, NCTU, Taiwan

48

Example 6.17: Random Incidence in a


Non-Poisson Arrival Process
Buses arrive on the hour and 5 minutes

after the hour.


Average interarrival time = 30 minutes
A person shows up uniformly randomly
within a particular hour.
E[length of the chosen interarrival
interval] = ?
1
11
Sol:
5

12

55

12

50.83,

Which is larger than the avg. interarrival


time
Broadband Communication, NCTU, Taiwan

49