By Debadrita Ghosh

Roll number: 07(MTech , Part 1)

 TRAFFIC THEORY
Random variables:
A random variable is a function defined by the characteristics of an arbitrary
random phenomenon. Random variables can be defined into two types:
• Discrete random variables
• Continuous random variables.
If the random variable is a continuous variable then an associated probability
density function (pdf) is defined.
A discrete random variable has either an associated probability distribution or
probability mass function (pmf), which reflects the behavioural characteristics
of the variable at discrete times.

Discrete Random Variables:

For a discrete random variable, X, the pmf P(k) of X is the probability that the
random variable X is equal to k, and is defined by the following function:
P(k) = P(X=k) of k = 0,1,2,….. (1)
It must satisfy the following conditions:
I. 0<=P(k)=1 for every k.
II. ∑P(k)=1 for all k.
Continuous Random Variable:
If a random variable can take an infinite number of values, it is called as
continuous random variables. For a continuous random variable, X, the pdf fx(x)
is a non-negative valued function defined on the whole set of real numbers (-∞,
+∞) such that for any subset s € (-∞, +∞).
Probability of (y€s) = ∫ 𝐹𝑥 (x) dx, where ‘x’ is simply a variable of integral. It must
satisfy the following conditions:
• Fx(x)≥0, for all x.
∞
• ∫−∞ 𝑓𝑥 (𝑥)𝑑𝑥 = 1.

Cumulative Distribution Function:

For all discrete (or continuous) random variable, a cumulative distribution
function (CDF) is represented by P (k) or Fx(x) indicating the probability that the
random variable X is less than or equal to k (or X), for every values k (or X).
Formally, the CDF is defined to be

1
P(k) = P (X≤k) for all k ------------ (3)
Fx(x) = P ( X≤ x) for -∞<x<∞ -----(4)
For a discrete random variable, the CDF is found by summing the probabilities
as follows:
P(k) = ∑ P (x=k) for all ≤ k --------(5)
For a continuous random variable, the CDF is the integral of its PDF, that is
∞
Fx(x) = ∫−∞ 𝑓𝑋 (𝑥)𝑑𝑥 --------------- (6)
Since, Fx(x) = P(X ≤ x), we have
𝑏
FX(a≤x≤b) = ∫𝑎 𝑓𝑋 (𝑥)𝑑𝑥 = FX(b) – FX(a) = P (a<x<b) -------- (7)

Probability Density Function:

The PDF of a continuous random variable is a function that can be integrated to
obtain the probability that the random variable takes a value in a given interval.
Formally, the PDF,
FX(x) of a continuous random variable x, is a derivative of the CDF,
𝑑
Fx(x) = (Fx(x)) -------------- (8)
𝑑𝑥

Discrete Random Variable:

• Poisson’s Distribution:
A Poisson’s random variable is a measure of the number of events that
occur in a certain time interval.
The probability distribution of having ‘k’ events is:
P(X=k) = (ƛke-ƛ)/k! , k=0,1,2………. (9)
The Poisson’s distribution has the expected value
E(X) = ƛ. and variance var(x) = ƛ.

• Geometric Distribution:
A geometric random variable indicates the number of trial required to
obtain the first success. The probability distribution of random variable ‘X’
is given by
P(X=k) = þ(1-P)k-1, k= 0,1,2…. -----(10)
Where þ is the success probability.
This geometric distribution has expected value
E(X) = 1/1-þ and variance, var(x)= þ/(1-þ)2

• Binomial Distribution:

2
 A binomial random variable represents the presence of k and only k, out
of n items and the number of success in a series of trials.
The probability distribution of random variable ‘X’ is
P(X=k) = (𝑛𝑘) þk (1-þ)n-k ----------- (11)
Where k = 0,1,2,….p is a success probability and (𝑛𝑘) = n!/k! (n-k)!
The binomial distribution has expected value
E(X)=np and var(x) = nþ(1-P)

Continuous Random Variable:

• Normal Distribution:
A normal random variable should be capable of assuming any real value,
though this requirement is obtained waived in actual practice. The PDF of
random variable ‘X’ is given by
2
FX(x) = (1/√2𝜋𝜎 )𝑒 −(𝑥−𝜇) /2𝜎2 for -∞<x<∞ -------- (12)
And the CDF can be obtained by
∞ 2
FX(x) = (1/√2𝜋𝜎) ∫−∞ 𝑒 −(𝑦−𝜇) /2𝜎2 dy ---------------- (13)
Where µ is the expected value and 𝜎 2 is the variance of the random
variable X.
Usually, we denote, X~ N(µ 𝜎2) integrating X as a normal random variable
with expected value µ and variance 𝜎2.
The case where µ=0 and 𝜎 =1 is called the standard normal distribution.

Uniform Distribution:
The value of a uniform random variable are uniformly distributed over an
interval. A continuous random variable X, is said to follow a uniform distribution
with parameters ‘a’ and ‘b’ if it’s PDF is constant within a finite interval [a,b] and
0 outside this interval (with ‘a ‘less than or equal to ‘b’). the probability density
distribution of random variable X is,
1
𝑓𝑜𝑟 𝑎 ≤ 𝑥 ≤ 𝑏
fX(x) = {𝑏−𝑎 ------------------------- (14)
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
and the PDF is
0 𝑓𝑜𝑟 𝑥 < 𝑎
𝑥−𝑎
FX(x) = {𝑏−𝑎 𝑓𝑜𝑟 𝑎 ≤ 𝑥 ≤ 𝑏 ------------------------- (15)
1 𝑓𝑜𝑟 𝑏 < 𝑥
The uniform distribution has expected value

3
 E [x] = (a+b)/2 and
Variance var(x) = (b-a)2/12

Exponential Distribution:
The exponential distribution is a very commonly used distribution in
engineering. Due to its simplicity, it has been widely employed even in cases
where it may not be applicable. The PDF of random variable X is given by
0 𝑓𝑜𝑟 𝑥 < 0
fx(x)={ −ƛ𝑥 ---------------------(16)
ƛ𝑒 𝑓𝑜𝑟 0 ≤ 𝑥 ≤ ∞
and the PDF is
0 𝑓𝑜𝑟 𝑥 < 0
EX(x) = {
1 − 𝑒 −ƛ𝑥 𝑓𝑜𝑟 0 ≤ 𝑥 < ∞
Where ƛ is the rate. The exponential distribution has expected value
E[X] = 1/ ƛ and variance var(X) =1/ ƛ2

Poisson’s Arrival Model:

A Poisson’s process is a sequence of events randomly spread in time. For
example, customers arrived in at a bank counter, clicks a similar two packets
arriving to a buffer. Similarly, in wireless networks, the sequence of calls being
initiated in a cell, is usually identified as a Poisson’s process. The rate ‘ƛ′ of a
Poisson’s process is the average number of events per unit time(over a long
time).
• Properties of a Poisson’s process:
For a time interval, [0,t] the probability of n arrivals in ‘t’ units of time is
Pm(t) = (ƛ𝑡 𝑚 )/n! 𝑒 −ƛ𝑡 ------------(18)
For two disjoint (non-overlapping) intervals (t1,t2) and (t3,t4),(that is
t1<t2<t3<t4), the number of arrivals in (t1,t2) is considered independent of
the number of arrivals in (t3,t4). For example in wireless networks, the
number of calls initiated within the time (t1,t2) may be independent of the
calls (t3,t4).
• Inter-arrival times of a Poisson’s process:
We pick an arbitrary starting point tt in time. Let T1 be the time until the
next arrival we have.
P(T1 >t) = P0 (t) = 𝑒 −ƛ𝑡 ---------------- (19)
Thus, the distribution function of T1 is given by:
P T1 (t) = P(T1 ≤ t) = 1-𝑒 −ƛ𝑡 -----------(20)

4
 And the PDF of T1 is:
P ‘T1 (t) = ƛ 𝑒 −ƛ𝑡
f T1 (t) = ƛ 𝑒 −ƛ𝑡 ---------------------------(21)
Therefore, T1 has an exponential distribution with min rate ‘ƛ’.
Let T2 be the time between first and second call arrivals. We can show that
P(T2 > T1 + t | T1 = ∆) =𝑒 −ƛ𝑡 , for ∆, 𝑡 > 0 --------(22)
Thus, the distribution function of T2 is given by
f T2 (t) = P(T2 ≤ T1 + t | T1 = ∆) =1 − 𝑒 −ƛ𝑡 ----------(23)
PDF of T2:
f ‘T2 (t) = ƛ 𝑒 −ƛ𝑡 ---------------------------(24)

Kendall’s Notation:
D.G. Kendall in 1951 proposed a standard notation for classifying system into
different types. The systems are described by the notation: A/B/C/D/E
Where,
A= distribution of inter arrival times of customers.
B= distribution of service times.
C= number of servers.
D= maximum number of customers in system.
E= calling population size.
A and B can take any of the following distribution types:
M= exponential distribution (Markovian)
D= degenerate(all deterministic) distribution.
EK=Erland distribution(k= shape parameter)
G=General distribution(arbitrary distribution)
HK=hyper exponential with parameter k.
Little’s Law:
Assuming a queueing environment operating on a stable steady state where all
initial transient has vanished, the key parameters characterising the system are
as follows:
ƛ= the mean steady state customer arrival rate.
N=the average number of customers in the system [ both in the traffic
and service].
T=The mean time spend by each customer in the system(time spent in the
queue+ service time)
It is intuitive to guess that N= ƛT. ----------------(25)

5
Markov Process:
A Markov process is the one in which the next state of the process depends only
on the present state, irrespective of any previous states taken by the process.
Birth-Death Process:
This is a special type of Markov process often used to model a population. If at
the same time, the population has the ‘m’ entities, then birth of another entity
caused the state to sense to ‘m+1’. On the other hand, a death would cause, the
state to change to ‘m-1’. Thus, we see that in any state, transition can be made
only to one of the two neighbouring states.
Similar arguments can be given to the number of calls in a cell of wireless
network. If a cell has calls being serviced by ‘n’ channels, then given the
probability of a new call being initiated and a call being completed. The
transition to servicing ‘m-1’ calls or ‘m+1’ calls, can be represented with
appropriate transition probabilities.

Figure 1: The state transition diagram of the continuous birth-death process.

In the state n, we have,
ƛn-1P(n-1) + µn+1 P(n+1) = ƛn Pn + µn Pn = (ƛn + µn) Pn ------------(26)
Where p(i) is the steady state probability of state ‘i’.
ƛi (i= 0,1,2,…)is the average arrival rate, and µi (i=1,2,…) is the average service
rate. A similar state equation can be written for states 0,1,2,3.
For state 0 we have,
ƛ0 P0 = ƛ1 P1 -------------------------------------------------------------- (27)
It should be noted that P0, P1, P2,…….. Pn are all steady state probabilities and the
equation also represents steady state equations. Solving the steady state of
equations obtained, we derive the relation between Pn and P0.
µ1 P1 = ƛ0 P0
P1= ƛ0/ µ1 (P0)
P2= ƛ1/ µ2 (P1) = ƛ1/ µ2 . ƛ0/ µ1 (P0)
Pn = ƛ0 ƛ1………. ƛn-1/ µ1 µ2………… µn (P0) ----------------------------------- (28)

M/M/1/∞ queueing system:

6
Here we deal with the simplest queueing system. This is called the M/M/1/∞
queue or M/M/1 queue.
When a customer arrives, in the system it will be served if the server is free,
otherwise the customer is queued. In an M/M/1/∞ queueing system customer
arrives according to the Poisson’s distribution and completes the service in a
FIFO or FCFS manner. The service time are independent, identically
distributed(IID) random variables. The common distribution being exponential.
In practice the M/M/1 queueing system is useful because many complex system
can be abstracted as a composition of a simple M/M/1 queueing system.
Theoretically, the M/M/1 queueing system has an accurate mathematical
solution in the terms of the mean arrival rate, ‘ƛ’, and the mean service rate ‘µ’.
Next, we give an analytical approach to have the M/M/1 queueing system.
Based on the preceding assumptions, M/M/1 queueing systems consists of a
birth- death process.
Let ‘i’ (i=0,1,2,….) be the number of customers in the system and let P(i) be the
steady state probability of the system having ‘i’ customers. For the wireless
networks the M in M/M/1 represents the inter arrival and service times of calls
in a cell and 1 indicates the single cell channel available in the cell. The state of
the Markov Model indicated the number of calls in progress within a cell.
Therefore, the state transition diagram of system is as shown in figure.

Figure 2: state transition diagram of the M/M/1/∞ queueing system.

From the state transition diagram, the equilibrium state equations are given by:
ƛ P0 = µP1 for i=0
(ƛ+µ)Pi= ƛ P(i-1) + µ P(i+1) , for i≥ 1 ------------ (29)
Thus we have,
P1= ƛ/ µ (P0) = 𝜌 P0
P2= ƛ/ µ (P1) = ƛ/ µ. ƛ/ µ P0 = (ƛ/ µ)2 P0 = 𝜌2 P0
P3=𝜌3 P0
.

Pƛ=𝜌𝑖 P0 ------------------------------------------------ (30)

Where 𝜌 = ƛ/ µ, it is called traffic intensity.
7
The normalized condition is given by
∑∞𝑖=0 𝑃𝑖 = 1 ------------------------------------------- (31a)
From the preceding equations we have,
∑∞ 𝑖 1 2
𝑖=0 𝜌 = 1+ 𝜌 +𝜌 +…… ∞ = 1/(1-𝜌)
=> ∑∞ 𝑖
𝑖=0 𝜌 P0 =1/(1-𝜌). P0 = P0/ (1-𝜌) ------------- (31b)
We know ∑∞ 𝑖 ∞ 𝑖
𝑖=0 𝑃 = 1 = ∑𝑖=0 𝜌 P0 => P0/ (1-𝜌) = 1 => P0 = (1-𝜌) ------- (32)
We know that P0 is the probability of the server being free. Since, P0>0, the
necessary condition of a system being in a steady state is 𝜌= ƛ/ µ <1, that is the
arrival rate cannot be more than service rate : otherwise the queue length will
increase to infinity and jobs will experience infinite waiting time. Therefore,
𝜌 = (1-P0) is the probability of server being busy.
From the equation, Pi=𝜌𝑖 P0 we have,
Pi=𝜌𝑖 (1-𝜌) ---------------------------------- (33)
We know that the equation Pi=𝜌𝑖 (1-𝜌) is a geometric distribution.
According to the probabilities, Pi the average number of customers in the system
is
Ls= ∑∞ ∞ 𝑖 ∞ 𝑖
𝑖=0 𝑖. 𝑃𝑖 = ∑𝑖=0 𝑖. 𝜌 𝑃0 = ∑𝑖=0 𝑖. 𝜌 (1 − 𝜌)
= 𝜌(1 − 𝜌) ∑∞ 𝑖=0 𝑖. 𝜌
𝑖−1

= 𝜌(1 − 𝜌)[1+2𝜌+ 3𝜌2 +……….∞]

= 𝜌(1 − 𝜌)[( 1+𝜌+ 𝜌2 +……….∞) + 𝜌( 1+𝜌+ 𝜌2 +……….∞)+
𝜌2 ( 1+𝜌+ 𝜌2 +……….∞)
= 𝜌(1 − 𝜌)[( 1+𝜌+ 𝜌2 +……….∞) ( 1+𝜌+ 𝜌2 +……….∞)]
= 𝜌(1 − 𝜌)[( 1 + 𝜌 + 𝜌2 + ⋯ … … . ∞)2 )]
= 𝜌(1 − 𝜌).1/(1 − 𝜌)2
= 𝜌/(1 − 𝜌)
= (ƛ/ µ)/(1- ƛ/ µ)
=ƛ/ (𝜇 − ƛ) ------------------------- (33b)
Ls = ƛ/ (𝜇 − ƛ)
Using Little’s Law, the average dwell time of a customer in the cell of a wireless
system is given by:
Ws = Ls / ƛ
= (ƛ/ (𝜇 − ƛ))/ ƛ
= 1/(𝜇 − ƛ) ------------------------- (34)
The average queueing length:
Lq = ∑∞ 𝑖=1(𝑖 − 1). 𝑃𝑖
= ∑∞ 𝑖
𝑖=1 (𝑖 − 1). 𝜌 𝑃0
= 𝜌(1 − 𝜌) ∑∞ 𝑖=2(𝑖 − 1). 𝜌
𝑖−1

8
 = 𝜌(1 − 𝜌)[ 𝜌 +2𝜌2 + 3𝜌3 +……….∞]
= 𝜌(1 − 𝜌)𝜌[ 1 +2𝜌1 + 3𝜌2 +……….∞]
= 𝜌2 (1 − 𝜌) [ 1 + 𝜌1 + 𝜌2 + ⋯ … … . ∞]2
= 𝜌2 (1 − 𝜌) 1/(1 − 𝜌)2
=𝜌2 /(1 − 𝜌)
= (ƛ/𝜇)2 /(1- ƛ/ µ)
= (ƛ2 × 𝜇)/𝜇2 (𝜇 − ƛ)
= ƛ2/(𝜇(𝜇 − ƛ)---------------------------- (35)
The average waiting time of customer is given by:
Wq = Lq / ƛ
= ƛ2/(𝜇(𝜇 − ƛ).ƛ
= ƛ/(𝜇(𝜇 − ƛ) ------------------------------ (36)

M/M/S/∞ Queueing System:

We consider a queueing system with arrival rate ƛ, but we assume that there
are multiple server S (≥ 1). This, each one with service rate µ, and they all shared
a common queue, i(i=0,1,2,…..) is the number of customer in the system. Let P(i)
is the steady state probability in the system having i customer, therefore, the
state transition diagram of the system is given below.

Figure 3: state transition diagram M/M/S/∞ queueing system.

From the state transition diagram, the equilibrium state equation is given by
ƛ P0 = µP1
(ƛ+iµ)Pi= ƛ P(i-1) + (i+1)µ P(i+1) , for 1≤i≤S
(ƛ+µS)Pi= ƛ P(i-1) + Sµ P(i+1) , for S≤i ---------------------- (37)
Thus we have,
Pi = 𝛼 𝑖 /i! P0 for i<s
𝑆 𝑖−𝑆
Pi = 𝛼 /S! (𝛼/𝑆) P0 for S≤i
Where 𝛼 = ƛ/𝜇
∑∞ 𝑆−1 𝑖 𝑆 ∞ 𝑖
𝑖=0 𝑃𝑖 = [∑𝑖=0 𝛼 /i! + 𝛼 /S! ∑𝑖=0(𝛼/𝑆) ] P0 = 1 ------ (38)
𝛼𝑖
=> P0 = [∑𝑆−1
𝑖=0 + 𝛼 𝑆 /S! ∑∞ 𝑖 −1
𝑖=0(𝛼/𝑆) ] ---------------- (39)
i!
If 𝛼 < 𝑆, we have,
∑∞ 𝑖
𝑖=0(𝛼/𝑆) = S/(S-𝛼) -------------------------------------------- (40)

9
This becomes,
P0 = [∑𝑆−1 𝑖 𝑆
𝑖=0 𝛼 /i! + 𝛼 /S! S/(S − 𝛼) ]
−1

𝛼𝑖 𝛼𝑆
= [∑𝑆−1
𝑖=0 + 1/(1 − 𝜌) ]−1 ------------------------ (41)
i! S!
Where 𝜌(= 𝛼/𝑆 = ƛ/𝜇𝑆) is called utilization factor.
Note that, for the queue to be stable we should have 𝜌 < 1.
According to the probabilities Pi’s the average number of customer in a system
is
Ls = ∑∞ 𝑆 2
𝑖=0 𝑖. 𝑃𝑖 = (𝛼+ 𝜌 𝛼 𝑃0 ) / S! (1 − 𝜌) ----------------- (42)
Using Little’s Law, the average dwell time of a customer in the system is given
by:
Ws = Ls / ƛ = 1/𝜇 + (𝛼 𝑆 𝑃0 ) / S.𝜇. S! (1 − 𝜌)2 ---------------- (43)
The average queueing length:
Lq = ∑∞𝑖=𝑆(𝑖 − 𝑆). 𝑃𝑖
= (𝛼 𝑆+1 𝑃0 ) / (S-1)! (S − α)2 ---------------------------- (44)
The average waiting time of customer is given by:
Wq = Lq / ƛ
= (𝛼 𝑆 𝑃0 ) / S.𝜇. S! (1 − 𝜌)2 ------------------------------- (45)

Problems:
Poisson’s Distribution:
1.A life insurance salesman sales on the average 3 life insurance policies/ week.
Use Poisson’s law to calculate the probability that in a given week he will sell:
a. Some policies.
b. Two or more policies but less than 5 policy
c. Assume that there are 5 working day/week, what is the probability that in
a given day he will sell 1 policy.

Solution:
Here 𝜇 = 3.
a. Some policies means one or more policies. We can work this out by
finding [1- (the 0 policy probability)] => P(X>0) = 1- P(𝑥0 )
Now, P(X) = (𝜇 𝑥 𝑒 −𝜇 )/x!
So, P(𝑥0 ) = (30 𝑒 −3 )/0! = 𝑒 −3 = 4. 9787 × 10−2
Therefore probability of one or more policies is given by:

10
 P(X>0) = 1 - 4. 9787 × 10−2 = 1- P(𝑥0 ) = 0.95021.

b. The probability of selling 2 or more policies but less than 5 policies:

P(2≤x≤5) = P(𝑥2 ) + P(𝑥3 ) + P(𝑥4 )
= (32 𝑒 −3 )/2! + (33 𝑒 −3 )/3! + (34 𝑒 −3 )/4!
= 0.61611.

c. The average number of policies sold per day 3/5=0.6.so, on a given day
:
P(X) = (0.61 𝑒 −0.6 )/1! = 0.32929.

2. 20 sheets of aluminium alloy were examined for surface flaws. The frequency
of the number of sheets with a given number of flaws per sheet was as follows:
Number frequency
of flaws
0 4
1 3
2 5
3 2
4 4
5 1
6 1
What is the probability to find a sheet chosen at random which contains 3 or
more surface flaws?
Solution:
Total number of flaws is given by:
(0×4) + (1×3) + (2×5) + (3×2) + (4×4) + (1×5) + (6×1) = 46.
So the average number of flaws for the 20 sheets is 𝜇 = 46/20 = 2.3.
Required probability is:
P(X≥3) = 1 – [P(𝑥0 ) + P(𝑥1 ) + P(𝑥2 )]
= 1 – [(2.30 𝑒 −2.3 )/0! + (2.31 𝑒 −2.3 )/1! + (2.32 𝑒 −2.3 )/2!]
= 0.40396.
3. If electricity power failures occur accordingly to a Poisson’s distribution with
an average of 3 failure every 20 weeks. Calculate the probability that there will
not be more than 1 failure during a particular week.
Solution:

11
 The average number of failures per week is: 𝜇 = 3/20 = 0.15.
Not more than one failure means probability for 0 failure and 1 failure:
P(𝑥0 ) + P(𝑥1 ) = (0.150 𝑒 −0.15 )/0! + (0.151 𝑒 −0.15 )/1!
= 0.98981.

4.Vehicles passes through a junction on a busy road at an average rate of 300/hr.

a. find the probability no one passes in a given minute.
b. what is the expected number of passing in 2 minutes.
c. find the probability that this expected number actually passes through
in a given 2 minutes period.
Solution:
The average number of cars per minute is 𝜇 = 300/60 =5.
a. P(𝑥0 ) = (50 𝑒 −5 )/0! = 6.7379 × 10−3
b. Expected number of vehicles each passing in 2 minutes:
E(X) = 2×5 = 10.
c. Now, 𝜇 = 10.
So, P(𝑥10 ) = (1010 𝑒 −10 )/10! = 0.12511

5. A company makes electric motors. The probability of an electric motor is

defective is 0.01. What is the probability that a sample of 300 electric motors
will contain exactly 5 defective motors?
Solution:
The average number of defectives in 300 motor is:
𝜇 = 0.01×300 = 3.
The probability of getting 5 defectives are:
P(X) = (35 𝑒 −3 )/5! = 0.10082

Binomial Distribution:
1. A die is tossed 3 times. What is the probability
a. Number 5 is turning up.
b. 1-5 turning up
c. 3-5 turning up
Solution:
This is a binomial distribution because there are only 2 possible outcomes
( we get 5 or not) . now n=3 for each part.

12
 Let X = number of 5’s appearing.
a. Here X =0, probability
P(X=0) = (𝑛𝑥) px qn-x
= (30) (1/6)0 (5/6)3-0
= 125/216
= 0.5787
b. Here X = 1, probability
P(X=1) = (31) (1/6)1 (5/6)3-1
= 75/216
= 0.34722
c. X=3, probability
P(X=1) = (33) (1/6)3 (5/6)3-3
= 1/216
= 4.6296× 10−3

2. Hospital records show that a patient suffering from certain disease. 75%
die of it. What is the probability that of 6 randomly selected patient 4 will
recover?

Solution:
P(X=4) = (64) (0.25)4 (0.75)2 = 0.0329595.

3. In the old days there was a probability of 0.8 success in any attend to make
a telephone call. (this often depends on the importance of the person
making the call or the operator’s curiosity). Calculate the probability of
having 7 success in 10 attempts.

Solution:
Let X be the success in getting through probability of 7 successive
in 10 attempts.
P(X=7) = (10
7
) (0.8)7 (0.2)3 = 0.20133.

4. A (blind folded) marksman finds that only average he hits the target 4
times, out of 5. If he fires 4 shots, what is the probability of
a. More than 2 hits.
b. At least 3 misses.

Solution:
a. Let X= number of hits
13
 𝑋0 = no hits, 𝑋1 = 1 hits, 𝑋2 = 2 hits etc.
P(X) = P(𝑋3 ) + P(𝑋4 )
= (43) (0.8)3 (0.2)1 + (44) (0.8)4 (0.2)0
= 0.8192

b. P(X) = P(𝑋1 ) + P(𝑋0 )

= (41) (0.8)1 (0.2)3 + (40) (0.8)0 (0.2)4
= 0.0272

5. The ratio of boys to girls at birth in Singapore is quite high at 1.09:1. What
proportion of Singapore families with exactly 6 children will have at least
3 boys? (Ignore the probability of multiple birth).

Solution:
Probability of getting a boy is : 1.09/1.09+1.00 = 0.5215
Let X be the number of boys in the family.
Here n=6, p = 0.5215, q= 1- 0.5215 = 0.4785
P(X=3) = (63) (0.5215)3 (0.4785)3 = 0.31097
P(X=4) = (64) (0.5215)4 (0.4785)2 = 0.25402
P(X=5) = (65) (0.5215)5 (0.4785)1 = 0.11074
P(X=6) = (66) (0.5215)6 (0.4785)0 = 2.0115× 10−2
P(X≥3) = 0.4785 + 0.31097 + 0.25402 + 0.11074 + 2.0115× 10−2 = 0.69565
6. A manufacturer of a metal pistons finds that on average 12 % of his
pistons are rejected because they are either oversized or undersized.
What is the probability that a batch of 10 pistons will contain
a. No more than 2 rejects
b. At least 2 rejects

Solution:
Here n= 10, p= 0.12, q= 0.88
a. No reject that is, X= 0
P(X=0) = (10
0
) (0.12)0 (0.88)10 = 0.2785
One reject, X= 1
P(X=1) = (10
1
) (0.12)1 (0.88)9 = 0.3797
Two rejects, X= 2
P(X=2) = (10
2
) (0.12)2 (0.88)8 = 0.23304

14
 Probability of getting not more than 2 rejects are:
P(X≤2) = 0.2785+0.3797+0.23304 = 0.89131

b. Probability of at least 2 rejects:

= 1- P(X≤1)
= 1- [P(X=0) + P(X=1)]
= 1- [0.2785+0.3797]
= 0.34173.

15
CELLULAR CONCEPT:
Cell area: A cell is in radio area covered by a transmitting section, or a BS. All
MSS in that area are connected and serviced by the BS. Therefore, ideally, the
area covered by a cell could be represented by a circular cell, with a radius R,
from the center of the DS. There are many factors that cause reflections and
refractions of the signals including elevation of the terrain, presence of a hill or
valley or a tall building, and presence of particles in the air. The actual shape of
the cell is determined by the required signal strength in the surrounding area.
Therefore, the coverage area may be a little distorted.
There are many possible models that can be used to represent a cell boundary
and the most popular alternative of hexagon, square and equilateral triangle. In
most modelling and simulation, hexagons are used. While, a square is employed
as the second choice.
SIGNAL STRENGTH AND CELL PARAMETERS:
Cellular systems depends on the radio signals received by an MS, throughout the
cell and only contours of signal strength emanating from the BSS of the two
adjacent cells, i and j, as in the figure.

Figure 4: Handoff region.

It is clear that signal strength goes down as one moves away from BS. As the MS
moves away from the BS of the cell, the signal strength weakens and at the same
point a phenomenon known as ‘Handoff’ occurs. MS moves away from cell ‘i’
and gets closer to cell ‘j’. Assuming that Pi(x) and Pj(x) represents the power
received at MS from BSi and BSj. in the figure at a distance x1, the received signal
from BSj is close to 0 and the signal strength at the MS could be primarily
attributed to BSi is negligible. To receive and interpret the signals correctly at
the MS, the radio received signals must be at a given minimum power level PMin,

16
and distance x3 and x4 represent two such points from BSj and BSi respectively.
This means that between the points x3 and x4 the MS can be served by either BSi
or BSj, and the choice is left to the service provider at the underlined technology.
If the MS has a radio link with BSi and is continuously moving away towards BSj
, then at some point it has to be connected to the BSj radio link and the change
of such linkage from BSi to BSj is known as handoff. Therefore, x3 to x4 indicates
the handoff area, where the perform handoff depends on many factors. One
option is to do handoff at x5, where 2BSS have equal signal strength [ping-pong
position]. A critical condition is that, the handoff should not take place too
quickly to make the MS, changes to BSi and BSj . too frequently if the MS, moves
back and forth between two cells area due to underlined terrain, or intentional
movements. To avoid such a ‘ping-pong’ effect, the MS is allowed to continue
maintaining a radiolink with the current BSi until the signal strength from BSj
exceeds that of BSi by some prespecified threshold value ‘E’ as shown in the
point XTH, thus besides transmitting power, the handoff also depends on the
mobility of the MS.
Another factor that influences handoff is the area and the shape of the cell. An
ideal situation is to have the cell configuration match the velocity of the MSS
and to have a larger boundary where the handoff rate is minimal. The mobility
of an illegal MS is difficult to predict, with each MS have a different mobility
pattern hence it is impossible to have an exact match between the cell shape
and subscriber mobility. For example: consider a rectangular cell of area ‘A’ and
sides ‘R1’ and ‘R2’ shown in figure assuming that X1 is the number of MSS having
handoff per unit length in the horizontal direction, and X2 is the similar quantity
in the vertical direction, then the total handoff rate is given by the equation:
X1 → X1cos𝜃 (R1) , X1sin𝜃(R2) ----------------- (i)
X2 → X2cos𝜃(R2) , X2sin𝜃(R1) ------------------(ii)
ƛH = R1 (X1cos𝜃 + X2sin𝜃) + R2 (X1sin𝜃 + X2cos𝜃)
Assuming that area A= R1× R2, is fixed, the equation is how to minimize ƛH for a
given 𝜃, this is done by substituting the value of
A= R1× R2
 R2 = A/ R1
Now, differentiating with respect to R1 and equating it to 0, which gives us
ƛH = R1 (X1cos𝜃 + X2sin𝜃) + A/ R1 (X1sin𝜃 + X2cos𝜃)
17
R12 = A(X1sin𝜃 + X2cos𝜃)/ (X1cos𝜃 + X2sin𝜃)------(iii)
R22 = A(X1cos𝜃 + X2sin𝜃)/ (X1sin𝜃 + X2cos𝜃)-------(iv)

ƛH = 2√A(𝑥1 sin𝜃 + 𝑥2 cos𝜃 )(𝑥1 cos𝜃 + 𝑥2 sin𝜃)------ (v)

 ƛH = 2√A(𝑥1 𝑥2 + cos𝜃sin𝜃(𝑥12 + 𝑥22 ) -----(vi)

equation (vi) is minimized when 𝜃 =0. Here from equation (v), (ii) and (iii) we
get, ƛH = 2√A(𝑥1 𝑥2 ) --------- (vii)
and R2/ R1 = X1/X2 ----------(viii)
 R2/ R1 = 1/ (X2/X1)
 R ∝ 1/X [let R2/ R1 = R, X2/X1 = X]

CAPACITY OF A CELL:
The load of a cell is typically characterized by following two important random
parameters:
(i) The average number of MSS requesting the service (average cell arrival
rate) ƛ.
(ii) The average length of time the MS requires the service (average
holding time) T.
The traffic load is defined as α = ƛT ------(ix)
For example, in a cell with 100 MS, for an average, if 30 requests are generated
during an hour, with average holding time T= 360 secs, then the request rate is:
ƛ = 30 request/3600 secs -----------------------(x)
a serving request that is kept busy for an hour is quantitatively defined as 1
Erlang. Hence, the over traffic load for the preceding example by Erlang is given
by
α = ƛT
 (30/3600) × 360
 3 Earlang. -----------------------------------------(xi)
The average arrival rate during a short interval ∆ƛ, assuming Poisson Distribution
of service, the probability P(n,t) for n cells to arrive in an interval of length ‘t’ is
given by

18
P(n,t) = (ƛ𝑡 𝑛 𝑒 −ƛ𝑡 ) /n! ----------------------------------(xii)
Assuming 𝜇 to be the service(departure) rate, the probability of each call to
terminate during the interval ‘t’ is given by 𝜇t. Hence, the probability that a given
call requires service for time ‘t’ or less, is given by:
S(t) = 1- 𝑒 −𝜇𝑡 -----------------------------------------------(xiii)
Then the probability that an arriving call is blocked is given by:
B(S,a) = (𝑎 𝑠 /𝑆! ÷ ∑𝑠𝑘=0 𝑎𝑘 /k!) --------------------------(xiv)
where, ‘S’ is the number of channels in a cell. B(S,a) is also called blocking
probability, probability loss or probability of rejection. Equation (xiii) is called
the Erlang ‘B’ formula. If ‘S’ is given as 2 with a=3, the blocking probability
B(2,3) = (32 /2! ÷ ∑2𝑘=0 3𝑘 /k!) ------------------------(xv)
= 0.6
Therefore, a fraction of 0.6 calls is blocked, and we need to reinitiate the call.
Thus, the total number of blocked calls is
= 30×0.6 = 18
Efficiency = traffic no. blocked/ capacity --------- (xvi) [output/input]
= (Erlang × portion of nonchannel traffic)/No. of channels
= 3(1-0.6)/2 =0.6 -------------------------------- (xvii)
The probability of an arriving call being delayed is
C(S,a) = [(𝑎 𝑠 /(S-1)! (S-a) ÷ 𝑎 𝑠 /(S-1)! (S-a) + ∑𝑆−1 𝑖
𝑖=0 𝑎 /i!)] ---- (xix)

= (SB(S,a))/(S- a[1- B(S,a)] --------------------- (xx)

This is called the Earlang ‘C’ formula.
If S=5 and a=3, we have;
B(5,3) = 0.11
Therefore, the probability of an arriving call delayed is
C(S,a) = (SB(S,a))/(S- a[1- B(S,a)]
C(5,3) = (5×0.11)/(5- 3[1 – 0.11] = 0.2360
FREQUENCY REUSE:
19
The same frequency band or channel used in a cell can be ‘reused’ in another
cell as long as the cells are far apart and the signal strength do not interfere with
each other. The distance between two cells using the same channel is known as
‘reused distance’ and is represented by D. In fact, there is a close relationship
between D and R(the radius of the cell) and N(the number of cells in a cluster),
which is given by

D= √3𝑁R ------------------------------- (xxi)

Therefore, the reuse factor is D/R = √3𝑁 -----(xxii)

The number of cells ‘N’ per cluster is given by
N= 𝑖 2 + 𝑗 2 + ij
Here, i represents the number of cells to be traversed along with the direction
of ‘i’ starting from the center of a cell and ‘j’ represents the number of cells in a
direction 60°to the direction of ‘i’. Substituting different values of i and j to the
equation N becomes, 1,3,4,7,9,12,14,16,19,21,28,…….; the most popular values
are 7 and 4.
If i=0 and j=1, then N=1.

7 cell clusters
Area = 3√3/2 𝑅 2 = 2.598 𝑅 2

20
Center to center distance between closest co-channel (interfering cells).

D = √(𝑖√3 𝑅)2 + √(𝑗√3 𝑅)2 - √2(i√3𝑅)( j√3𝑅) cos 120°

 R√3(𝑖 2+𝑗 2 +ij)
 R√3𝑁
Prove that D= R√3𝑁
It is also called reused distance.
COCHANNEL INTERFERENCE:
There are many cells using the cell frequency band. All the cell using the same channel are
physically located apart by atleast reuse distance. Even though the power level is controlled
carefully, so that such co-channel do not create a problem for each other. There is still some
degree of interference due to non-zero signal strength of such cell. In the cellular system with
the seven cells, there will be six cells using co-channel at a reuse distance. The second tire co
channels are at two times reuse distance apart and their effect on the serving BS is negligible.
The co channel interference ratio(CIR) is defined as:
CIR= C/I = carrier/ interference = C/∑𝑀 𝐾=1 𝐼𝐾 --------(xxiii)
Where IK is co channel interference and M is the maximum number of co channel interfering
cells. For the cluster size of 7, M=6, CIR is given by
CIR = C/I = C/∑𝑀𝐾=1(𝐷/𝑅)
𝛾

Where 𝛾 is the propagation path loss and varies 2 and 5.

First time co-channel

base station

D2

D3
D1

M
D6
D4

D D5

Second channel 21 SERVING BASE

base station STATION
MOBILE STATION
Figure 5: Cell with co channel and their forward channel interference on
transmitted signal.
Reuse distance = ∞
Radius of distance = R
Where [D1= D2= D-R],[D3 =D6 =∞] [D4= D5= D+R]
The co-channel interference is the worth case for the forward channel.
Downlink is given by:
C/R = 1/[2(𝑞 − 1)−𝛾 + 2𝑞 −𝛾 + 2(𝑞 + 1)−𝛾 ---------------------(xxv)
D1= D2 = (D-R)/R = D/R -1 = 2(𝑞 − 1)−𝛾
D3= D6 = D/R = D/R -1 = 2𝑞 −𝛾
D4= D5 = (D+R)/R = D/R +1 = 2(𝑞 + 1)−𝛾
Where q= D/R is the frequently reuse factor- to reduce interference technique.
CELL SPLITTING:
We have been considering the same size cell across the board. This implies that
the BS of all the cell transmit information at the same power level so that the
net coverage area for each cell is same as at times they may not be visible and
in general, this may not be desirable. Service providers would like to service user
in a cost effective way and resource demand may depend on the concentration
of users in a given area. Change in number of users could also occur over a period
of time. One way to cope with increased traffic is to split a cell into several
smaller cells. This implies that additional BS used to be established at the cluster
of each new cells that has been added, so that the higher density cells can be
handled effectively. As the coverage area of new split cell is similar. The
transmitting power level are lower and this helps in reducing co channel
interference.
CELL SECTORING:
We have been primarily concentrating on what is known as omni-directional
(360°) antennas, which allow transmission of radio signals with equal power
strength in all directions. It is difficult to design such antennas and most of the
time. An antenna covers an area of 60° or 120°. These are called dimensional
antennas and cell served by them are called sectored cells. For a particular point

22
of view, many sector are mounted on a single microwave tower located at the
center of the cell and adequate non-power antennas was placed to cover the
whole 360° of the cell. For example, the 120° sectored cell requires 3 directional
antennas. In practice, the effect of an omni directional antenna, can be achieved
by employing several directional antennas to cover the whole 360°.
The advantages of sectoring are that it requires coverage of a smaller area by its
antenna and hence, lower power is required to transmit radio signals. It is also
observed that this spectrum efficiency of the overall system is enhanced. It is
found that a quad-sector architecture has higher capacity for 90° area, coverage
than a trisector cells.

The co channel interference for cells using directional antennas can also be
computed and the worst case for the three sector directional antenna can be
obtained as:
CIR = C/I = 1/𝑞 −𝛾 + (𝑞 + 0.7)−𝛾 --------------------------- (xxvi)
The co channel interference ration is the worst case of the six sectors, directional
antenna when 𝛾 = 4 can be given by:
CIR = C/I = 1/(𝑞 + 0.7)−𝛾 = (𝑞 + 0.7)𝛾 = (𝑞 + 0.7)4-------------------------- (xxvii)
Thus we can see that the use of directional antenna is helpful in reducing co-
channel interference.

2
D= √(9/2𝑅)2 + √3/2𝑅 = √21𝑅 = 4.58 R

23
 2
D’ = √(5𝑅)2 + √3𝑅 = √28𝑅 = 5.29 R = D+0.7R

CHANNEL ALLOCATION:
Whenever a MS wants to make a new call, it must submit a request for a channel
to BS. The BS can grant such an access provided that a channel is readily available
and in this way the probability of new call to be blocked or blocking probability
for calls originated in a cell can be minimized. A limited frequency band is
allocated for wireless communication so there is a limit to be maximum number
of channels in each cell. Interference in FDMA/TDMA is avoided by using
appropriate reused distance. One simplistic approach is to divide the channels
equally among the cells. If ‘Stotal’ is the total number of channels and ‘N’ is the
size of reused cluster, then
S= the number of channels per cell = Stotal/N ---------(i)
Looking at this relation, we will think that reducing the value of N to increase
the number of channels per cell which is against the philosophy of frequency
reuse. Any reduction in the reuse distance increased the interference. It is
reasonable to assign an equal no of channel to each cell. An ideal situation would
be when all parameters are assumed to be the same. This means that the
location of MSS over an area is informing this and the probability of each MS
making a call is also assumed to be in same.
STATIC VS DYNAMIC ALLOCATION:
There are two ways traffic channels can be allocated to different cells of a
cellular system that is static and dynamic. In static allocation, a fixed no of
channels are allocated to each cell, while in dynamic allocation implies that
allocation of channels to different cells is done dynamically, as needed. Possibly,
from a central pole. In a static scheme an equal no of channels can be allocated
to each cell. That is fixed channel allocation will be done on the amount of traffic
in different cells.
In brief, generalization schemes can be divided as follows:
(i) Fixed channel allocation (FCA) schemes
(ii) Dynamic channel allocation (DCA) schemes.
(iii) Hybrid channel allocation (HCA) schemes.
FIXED CHANNEL ALLOCATION:

24
In fixed channel allocation, a set of channels are primarily allocated to each cell
of a system. If the total no of available channels in any system is divided into
sets, maximum no of channels sets require to serve the entire coverage area is
related to the frequency reuse distance D and radius R of each cell as follows:

N = D/√3𝑅 ------------------------------------(ii)
One approach to addresses increase the traffic of new organization and handoff
call in a cell is to borrow free channel from neighbouring cell. There is no
interference each cell associated within reuse distance. There are many possible
channel burrowing schemes from simple to complex and they can be selected
based on employed controller software and the visibility of burrowing under
given condition:
(i) Simple burrowing scheme(SBS)
(ii) Complex burrowing scheme(CBS)
Simple burrowing scheme(SBS) : a SBS implies that if all channels allocated to a
cell are being used, then additional channels can be borrowed from any cells
that has some free unused channels. Such a cell is called a donor cell. And
obvious choice is to select a donor from along an adjacent cells that has largest
no of free channels. This is known as borrowing from the richest. Further
consequences is to return the borrowed channels to the donor. If a channel
becomes available in the cell that initially borrowed a channel. Such an algorithm
is defined as basic algorithm in reassignment. Another alternative is to select
first free channel found a borrowing when the search followed a pre-defined
sequences, this is known as borrow first available scheme.
Complex borrowing scheme(CBS): basic strategy of complex scheme is to divide
into two groups: one group assign the each cell permanently and the 2 nd group
kept reserve as donor to borrowed by neighbouring cells. The ratio between the
two groups of channels is determined by a priory and can be based on estimated
traffic in a system. An alternative, known as borrowing with channels ordering
is to assign priorities to all channel of each cell, with highest priority channel
being used in sequential order for local call in the cell while channel borrowing
is done starting from lowest priority channels. As mentioned earlier, every
attempt must be made to minimize interference. Therefore, if channel
borrowing is done such that a particular channel is available in nearly co channel
cells, then the channel can be borrowed such schemes is known as borrowing

25
with directional channel locking. Since this scheme, imposes additional
constraints, the no of channels available is reduced.

The basic sectoring technique can be used to allocate channels temporarily. In

the following section, we look into channels borrowing in such a scenario and
discuss why cell sectoring is useful? How it influences the selection of donor cells
and what kind of impact it has on channel interference. One way of using sector
cells is to share with bias, which implies borrowing a channel from one of the
two adjacent sectors of neighbouring cells. This can be further enhanced by a
scheme known as channel assignment with borrowing and reassignment by
ensuring the following process. Minimum impact on future call blocking
probability in neighbouring cells and reassignment of borrow channel is done to
provide maximum help to the neighbourhood. The channels can even also be
ordered based on which channels provides better performance, this can be
useful in selecting lower order channel for borrowing. In addition, borrowed
channel to be return to donor cells. If the channel becomes available in the
borrowing cells. This scheme is known as “ordered channel assignment scheme
with rearrangement.”
7- adjacent cluster that could have co-channel interference are shown in above
figure. Let us assume that each sector of clusters use the same frequency and or
channels to maintain reuse distance with such an arrangement and fixed

26
distribution of channel to different sectors, interference to be kept to a
minimum desired level. Let us assume that sector ‘X’ of cluster(A3) need borrow
channel from adjacent cells. Let us say from sector A of A1. There could be
potential violation of reuse distance and there could be interference between
the borrowed channels in sector, with the same channels of all ‘a’ channels of
clusters A2,A3,A4,A5,A6,A7, looking at the distance between X and the sector ‘Q’ of
other clusters A5,A6,A7 satisfy the reuse distance requirements, while cluster
A2,A3,A4 violates the reuse distance from ‘x’. clearly, the sector ‘a’ for both cluster
A2,&,A4 are in different direction from X1 and simultaneous use of the same
channel in this area will not cause any additional interference as normally
expected. The only question that needs to be addressed is interference between
sector X and sector A., of cluster A3 and even though the reuse distance is
violated, their direction are such that they would most likely not interfere with
each other. If the cells are not sectored, then in figure borrowed channel will be
marked ‘X’ and would cause interference with the cell (a,b,c) of cluster A2,A3,A4.
This borrowed channel can’t be used in the clusters as well. Therefore, we can
see the obvious advantage of sector cells. Therefore two steps of verifying
potential interference and potential prohibition of their borrowed channel from
other cells are first checking the reuse distance with other closest using those
borrowed channel and 2nd looking at the center direction of all cells not
satisfying the reuse distance.
DYNAMIC CHANNEL ALLOCATION:
Dynamic channel allocation implies that channels are allocated dynamically as
new calls arrive in the system and is achieved by keeping all free channels in a
central pool. In this way, it is fairly straightforward to select the most
appropriate channel for any new call with the aim of minimizing the
interference, as current allocation of different channels and current traffic is
known. In this way, a dynamic channel allocation scheme overcomes the
problem of a fixed channel allocation scheme. In fact, a free channel can be
allocated to any cell, as long as interference constraints in that cell can be
satisfied. The selection of a channel could be very simple and could involve one
or more considerations, including future blocking probability in the vicinity of
the cell, reuse distance, usage frequency of the candidate channel, average
blocking probability of the overall system, and instantaneous channel occupancy
distribution. The control could be centralized or distributed, and accordingly,

27
dynamic channel allocation schemes are classified into two types- centralized
and distributed schemes.
CENTRALIZED DYNAMIC CHANNEL ALLOCATION SCHEMES:
In these schemes, a channel is selected for a new call from a central pool of free
channels, and a specific characterizing function is used to select one among
candidate free channels. The simplest scheme is to select the first available free
channels that can satisfy the reuse distance. An alternative is to pick a free
channel that can minimize the future blocking probability in the neighbourhood
of the cell that needs an additional channel. This is defined as locally optimized
dynamic assignment. Another scheme of channel reuse optimization maximizes
the use of every channel in the system by appropriate allocation of channels,
thereby maximizing system efficiency.
For a given reuse distance, cells can be identified that satisfy minimum reuse
distance and all these cells are defined as co-channel cells. If a cell need to
support a new call, then a free channel from the central pool is selected that
would maximize the number of members in its co-channel set.
Dynamic channel allocation schemes handle new calls being generated
randomly and hence cannot maximize overall channel reuse. Therefore, those
schemes are observed to carry less traffic as compared to fixed channel
allocation, especially for higher traffic rates.
DISTRIBUTED DYNAMIC CHANNEL ALLOCATION SCHEMES:
Centralized schemes can theoretically provide near optimal performance, but
the amount of computation and communication among the BSS lead to
excessive system latencies and make them impractical. Therefore, schemes
have been proposed that involve scattering of channels across a network.
However, centralized schemes are still used as a benchmark to compare various
decentralized schemes.
Distributed dynamic channel allocation schemes are primarily based on one of
the three parameters: co-channel distance, signal strength measurement and
signal- to- noise interference ratio. In a cell-based distribution scheme, a table
indicates if other co-channel cells in the neighbourhood are not using one or
more channels and selects one of the free channels for the requesting cell. In an
adjacent channel interference constraints scheme, in addition to co-channel
interference, adjacent channel interference is taken into account while choosing

28
a new channel. The main limitation of this scheme is that a maximum packing of
channels may not be possible in the MSS location is not taken into account. In a
signal, channels are allocated to a new call if the anticipated CIR is above a
threshold. This could cause the CIR ratio for some existing calls to deteriorate
and hence those would require finding new channel that could satisfy a desired
CIR ratio. Otherwise, those interrupted calls could be dropped prematurely or
may also have a further ripple effect, possibly leading to system instability.
HYBRID CHANNEL ALLOCATION:
Hybrid channel allocation schemes are a combination of fixed and dynamic
channel allocation schemes with the channels divided into fixed and dynamic
sets. This means that each cell is given a fixed number of channels that is
exclusively used by the cell. Request for a channel from the dynamic set is
initiated only when a cell has exhausted using all its channels in the fixed set. A
channel from the dynamic set can be selected by employing any of the dynamic
channel allocation schemes. It has been observed that for a fixed to dynamic
channel ration of 3:1, the hybrid allocation leads to better service than the fixed
scheme for traffic upto 50% and beyond that load, fixed schemes performs
better. Doing a similar comparison, with dynamic schemes, when the load varies
from 15% to 32% to 40%, the corresponding best values vary from most to
medium to no dynamic channels. A lot of time is required if simulation is to
determine the behaviour of a large system, and an analytical approach is
desirable. However, exact analytical models are much more difficult to define
for hybrid schemes and if data traffic also need to be incorporated, it is almost
impossible to have even an approximate model. This is an interesting area that
needs further investigation.
FLEXIBLE CHANNEL ALLOCATION:
The idea behind a flexible channel allocation scheme is similar to a hybrid
scheme having available channels divided into fixed and flexible(emergency)
sets, with fixed sets assigned to each cell to handle lighter loads effectively. The
flexible channels are called for by cells only when additional channels are
needed after exhausting the fixed set. Flexible schemes require centralized
control, with up to date traffic pattern information, to assign flexible channels
effectively. There are two different strategies used in allocating channels:
scheduled and predictive. In scheduled assignment, a priori estimates about
variation in traffic (i.e. peaks in time and space) are needed to schedule
emergency channels at predetermined peaks of traffic change. In a predictive
29
strategy, the traffic intensity and blocking probability is monitored in each cell
all the time that flexible channels can be assigned to each cell according to its
needs. This is similar to allocating additional channel S as needed, rather than
assigning extra channels during the peak hours of 8 a.m. to 5 p.m.
REUSE PARTITIONING BASED CHANNEL ALLOCATION:

Figure : Concentric zone of a cell.

In a reuse partitioning based allocation strategy, each cell is divided into multiple
co- centric, equalize zones, as illustrated in figure. The basic ideas that the inner
zone, being closer to the BS, would require lesser power to attain a desired
carrier to interference ratio(CIR) or signal to interference ratio(SIR). The net
effect is to have a lower value of reuse distance for inner zones as compared to
outer zones, thereby enhancing spectrum efficiency. Such reuse partitioning
schemes could be based on either fixed or adaptive allocation.
In simple reuse partitioning, mobile subscriber with the rest SIR are assigned a
group of channels that have the smallest reuse distance. A similar strategy is
used to allocate channels with the largest reuse distance and worst SIR.
Appropriate adjustment in reuse group channels needs to be performed
whenever the SIR for an MS changes. An alternative is to measure the SIR of all
MSS in the cell, sort them and assign channels starting from the inner zone to
the outer zone in descending SIR values of the mobile user. The concentric zones
are formed to help enhance channel utilization and the number of zones and
size of each zone is not fixed. Moreover, in actual practices, the zone shape and
size may not exactly correspond to a given SIR value. Therefore, many dynamic
reuse partitioning scheme have proposed.
OVERLAPPED CELLS BASED ALLOCATION:

30
One such example is shown in figure. Wherein a cell is split into seven microcells,
with separate BS and microwave tower placed at the center of each microcell.
There are many different alternatives for allocating channels. One way to assign
channels for the cell and the microcells is to characterize the mobility of each
MS into highly mobile and slow-moving groups. For slow moving MSs, channels
are assigned from one of the microcells, based on the current location. Fast
moving MSs would have more frequent handoff if channels associated with the
microcells are assigned for the same. For this reason, highly mobile MSs are
given channels from the cell. Therefore, channel allocation from cells/microcells
is matched with the speed of the MSs. In such a multitier cellular system, the
number of channels allocated to each tier depends on the total number of
channels, the area to be covered, the average speed of MSs in each tier, the call
arrival rate and duration information in each tier, desirable blocking and
dropping probabilities and the number of channels set aside for handoffs.
Optimization of such a system is fairly complex and beyond the scope of this
chapter. One approach to handle increased traffic in a cell is to split it into a
number of smaller cells inside a cell and such partitioning into smaller cells is
called microcell and picocell.
An alternative to use cells and microcells are shown in figure is to change the
logical structure dynamically, starting with only the main cell being used and
other microcells being switched off under the control of the cell for low traffic.
As traffic increases in one or more parts of the cell, the corresponding microcells
are turned on if unacceptable level of co channel interference or unavailability
of resources lead to forced cell blocking. Switching on the microcell nearest to
the MSs requesting channels makes the microcell BS physically closer to them,
thereby enhancing the CIR values of traffic decreases, then the cell switches off
selected BSs located at the microcells, thereby automatically adapting to
instantaneous call traffic density and lowering the probability of calls being
terminated. Simulation results in such multi tier networks approach a drastic
reduction in the number of handoffs and optimal partitioning, of channels
among the cells and microcells is a complex function of numerous parameters,
including the rate at which switching on and off can be done and threshold
parameters. Another possibility is to have an overlap of cell areas between two
adjacent cells, as shown in figure. In such an overlapped cells scheme, either
directed retry or directed handoff can be used. In directed retry, if an MS located
in the shaded area cannot find any free channels from cell A, then it can use a
free channel from cell B, if the signal quality is acceptable. In directed handoff,

31
another extreme step is taken to free up a channel by forcing come of the
existing connections in the shaded area of cell A to do forced handoff to cell B,
if a new calls in cell A do not find free channel. A similar measure can be taken
for other parts of cell A as well. Both of these approaches are observed to
improve system performance and many factors, including the ration of
overlapped area, total cell area, influence the blocking probabilities of
originating calls. A detailed investigation is needed to determine an appropriate
overlap so that all calls can be served and unavailability of free channels can be
minimized. Given the number of channels, we next consider how the rates for
new originating and handoff calls can influence blocking probability and hence
system performance.

CHANNEL MODELLING:
If S channels are allocated to a cell, then they have to be used both for the
originating calls in the cell and handoff calls from adjacent cells. These call rates
influence the probability of call acceptance, and it is relatively difficult to model
an exact scenario. Therefore, some simplistic assumptions are made to obtain
an approximate model of the system.
1. All MSs are assumed to be uniformly distributed through the cell.
2. Each MS moves at a random speed and to an arbitrary random
direction.

32
3. The probability of originating calls is given by ƛ0.
4. The probability of handoff calls is given by ƛH.
5. Originating and handoff calls are given equal priority.
6. The call service rate is given by 𝜇.
7. All assumptions are equally applicable to all cells in the system.
As the origination and handoff calls are treated equally by S channels of
the cell, the calls are served as they arrive and both kinds of requests are
blocked if all s channels are busy. The state of a cell should be modelled
as shown in figure. The cell state could be represented by the (S+1) states
Markov model, with each state indicating the number of busy channels
within the cell. The total request rate becomes ƛ0 + ƛH, while the service
rate remains. This leads to a state transition diagram of the M/M/S/S
model, as shown in figure. The following parameters are in effect:
P(i): the probability of i channels to be busy.
ƛ0: the probability of an originating call in the cell.
ƛH: the probability of a handoff call from neighbouring cell.
B0: the blocking probability of handoff calls.
S: the total number of channels allocated to a cell.
𝜇 : the call service rate.
𝜇c: the average call duration.
𝜇c- dwell : the outgoing rate of MSs.

33
From figure, the state equilibrium equation for state i can be given as:
P(i) = (ƛ0 + ƛu)/ i𝜇 P(i-S); 0<=i<=S ----------------(3)
Using the preceding equation recursively, along with the assumption that
the system will be in one of the (S+1) states, the sum of all states must to
be equal to one:
∑𝑆𝑖=0 𝑃(𝑖) = ƛ ---------------------------------------------- (4)
The steady state probability p(i) is easily found as follows:
P(i) = (ƛ0 + ƛu)i/ i! 𝜇 iP(0); 0<i<S --------------------------------------- (5)
Where
P(0) = 1/ ∑(ƛ0 + ƛ𝐻 )^i/ i! 𝜇 i ---------------------------------------------(6)
The blocking probability for an originating call can be expressed by
B0 = P(S) = [(ƛ0 + ƛH)S/ S! 𝜇S ]/ ∑𝑆ƛ=0(ƛ0 + ƛ𝐻 )^i/ i! 𝜇 ^i ------------(7)
The blocking probability of a handoff request is BA=B0 ------------- (8)
The equation 7 is known as the Erlang- B formula.

MODELLING FOR HANDOFF CALLS:

It is well known that of an originating call is blocked, it is not as disastrous
as opposed to a handoff call being dropped. Therefore, it is important to
provide a higher priority to handoff calls so that ongoing calls can be
maintained. One way of assigning priority to handoff requests is by
assigning SR channels exclusively for handoff calls among the S channels
in a cell. The remaining Sc = (S-SR) channels are shared by both originating
calls and handoff requests. An originating call is blocked if channels have
been allocated. A handoff request is blocked if no channel is available in
the cell. The system model must be modified to reflect priorities, as shown
in the figure.

34
The probability P(i) can be determined in a similar way, with the state
transition diagram shown in the figure. The state balance equations can
be obtained as

i 𝜇 p(i) = (ƛ0 + ƛH) P(i-1); 0<=i<=SC

i 𝜇 p(i) = ƛH P(i-1); SC<i<=S --------------------- (9)
using these equations recursively, and the addition of all (S+1) states or
∑𝑆𝑖=0 𝑃(𝑖) = ƛ ------------------------------------------------------------------------ (10)
The steady state probability P(i) can be obtained:
P(i) = [ (ƛ0 + ƛ𝐻 )𝑖 / i! 𝜇i] P(0) ; 0<=i<=SC

= [[ (ƛ0 + ƛ𝐻 ) 𝑆𝑐 ƛ𝐻 𝑖− 𝑆𝑐 ] / i! 𝜇𝑖 ] / P(0) ; SC<i<=S ------------(11)

B0 =
𝑆
[∑𝑆𝑖=0(ƛ0 + ƛ𝐻 )^i/ i! 𝜇 ^i + ∑𝐼=𝑆
𝐶
𝐶 +1
(ƛ0 + ƛ𝐻 )^𝑠𝑐 + ƛ𝐻 𝑖− 𝑆𝑐 / i! 𝜇 ^i]

---------------------- (12)

35
 The blocking probability B0 for an originating call is given by
B0 = ∑𝑆𝑖=𝑆𝐶 𝑃(𝑖) --------------------------------------------------------------------- (13)

The blocking probability BH of a handoff request or the forced

determination probability is

BH = P(s) =[[ (ƛ0 + ƛ𝐻 ) 𝑆𝑐 ƛ𝐻 𝑆− 𝑆𝑐 ] / S! 𝜇 𝑆 ] / P(0) -------------------------(14)

The relations of equations 13 and 14 re not equal as priority is given to handoff
calls. In fact, another possible improvement in servicing handoff calls is to
provide buffers for such calls so that BH can be minimized and serviced latter
even if no channels are available instantaneously.

36