JOURNAL OF COMPUTING, VOLUME 3, ISSUE 7, JULY 2011, ISSN 2151-9617

HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG

The Impact of Queue on Opportunistic based

Cognitive Radio Networks
Samira Homayouni, Seyed Ali Ghorashi
AbstractIn cognitive radio systems which include primary and secondary networks, dynamic spectrum sharing methods are
implemented in order to provide more spectrum access opportunities for secondary users without interference to the primary
ones. In this paper, we investigate the impact of queuing secondary users on the cognitive radio system performance. Resource
sharing process in the considered system is modeled by a two-dimensional Markov chain and performance metrics such as
blocking and dropping probabilities, mean waiting time and channel utilization for different populations of secondary users are
computed and compared. Both analytical and simulation results show that in the presence of queue, the dropping probability of
secondary users decreases intensively with the cost of increasing secondary users' blocking probabilities and mean waiting
time.
Index Terms Blocking and dropping probabilities, Cognitive radio, Dynamic spectrum access, Markov chain.

1 INTRODUCTION

raditional wireless networks use fixed spectrum allocation methods for licensed users. Recent studies on
the measurement of the spectrum show that by conventional spectrum allocation methods, the average utilization of the spectrum is low [1], [2]. Nowadays, exploiting the scarce spectrum resources has been significantly
improved with emergence of dynamic spectrum access
methods [3], [4] as well as cognitive radio technology [5].
Networks with opportunistic access to the spectrum allow the secondary users to use the primary users spectrum, if they are idle in a specific time or location. Each
secondary user is supposed to have the capability of realization and usage of these idle channels [5]. Thus, secondary users can coexist with primary users in a given
service area and share the spectrum with primary users
without any harmful impact on them.
A number of schemes related to opportunistic spectrum access have been investigated in the literature. For
example, [6] analyzed the modeling of interference in
spectrum access based on the listen before talk scheme. In
[7], a measurement-based model for statistically describing the idle and busy periods of a WLAN is proposed. [8]
proposed a network in which a primary user coexists
with N secondary users with different types of traffic.
They showed that the spectral efficiency can be increased
by giving permission to reuse free bands of secondary
users [9]. In [10], a channel reservation method for a spectrum sharing system has been introduced. Also several
works has been done to use queues in cognitive radio
networks; [11] analyzes the performance of the network
with a delay and discard queue. Authors in [12] have

Samira Homayouni is with the Cognitive Telecommunications Research

Group, Department of Electrical Engineering, Shahid Beheshti University
G. C., Evin 198396113, Tehran, Iran,
Seyed Ali Ghorashi is Assistant Professor of the Department of Electrical
Enginireeing, Cognitive Telecommuications Research Group, Shahid Beheshti University G. C.,Evin 1983963113, Tehran, Iran.

used channel bonding/aggregation for multi-channel

cognitive radio networks in order to take more advantage
of bandwidth. Using this method, a secondary user can
utilize separate parts of spectrum as a channel. In [13],
two channel aggregation methods (fixed and variable) are
compared. [14] proposes a performance model of an opportunistic spectrum sharing scheme. Since the detection
mechanism in practical systems is associated with errors,
then the authors have been extended their model to the
unreliable sensing cases in [15] and [16]. In these models,
secondary users can detect unused channels and make
use of them by utilizing different spectrum sensing techniques. However, due to the priority of primary users
over secondary ones, the channels cannot be used by secondary users, continuously. Thus, a secondary user upon
detecting a primary user in its current channel, should
vacant it and switch to another vacant channel, if there is
any available one. This procedure is called spectrum
hand-over. Otherwise, the secondary user connection
moves to the queue and waits to achieve an opportunity
to resume the service before the end of a maximum waiting time.
In this paper, a method of opportunistic spectrum
sharing in cognitive radio networks has been investigated
in which primary and secondary users are in the same
spectrum band. In contrast to [14], [15] and [16] in which
the number of primary and secondary users in the system
should be equal, our proposed model is a general one that
can be easily matched to different scenarios. The spectral
efficiency of channels with the presence of the secondary
users can be increased by putting secondary users in the
queue. This increases the probability of the channel being
occupied and thus increases the channel efficiency. In this
paper, we have investigated the effect of using queue in
these cognitive radio systems. Different performance
metrics such as blocking and dropping probabilities,
mean waiting time and channel utilization with and
without queue for different populations of secondary us-

JOURNAL OF COMPUTING, VOLUME 3, ISSUE 7, JULY 2011, ISSN 2151-9617

HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG

10

ers are computed and compared.

The reminder of the paper is organized as follows. Section 2 explains the system model and describes the performance analysis of channel allocation by the Markov
process. Measurement metrics of network performance
are derived in section 3. Simulation and analytical results
of a cognitive radio system with primary and secondary
users in two cases with and without queue are compared
in section 4. Finally, the paper is concluded in section 5.

2 MODEL AND PERFORMANCE ANALYSIS OF

CHANNEL ALLOCATION PROCESS
Consider a wireless cognitive network with access
points that cover a certain service area and each base station puts a set of users under the service. Spectrum owners are defined as the primary users (PUs). They can allow a secondary network to use the same spectral bands,
provided that the secondary users (SUs) only use those
parts of the spectrum that are not used by the PUs. When
a PU appears on a SUs channel, the ongoing SU must
stop its data transmission to it, and if there is another idle
channel, it has to switch. If there is no idle channel, in the
absence of any queue, the SUs service forces to stop. To
increase the spectral efficiency, a queue can be considered
for SUs. When there is no idle channel in system, the SU
can go to the queue and wait for a channel to become free
in a FIFO order. We assume that both the primary and
secondary networks are infra-structured and resource
allocation is performed centrally by their access points,
AP1 and AP2, respectively (Fig. 1).
In dynamic spectrum sharing, where SUs share unused
parts of the licensed spectrum opportunistically, PUs
have the priority to access the channel. When a call request of PUs arrives, the PU occupies the channel if it is
available. If the channels are occupied by SUs, they vacate
the channel for PUs. Therefore, PUs' request is blocked
only when all of the channels are occupied by other PUs.
When a SU arrives, it occupies an idle channel if it is
available. Otherwise, the SUs request is blocked. If some
channels are occupied by PUs and some others are used
by SUs, when a PU needs to use a certain channel, the SU
should release that channel immediately, and switch to
another channel, if there is any free. At this time, if all
channels are occupied, the SUs call is placed in a queue
(if there is queue) and temporarily avoids blocking.
Now we explain our model assumptions; N channels
are used to carry the traffic of both primary and secondary networks. We assume all cells are statistically identical, so we analyze our model in a cell. In each cell, the arrival time of both PU and SU calls follows an independent
p respectively.
s
Poisson processes with rates
and
Also,
the holding times of PU and SU calls are exponentially
distributed with means
and
, respectively.
We
1/
hp
1/
hs
assume that the resident time of PU and SU calls in the
service area is exponentially distributedwith mean and
,1/
respectively.
Also
rp
1/
rs channel holding times are set to the
minimum holding time of calls and residence time in the
area under the service. Therefore,

Fig. 1. Spectrum sharing model in a primary/secondary in the order

network.

channel holding time for PU and SU calls is exponentially

distributed with mean=
P1 1 / ( hp + rp )and S1 =
1 / ( hS + rS ), respectively [17]. For the sake of simplicity,
we also assume that both PU and SU calls occupy only
one channel for each call. Now we write ( X 1 (t ), X 2 (t ))
as a 2-D Markov process with state spaces S = {( k1 , k2 ) |
0 k1 N , 0 k2 N ) in which X 1 (t ) and X 2 (t ) are
the number of channels that are used by PU and SU at
time t, respectively. The transition rate from state (k1 , k2 )
k ', k '
to (k1 ', k2 ') is denoted by Tk11, k2 2 . Also, the indicator function 1{ x} , is defined as 1 if x is true and is 0 otherwise.
Based on the above definitions, we can define three types
for the state (k1 , k2 ) in Markov process; type-1: (k1 + k2 < N )
, type-2: (k1 + k2 =
N )and type-3: (k1 + k2 > N ) . In the next
section, we express the system model for three types of
occupancy channel status.

2.1 Morkov Model (for the case of No Queue)

The state diagram without queue for type-1and type-2
are shown in Fig. 2(a) and Fig. 2(b), respectively. Due to
the lack of queue, type-3 does not happen in this case. For
the same reason in this case we do not have the transitions from (i, j ) to (i + 1, j ) and (i, j + 1) . Moreover,
for transition from (i, j ) to (i, j 1) and (i 1, j ) there
should be at least one PU and one SU in the system. Transition from state (i, j ) to (i + 1, j 1) occurs when the
PU arrives, while all channels are occupied and there is at
least one SU in the system. In this case, the PU gets the
SUs channel by dropping its call. This transition is eliminated for type-1, because at least one channel in the existing system is empty.
Therefore, the transition rate Tkk11, k',2k2 ' for Markov model
with no queue is obtained as follows:

Tkk11, k+21, k2 = p 1{0 k1 N 1,0 k2 N k1 1}

Tkk11, k21, k2 = k1 p 1{1 k1 N ,0 k2 N k1 }
Tkk11, k, k22 +1 = s 1{0 k1 N 1,0 k2 N k1 }
Tkk11, k, k22 1 = k2 s 1{0 k1 N 1,1 k2 N k1 }
Tkk11, k+21, k2 1 = ( k2 ) p 1{0 k1 N , k2 = N k1 }

(1)

JOURNAL OF COMPUTING, VOLUME 3, ISSUE 7, JULY 2011, ISSN 2151-9617

HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG

11

p
p

s
,

(+ )s

,+

p
(+ )p

+ ,

s
,

( j ) p
+ ,

(a) (k1 + k2 < N )

N)
(b) (k1 + k2 =

Fig. 2. state transition diagram in the case of without queue

where ( k 2 ) = 0 for every k 2 = 0 and is equal to 1 in (Tkk1, k+1, k2 + Tkk1, k1, k2 + Tkk1, k+1, k2 1 + Tkk1, k, k2 +1 + Tkk1, k, k2 1 ) I{k , k }
1 2
1 2
1 2
1 2
1 2
1 2
other cases. A forced SU disconnection depends on the
k1 , k2
k1 , k2
{k1 , k2 } Tk1 +1,k2 I{k1 +1,k2 } {k1 + 1, k2 } + Tk1 1,k2 I{k1 1,k2 }
remaining number of channels and users in the system.=
The transitions rate Tkk1, k+1, k2 1
, represent the state in
1
2
{k1 1, k2 } + Tkk11,k,k22+1 I{k1 ,k2 +1} {k1 , k2 + 1} + Tkk11,k,k221
which a secondary user is interrupted from the system.

I{k1 ,k2 1} {k1 , k2 1}

2.2 Markov Model (for the case of With Queue)
To avoid dropping SUs calls when all the channels are
occupied by PUs and SUs, we consider a queue of length
N. we choose the length of the queue equal to the number
of channels, therefor, all of the on-going SU calls can
transfer to the queue to avoid SU dropping. Transition
state diagrams in the case of using queue for the type-2
and type-3 cases are shown in Fig. 3 (a) and (b), respectively. The state diagram for type-1 is the same as the case
with no queue (Fig. 2. (a)). In this case, as the number of
users in the system is less than the number of channels,
queue is not used in practice. The transition rates for the
case of queue can be written as follows:

{k , k }I
1

=
k 2 0=
k1 0

{ k1 , k2 }

=1

(4)

where, {k1 , k2 } is the steady state probability for the state

(k1 , k2 ) . Now, based on the steady state probabilities
achieved above, different metrics such as blocking probabilities for the PUs and SUs, dropping probability for SUs
and the channel utilization of system can be obtained.

In this section we describe different measures of interest, such as blocking and dropping probabilities for two
cases of with and without queue, as well as channel utilization.

= k1 p 1{1 k1 N ,0 k2 N }

Tkk11,k,k22 +1 = s 1{0 k1 N 1,0 k2 N k1 }

Tkk11,k,k22 1

and

3 PERFORMANCE METORICS

Tkk11,k+21,k2 = p 1{0 k1 N 1,0 k2 N }

k1 1, k2
k1 , k2

(3)

k2 s 1{0 k1 N 1,1 k2 N k1 } +

3.1 Call Blocking Probability for SUs

[( N k1 ) s + (k2 N + k1 )rs ]1{1 k1 N , N k1 +1 k2 N }

(2)

The SU call request is blocked, when a SU call request

arrives, and all the channels are occupied by PUs and/or
SUs. The blocking probability of the SUs is denoted by
PBSU and can be written as follows:

We define I{k1 , k2 } as a function which has two cases; it is

equal to 1 (with or without queue), if 0 k1 + k2 N (for=
PBSU
the case of with queue), and if k2 N (for the case of
without queue), and I{k , k } = 0 otherwise. Then, ba- and
1 2
lanced state equation for two Markov models (with and
without queue) can be expressed as:

(k , N k )

k1 = 0

PBSU =

k1= 0 k2= N k1

(5)

(k1 , k2 )

(6)

JOURNAL OF COMPUTING, VOLUME 3, ISSUE 7, JULY 2011, ISSN 2151-9617

HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG

there is queue in the system. Note also that, in this case

there should be at least one SU in the system. The dropping probability of the SUs is denoted by PDSU , and it is
obtained as:

s
,

s +s

(a)

,+

s +(+ + )s

(b)

(9)

The ratio of the average number of busy channels to the

number of all channels is called channel utilization and
is denoted by and can be written as follows:

1 N N k1
k0 ((k1 + k2 ) (k1 , k2 ))I{k1 ,k2 }
N=
k1 0=
2

(10)

1 N N k1
[ ((k1 + k2 ) (k1 , k2 ))I{k1 ,k2 }
=

N=
k1 0=
k2 0
N

p
( + )p

+
+ ,

(k1 + k2 > N )

3.2 Call Blocking Probability for PUs

The primary call request is blocked, when a PU call
request arrives, and all the channels are occupied by the
PUs. The blocking probability of the PUs is denoted by
PBPU and can be written as follows:

PB
=
=
(k1 N=
, k2 0)
PU

(7)

and
N

(N , k )
2

k1 =1 k2 = N k1 +1

( N (k1 , k2 ))I{k1 ,k2 } ]

(11)

for the cases of without and with queue, respectively.

In equation (11), the first term is defined as the channel
utilization of both type-1 and type-2 and the second one is
for type-3.

4 NUMERICAL AND SIMULATION RESULTS

,+

for the cases of without and with queue, respectively.

k2 = 0

(N k , k )

k2 =1

and

Fig. 3. State transition diagram when queue is used.

PBPU =

3.4 Channel Utilization Probability

(k1 + k2 =
N)

+ ,

(+ )p

s +( + )s
,

=
PDSU

12

(8)

for the cases of without and with queue, respectively.

3.3 Call Dropping Probability for SUs

The ongoing SUs call is dropped, when a PU call request arrives, all the channels are occupied by either PUs
and/or SUs, and at that time there is no empty channel to
switch to it. Note that, call dropping only happens when

In this section, we present the numerical and simulation

results to show the applicability and the performance of the
proposed channel allocation method. To obtain numerical
results, we consider a cellular system in which each cell has
10 channels available. We assume that both PUs and SUs
arrivals follow Poisson process with parameters
p = s
= 1 call / hour. The mean call time of each user is 30
seconds which follows the exponential distribution. Channel
allocation strategy is simulated for two cases. In the first
case, there is no queue for the SUs. When PUs arrive, one of
the ongoing SUs will be dropped from the channel, if there is
no other available channel. If all channels are occupied by
other PUs, the incoming PU will be blocked. In the second
case, to reduce of the SUs blocking probability, a queue is
used whose length is equal to the number of channels in the
system. We also assume that the maximum waiting time of
SUs in the queue is set to 5 sec, which means that after this
time if there is no free channel to be allocated to these SUs,
they will be dropped. It is assumed that the number of PUs
in each cell is constant and it is set to 500. However, the
number of SUs is variable. This is because of determining the
impact of the SUs population on the PUs performance.
First, we validate our analysis of the blocking probabilities by comparing the results of the simulation and numerical solution of (3) and (4) for both cases of with and without
queue. Notice that the SUs have no effect on the PUs activities, therefore having queue has no effect on the blocking
probability of PUs. Blocking probabilities of PUs and SUs are

JOURNAL OF COMPUTING, VOLUME 3, ISSUE 7, JULY 2011, ISSN 2151-9617

HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG

13

0.9

0.45

0.35
0.3

Numerical SU with queue

Simulation SU with queue
Numerical SU wihtout queue
Simulation SU without queue
Numerical PU
Simulation PU

0.8

0.25
0.2
0.15

0.75
0.7
0.65
0.6

0.1

0.55

0.05

0.5

0
100

200

800
700
600
500
400
300
Number of SUs per cell (numer of PUs=500)

with queue
without queue

0.85

channel utilization

SU and PU call blocking probabilities

0.4

900

0.45
100

1000

Fig. 4. Comparision of PU and SU call blocking probabilities, for

simulation and numerical results. Notice that the blocking probability
of PUs is independent of queue presence, as the priority is with PUs
and channel sensing is assumed to be ideal.

200

900

1000

Fig. 5. Channel utilization in two cases of with and without queue.

0.45
SU with queue
SU without queue
PU with queue
PU without queue

SU and PU call blocking probabilities

0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
100

200

300

800
700
600
500
400
number of SUs (number of PUs = 500)

900

1000

Fig. 6. SU and PU call blocking probabilities are compared for the

cases of with and without queue.
0.16
without queue
with queue

0.14

SU call Dropping probability

both calculated and simulated over a wide range of the users traffic with different populations. Fig. 4 shows a good
compatibility between numerical and simulation results of
the proposed schemes. The simulation was performed in
MATLAB.
Fig. 5 shows the channel utilization for two cases of with
and without queue. It can be observed that the channel utilization is slightly increased when queue is used for SUs. This
is due to SUs reconnection into the channel that increases
the probability of busy channel consequently.
Fig. 6 shows the blocking probability of the PUs and SUs
in two cases of with and without queue. It can be seen that in
both cases, increasing the SUs number has no effect on
blocking probability of the PUs. The reason is that we have
assumed an ideal channel sensing performed by SUs. The
blocking probability of the SUs in the case of using queue is
more than the case of without it. This can be explained as
follows. When there is a queue in the system, the channel
utilization increases and this will increase the probability of
busy channels. Then at the SUs arrival times, there is more
probability for them to see the channel busy and therefore
the blocking probability will increase.
Fig. 7 demonstrates the call dropping probability of the
SUs as the SUs traffic increase in both cases. According to
this figure, the dropping probability of the SUs in the case of
using queue is dramatically lower than the case with no
queue, because as the SUs are dropped by the PUs, they
move to the queue and the queued secondary calls reconnect
back to the system as soon as they find an idle channel.
Fig. 8 shows the mean waiting time of the SUs in the
queue. It can be seen that as the SUs population increases,
the mean waiting time will increase as well.
According to the results of figures 5 to 8, it can be said
that an increase of the SUs population does not affect the
performance of the primary users. If there is a queue, the
dropping probability of the SUs will significantly reduce at

300
400
500
600
700
800
number of SUs (number of PUs = 500)

0.12
0.1
0.08
0.06
0.04
0.02
100

200

800
600
700
400
500
300
number of SUs (number of PUs = 500)

900

1000

Fig. 7. SU call dropping probability for two cases of with and

without queue.

JOURNAL OF COMPUTING, VOLUME 3, ISSUE 7, JULY 2011, ISSN 2151-9617

HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG

SU mean waiting time in queue (seconds)

0.5

[4]

0.45

[5]
0.4
0.35

[6]

0.3

[7]

0.25
0.2

[8]

0.15
0.1
100

[9]
200

300
400
500
600
700
800
number of SUs (number of PUs = 500)

900

1000

Fig. 8. SU mean call waiting time in queue.

the expense of an increase in the blocking probability and

the waiting time of SUs. In practice and, from the users
point of view, call dropping is more unpleasant than call
blocking. Therefore, putting the queue for SUs reduces the
probability of dropping, which leads to more satisfied secondary network customers, while primary network costumers are not affected at all.

5 CONCLUSION
In this paper, we used Markov model to evaluate the impact of using queue in opportunistic radio management of
secondary users in a cognitive radio network. Metrics such
as the blocking and dropping probabilities for users in cases
of with and without queue are derived and computed. Both
numerical and simulation results show that opportunistic
spectrum sharing increases the channel utilization and if
there is a queue for SUs, this increase is even more. It also
showed that the presence of queue reduces the dropping
probability of SUs while slightly increases their blocking
probability and leads to experiencing some delays for SUs,
which is a well-known trade-off between blocking and
dropping probabilities. In general, it is demonstrated in this
paper that at least as far as we assume a perfect channel
sensing mechanism, using queue for SUs is beneficial for
secondary network users.

ACKNOWLEDGMENT
This work is supported by Iranian Education and Research Institute for Information and Communication Technology (ERICT), under the grant number 500/8974.

REFERENCES
[1]
[2]
[3]

M. McHenry, Frequency agile spectrum access technologies, in

Proc. FCC Workshop on Cognitive Radio, May 2003.
G. Staple, and K. Werbach, The end of spectrum scarcity, IEEE
Spectrum, vol. 41, pp. 4852, March 2004.
I. F. Akyildiz, W.-Y. Lee, M. C. Vuran, and S. Mohantly, Next
generation/ dynamic spectrum access/cognitive radio wireless

[10]

[11]

[12]

[13]

[14]

[15]

[16]

[17]

14

network: a survey, Elsevier Computer Networks, vol. 50, pp. 21272159, Sept.2006.
A. Attar, M. S.A. Ghorashi, Sooriyabandara and A.H. Aghvami,
Challenges of real-time secondary usage of spectrum, Computer
Networks 52 (2008) 816830.
J.Mitola III, Cognitive radio: an integrated agentarchitecture for
software defined radio, PhD thesis, KTHRoyal Institute of Technology, Stockholm, Sweden, 2000.
A. E. Leu, M. McHenry, and B. L. Mark, Modeling and analysis of
interference in listen-before-talk spectrum access schemes, Int. J.
Network Mgmt, vol. 16, pp. 131147, 2006.
S. Geirhofer, L. Tong, and B. Sadler, A measurement-based model
for dynamic spectrum access in WLAN channels, in Proc. IEEE
MILCOM06, pp. 17, Oct. 2006.
A. S. Zahmati, X. Fernando, and A. Grami Steady-State markov
chain analysis for heterogeneous cognitive radio networks, IEEE
sarnoff symposium, Princeton, pp. 1-5, May, 2010.
S. Tang and B. L. Mark, Performance analysis of a wireless network with opportunistic spectrum sharing, in Proc. IEEE Globecom07, Washington, D.C., USA, Nov. 2007.
X. Zhu, L. Shen and T.-S. P. Yum, Analysis of cognitive radio
spectrum access with optimal channel reservation, IEEE Commun.
Letters, vol. 11, no.4, pp.304-306, Apr. 2007.
M. Zhang, Sh. Jiang, G. Wei, and H. Wang, Performance analysis of the
cognitive radio network with a call level queue for secondary users,
Wireless communications, Networking and Mobile computing, Wicom09,
IEEE. Beijing, pp. 1-4, Sept. 2009.
L. J. Pla, V. Li, and F. Y, Analysis on channel bonding/aggregation
for multi-channel cognitive radio networks, IEEE (EW), European.
Lucca, pp. 468-474, Apr. 2010.
L. Jongheon, S. Jaewoo, Analysis of cognitive radio networks with
channel aggregation, IEEE (WCNC). Sydney, Australia. pp. 1-6,
Apr. 2010.
S. Tang and B. L. Mark, Performance analysis of a wireless network with opportunistic spectrum sharing, in Proc. IEEE Globecom'07, Washington, D.C., USA, Nov. 2007.
S. Tang, M. B. L, Modeling and analysis of opportunistic spectrum
sharing with unreliable spectrum sensing, IEEE Trans. On wireless
communication, pp. 1934-1943, Apr. 2009.
S. Tang, Performance modeling of an opportunistic spectrum
sharing wireless network with unreliable sensing, IEEE (ICNSC),
pp. 101-106, Apr. 2010.
Y. Fang, Y.-B. Lin, and I. Chlamtac, Channel occupancy timesand
handoff rate for mobile computing and PCS networks, IEEE Trans.
on Computers, vol. 47, pp. 679692, June 1998.

Samira Homayouni is a M.Sc. student in Cognitive Telecommunications Research Group, at the Department of Electrical Engineering,
Shahid Beheshti University, Tehran, Iran. Her research is focusedon
Cognitive Radio Networks, opportunistic spectrum access, dynamic
spectrum sharing and queuing theory.
Seyed Ali Ghorashi Received his B.Sc. and M.Sc. Degrees in electrical engineering from the university of Tehran, Iran, in 1992
and1995, respectively. Then, he joined SANA Pro. Inc., where he
worked on m odeling and s imulation of OFDM based wireless LAN
systems and interference cancellation methods in WCDMA systems.
Since2000, he worked as a research associate at Kings College
London on Capacity enhancement methods in Multi-leyer WCDMA
systemssponsored by Mobile VCE. In 2003 he received his PHD at
Kings College and since then he worked at Kings College as a
research fellow. In 2006 he joined Samsung Electronics (UK) Ltd as
a senior researcher and no w he is a f aculty member of Cognitive
Telecommunications Research Group, Department of Electrical
Engineering, Shahid Beheshti University G.C., at Tehran, Iran, working on wireless communication.