Stochastic system exam

DNH
FINAL-19-sol.pdf

12/16/2019

ESE 503

Stochastic Systems

Final

1. Suppose you have just landed on JFK after making a tight connection

in Atlanta, GA, and you are waiting for your suitcase to appear on the

carousel. Once the carousel starts moving, all the luggage is delivered

within 10 minutes. Other passengers have started to get their bags,

but you keep waiting, and waiting, and waiting. Assume that the a

priori probability that the luggage made the connection is 1/2. Let

T = {1, 2, · · · , 10} be time elapsed in minutes once the luggage started being delivered. If the luggage is on the plane, it is delivered with

equal probability within any minute from 1 to 10. Five minutes passed

(T = 5), and your luggage has not shown up yet. What are the chances

that your luggage made the connection from Atlanta to JFK?

Solution: Let X = {0, 1} be the random variable denoting if the bag was not or was on the plain, respectively.

Let Y = {0, 1} be the random variable denoting if the bag was not or was on the carousel, respectively.

We want to find

P(X = 1|T = k,Y = 0) = P(Y = 0|T = k,X = 1)P(X = 1)

P(Y = 0|T = k)

We have

P(Y = 0|T = k) = P(Y = 0|T = k,X = 1)P(X = 1)

+ P(Y = 0|T = k,X = 0)P(X = 0)

We write

P(Y = 0|T = k,X = 1) = 10 −k

10

P(Y = 0|T = k,X = 0) = 1

Thus,

P(X = 1|T = k,Y = 0) = 10−k 10 × 1

2 10−k 10 × 1

2 + 1 × 1

2

= 10 −k 20 −k

F-1

When k = 5, we get

P(X = 1|T = 5,Y = 0) = 1

3

2. The lifetime of a machine is modeled by a Weibull random variable X,

where

fX(x) = λx β−1e−λx

β/β, x ≥ 0

What is the pdf of the density of the lifetime of the machine if it is

still operating at time x = T, i.e., fX(x|X > T)?

Solution: We want to find

FX(x|X > T) = P(X ≤ x∩X > T)

P(X > T)

= FX(x) −FX(T)

1 −FX(T) , x > T

Since

FX(x) =

∫ x 0 fX(u)du = 1 −e−λx

β/β

we get

FX(x|X > T) = e−λT

β/β −e−λx β/β

e−λT β/β

, x > T

= 1 −eλT β/β−λxβ/β, x ≥ T

Clearly, the density of X is then

fX(x|X > T) = dFX(x|X > T)

dx

= λxβ−1e −λ

β (x β−Tβ), x ≥ T

3. Let X1,X2, and X3 be three independent exponential random variables

with pdfs given by

fX1 (x) = λ1e −λ1x, x ≥ 0

fX2 (x) = λ2e −λ2x, x ≥ 0

fX3 (x) = λ3e −λ3x, x ≥ 0

We define the random variable Y = min(X1,X2,X3). Find the pdf of

Y , fY (y).

F-2

Solution:

FY (y) = P(min{X1,X2,X3}≤ Y )

= 1 −P(X1 > y,X2 > y,X3 > y)

= 1 −P(X1 > y)P(X2 > y)P(X3 > y)

= 1 −e−λ1ye−λ2ye−λ3y

= 1 −e−(λ1+λ2+λ3)y, y ≥ 0

For the pdf of Y we get

fY (y) = dFY (y)

dy

= (λ1 + λ2 + λ3)e −(λ1+λ2+λ3)y, y ≥ 0

4. Let

fX,Y (x,y) =

{ 2e−xe−y, 0 ≤ y ≤ x < ∞ 0, otherwise

Find E(Y |x) and Var(Y |x).

[Note: ∫ xexdx = ex(x− 1) and

∫ x2exdx = ex(x2 − 2x + 2).]

Solution: We write

fY (y|x) = fX,Y (x,y)

fX(x)

where

fX(x) =

∫ ∞ 0

fX,Y (x,y)dy

=

∫ x 0

2e−xe−ydy

= 2e−x(1 −e−x), 0 ≤ x < ∞

and

fY (y|x) = 2e−xe−y

2e−x(1 −e−x)

= e−y

1 −e−x , 0 ≤ y ≤ x.

This conditional density is a truncated exponential.

F-3

Next we find E(Y |x) and write

E(Y |x) = ∫ x 0 y

e−y

1 −e−x dy

= 1

1 −e−x ( e−y(−1 −y)

)∣∣x 0

= 1 − xe−x

1 −e−x

where we used ∫ xe−xdx = e−x(−x− 1)

For the variance we use

Var(Y |x) = E(Y 2|x) −E ( Y |x

)2 Thus, now we find E(Y 2|x). We write

E(Y 2|x) = ∫ x 0 y2

e−y

1 −e−x dy

= 1

1 −e−x ( e−y(−y2 − 2y − 2)

)∣∣x 0

= 2 −e−x(x2 + 2x + 2)

1 −e−x

where we used ∫ x2e−xdx = e−x(−x2 − 2x− 2)

Finally,

Var(Y |x) = E(Y 2|x) −E ( Y |x

)2 =

2 −e−x(x2 + 2x + 2) 1 −e−x

− (

1 − xe−x

1 −e−x

)2 = 1 −

x2e−x

(1 −e−x)2

5. Let the observation Zn be given by

Zn = Sn + Wn

where Sn is a sinusoid defined by Sn = a cos(ωn + Φ), where Φ is

a uniformly distributed random variable on [0, 2π) and a and ω are

constants. The process Wn is a noise process with autocorrelation

F-4

rw[k]. The processes Sn and Wn are independent processes. Find the

autocorrelation function of Zn. Is Zn a WSS process?

Solution: For the mean of Zn we have

E(Zn) = E(Sn) + E(Wn) = 0

= E ( a cos(ωn + Φ)

) + E(Wn)

=

∫ 2π 0

a cos(ωn + φ) 1

2π dφ + 0

= 0

and

E(Zn+kZn) = E ((Sn+k + Wn+k)(Sn + Wn))

E(Sn+kSn) + E(Sn+kWn) + E(Wn+k)(Sn) + E(Wn+k)Wn)

= E(Sn+kSn) + rw[k]

For E(Sn+kSn) we write

E(Sn+kSn) =

∫ 2π 0

a cos(ω(n + k) + φ)a cos(ωn + φ) 1

2π dφ

= a2

∫ 2π 0

cos(ω(n + k) + φ) cos(ωn + φ)dφ

= a2

∫ 2π 0

( cos(ω(2n + k) + 2φ)

2 +

cos(2πk)

2

) dφ

= a2

( 0 +

2π cos(2πk)

2

) = a2

2 cos(2πk)

So the process Zn is WSS with autocorrelation function

rz[k] = a2

2 cos(2πk) + rw[k]

We note that Zn is a WSS process.

F-5