Int. J. Communications, Network and System Sciences, 2009, 2, 917-926
doi:10.4236/ijcns.2009.29107 Published Online December 2009 (http://www.SciRP.org/journal/ijcns/).
917
Enhanced Spectrum Utilization for Existing Cellular
Technologies Based on Genetic Algorithm in Preview
of Cognitive Radio
K. SRIDHARA1, Aritra NAYAK2, Vikas SINGH2, P. K. DALELA2
1Member Technology, Government of India, New Delhi, India
2C-DOT, New Delhi, India
Received September 2, 2009; revised October 21, 2009; accepted November 29, 2009
Abstract
This paper attempts to find out the distributed server-based dynamic spectrum allocation (DSA) within liber-
alized spectrum sharing regulation concept as an alternative to existing regulation based on fixed frequency
spectrum allocation schemes towards development of cognitive radio for coverage-based analogy. The pre-
sent study investigates a scenario where a block of spectrum is shared among four different kinds of exem-
plary air interface standards i.e., GSM, CDMA, UMTS and WiMAX. It is assumed to offer traffic in an
equally likely manner, which occupy four different sizes of channel bandwidths for different air interfaces
from a common pooled spectrum. Four different approaches for spectrum pooling at the instance of spectrum
crunch in the designated block are considered, viz. channel occupancy through random search, existing
regulation based on fixed spectrum allocation (FSA), FSA random and channel occupancy through Genetic
Algorithm (GA) based optimized mechanism to achieve desired grade of service (GoS). The comparisons of
all the approaches are presented in this paper for different air interfaces which shows up to 55% improve-
ment in GoS for all types of air interfaces with GA-based approach in comparison to existing regulations.
Keywords: DSA, GA, GSM, CDMA, UMTS, WiMAX, Cognitive Radio
1. Introduction
The sophistication possible in a software-defined radio
(SDR) [1–3] has now reached the level where each radio
can conceivably perform beneficial tasks that help the
user, network, and minimize spectral congestion. Radios
are already demonstrating one or more of these capabili-
ties in limited ways [4,5]. A simple example is the adap-
tive digital European cordless telephone (DECT) wire-
less phone, which finds and uses a frequency within its
allowed plan with the least noise and interference on that
channel and time slot [6]. Of these capabilities, conser-
vation of spectrum is already a national priority in inter-
national regulatory planning. As on date, there are cer-
tain rules [4] by which a fixed spectrum is allocated to
designated technology, and other technology/service pro-
vider cannot use this spectrum. We are interested to in-
vestigate this hypothesis in the case of cognitive radio [5,6]
i.e., in case of availability of spectrum anywhere, any
technology/service provider user can use that to accom-
modate maximum subscribers within limited spectrum.
As an example of the potential for utilizing the time
varying nature of the traffic, we consider four different
radio networks: GSM, CDMA, UMTS and WiMAX. We
also assume that these radio networks might be used to
support different services, e.g. voice telephony on GSM
[7], CDMA, broadband internet access along with video
streaming on UMTS (for individual subscribers) and
WiMAX (mainly for corporate connections). The traffic
pattern (and therefore demand for frequency spectrum)
seen on each of these networks would vary throughout
the day. Example traffic patterns are shown in Figure 1
based on the assumption that voice telephony and corpo-
rate connection demands will be high during office time
while individual broadband internet subscriber demand
will be high before and after office hours. Here GSM
traffic variation has been drawn with the help of refer-
ence [7] whereas the traffic variation of CDMA, UMTS
and WiMAX are drawn based on above assumption.
Here we assumed that a block of spectrum is shared
among four different kinds of exemplary air interface
standards i.e., GSM, CDMA, UMTS and WiMAX which

K. SRIDHARA ET AL. 918
DAY TIME TRAFFIC
0.0000
0.2000
0.4000
0.6000
0.8000
1.0000
1.2000
12:00:00 AM04:48:00 AM09:36:00 AM02:24:00 PM07:12:00 PM12:00:00 AM04:48:00 AM
TIME
USERS NORMALIZED
GSM
CDMA
UMTS
WIMAX
Figure 1. The peak traffic variations of the four types of technologies that share the spectrum over a 24-hour period.
occupies 200 kHz, 1.25 MHz, 5 MHz and 10 MHz frequency
bandwidths respectively. The relation among different tech-
nologies is spectrum occupancy related for coverage-based
analogy that in case of secondary users, i.e., if UMTS
allocated spectrum is fully occupied and UMTS subscriber
(which will be secondary user for other technology spec-
trum) wants to grab channel in some other allocated
technology spectrum e.g. GSM, then it would also re-
quire full 5 MHZ instead of 200 KHz for GSM. The dis-
tributed server instead of centralized server-based approach
has been taken to reduce computing time. The fixed fre-
quency spectrum has been allocated to these different
technologies. Initially, some traffic patterns based on actual
traffic load have been assumed for all these four technolo-
gies during a day. As per present regulations, initially as
traffic (number of users) increases, the specific technology
user tries to grab channel within its allocated frequency
spectrum slot and in case of unavailability of frequency
resources user would be dropped. The fixed spectrum
allocation (FSA) does have some disadvantages. For
example, most communication networks are designed to
cope with a certain maximum amount of traffic. The di-
mensioning of the network is based on the “busy hour”,
which is the time of the peak use of the network. If this
network uses its allocated spectrum fully during this hour,
then the rest of the time the spectrum is not fully utilized.
A similar pattern is also seen with other services, hence,
with the help of dynamic spectrum allocation the dropped
users can be reduced and hence GoS can be enhanced. This
paper leads through the technologies and regulatory con-
siderations to support spectrum management and optimiza-
tions that raise SDR’s capabilities and make it a cogni-
tive radio. Many technologies have come together to result
in the spectrum efficiency and cognitive radio technolo-
gies may be considered as an application on top of a ba-
sic SDR platform. In the present paper, biologically in-
spired Genetic Algorithm (GA) [8–10] based dynamic
spectrum access (DSA) [4,11,12], with distributed server
based approach [13], as one of its intended applications
have been proposed to reduce blocked users i.e., GoS by
utilizing unutilized spectrum. This paper is organized as
follows. In Section 2, simulation model for GoS of existing
regulation based on FSA and other approaches which in-
cludes GA-based optimized mechanism of channel grab-
bing has been explained along with traffic model. In Sec-
tion 3, a brief review to GA and its applicability in simu-
lation has been explained. In Section 4, simulation re-
sults are shown and Section 5 concludes this study.
2. Simulation Model
In this section, the concepts behind simulation of GoS
with time for 4 different scenarios i.e. Fixed Spectrum
Allocation (FSA), FSA random (FSA_RAND), total
random (TOT_RAND) and GA optimized mechanism
have been elaborated.
The random spectrum allocation situation is analogous
to road traffic control for multiple lanes dedicated to a

K. SRIDHARA ET AL. 919
particular type of vehicle philosophy i.e., where alloca-
tion of frequency spectrum is fixed for each technology
within a certain frequency range. The different sizes of
vehicles can be compared to different channel bandwidth
requirements for the different air interface standards. A
deregulated regime is analogous to having a traffic circle
at the road junction wherein every vehicle finds a suit-
able slot proportional to its size in the circle; the circle
itself represents the available spectrum pool. It is assumed
that a pool of F=120MHz of spectrum is available for
four different bandwidths, viz. B1=0.2MHz, B2=1.25MHz,
B3=5MHz and B4=10MHz, all operating in Time Division
Duplex mode, i.e., pairing of frequencies for uplink and
downlink is not considered. The entire band of 120MHz
is quantized in steps of f=0.05MHz for simulation purposes.
The above analogy is repeated with fixing slots of fre-
quency spectrum for different technologies. At the time of
congestion pertaining to one technology, the additional
amount of required frequency spectrum can be borrowed
from other technology slots if they have spare frequency
spectrum at that moment. The optimization of bandwidth
borrowing and lending is proposed by GA i.e., introduc-
ing regulations based on DSA with GA. Four different
approaches for spectrum pooling at the instance of spec-
trum crunch in the designated block are considered. In first
approach, channel occupancy through random search in
complete pooled frequency spectrum is simulated. This
is done by allocating chunks of the quantized spectrum to
the various users of the four technologies. For example, a
GSM user gets four blocks of 50 KHz i.e., 200 KHz for a
call but this allocation is done on the basis of a random
channel grabbing where the channel to be grabbed is gener-
ated by a random generator. In second approach, the chan-
nel occupancy through existing regulations based on fixed
spectrum allocation (FSA) is simulated. This is done by
allocating users in their fixed spectrums one after the
other is a sequential manner until space runs out for new
users on which we simply do a sequential scan to find if
there is any empty space to accommodate the new user
else the call is dropped. In third approach, FSA random i.e.,
allocation of resources to different technologies in the
designated slots only through randomized search is simu-
lated. For this scheme of allocation, the users are allocated
space only in their respective spectrums just like FSA the
difference being that the users grab channels within the
spectrums allocated with the help of a randomly gener-
ated channel number. Lastly the channel occupancy in des-
ignated slots through Genetic Algorithm (GA) based opti-
mized mechanism is simulated to achieve the desired grade
of service (GoS). The scheme of which will be explained in
Section 3. The comparisons of all the four approaches for
individual and combined traffic are presented in Section 4.
Traffic Model: For the simulation purpose we assume
a perfect channel (either idle or busy). Poisson random
process [14] is used to model the arrival traffic with rate
λ and the inter arrival time is negative-exponentially dis-
tributed with mean 1/λ. The hold time duration is gener-
ated from Gaussian random process [14] which is nega-
tive-exponentially distributed with mean 1/μ. It is as-
sumed that there is some time spent for spectral scanning
and this is required to be less than the inter arrival rate.
3. A Brief Review to GA and its Applicability
in Simulation
A Genetic Algorithm (GA) [10] is a search algorithm based
on the principles of evolution and natural genetics. It
combines the exploitation of past results with the explo-
ration of new areas of the search space. By using survival
of the fittest techniques combined with a structured yet
randomized information exchange, a GA [10] can mimic
some of the innovative flair of human search.
In our case, the GA [10] is used to maximize the spec-
tral utilization by using the least bandwidths to create spec-
trum opportunities for competing users in the spectrum.
We can describe GA [10] to find solution for blocked users
as follows. For example, in a network with d users, the
number of different ways (Γ) [15] these users can be al-
located to a spectrum bandwidth (assuming reuse factor
=1) without repetition can be computed using the fol-
lowing equation.
1
()
d
d
r
r


(1)
!
() !( )!
d
r
d
rd r
 (2)
This would lead to a total of гm [15] different combi-
nations for the m spectrum bands. For instance, a system
with blocked users d=15 and four spectrum bands for
different technologies (m=4) would have approximately
256 possible allocations and to find the optimal solution,
exhausting all combinations would not be efficient, as a
processor checking one billion solutions per second re-
quires approximately 2.3 years to analyze all permutations.
Therefore analytical methods may not be suitable for such
type of problems. Thus, GA [10] has been successfully ap-
plied to this class of combinatorial optimization problems.
Simplicity of operation and power of effect are two
main attractions of the GA [10] approach. The effective-
ness of the GA [10] depends upon an appropriate mix of
exploration and exploitation. Three operators to achieve
this are selection, crossover, and mutation [10]. GA has
parameters and variables to control the algorithm. There
are evolution operation, genetic operations and parameter
settings in GA. First evolution operation is selection. Typi-
cal methods for selection [10] are encoding scheme, fitness
function and seeding. Second genetic operations mainly
have crossover and mutation operations. The selection
parameters are defined as follows:

K. SRIDHARA ET AL. 920
1) Encoding Scheme: Encoding scheme [10] is one of the
important and crucial aspects to control the performance
of Genetic Algorithm [10]. It refers to the method of
mapping the problem parameters into a chromosome [10],
which decides the nature of being a weak or strong cod-
ing. Encoding scheme can be strong or week in terms of
its capability to explore the search space and strong encod-
ing scheme exploits more features of the solution domain.
The encoding method is a global approach to the problem.
It is global because any chromosome has enough informa-
tion to describe a set of channel- borrowings for the en-
tire network.
A chromosome is composed in the following way. For
every slot of technology in the spectrum, there is a major
chromosome slot or super-gene. Within each super-gene,
there are four actual genes. These genes represent the four
technologies within a spectrum [8].
A Gene is an array of length 10. At the first location of
the array, we keep the number of blocked slots and sec-
ond location has the number of free slots which were
formed by quantization of the spectrum. Next four loca-
tions contain the data about number of slots borrowed
from other technologies and last four locations contain in-
formation about slots lent to other technologies. Thus the
super-gene is formed as matrix of order 4 X 10 where each
row of the gene represents one particular technology.
Gene Structure
1 2 3 4 5 6 7 8 9 10
The allocated spectrum for different technologies in the
present study is in the ratio of 1:2:4:5 for GSM, CDMA,
UMTS, and WiMAX respectively. However, to show
results independent from allocated frequency, spectrum slots
for individual technologies GoS has been evaluated.
2) Fitness Function: The fitness function [10] is meas-
uring mechanism to rank the quality of a chromosome. It
serves the only link between the problem and algorithm
to search the optimal solutions.
(exp )
fitniess accommodated
unserviced
ected congestionserviced



 





（3）
where α, β, µ are constant values to suit the environment;
such that, α ε {w1, w2, w3} and w1>w2>w3; β<0 and
µ<0 for all cases.
accommodated is the measure of whether the given
number of users can be serviced by the free space and
depending upon this α takes values from w1, w2 and w3.
unserviced is the number of users blocked during the
congestion and expected congestion is calculated by work-
ing out possibly how much traffic is going to arrive in
each band and finding a ratio to this expected traffic and
the maximum capacity of the band.
3) Seeding: We have presented the solution having
initial gene to zero which is the first step of seeding [10].
While with channel-borrowing heuristic, the initial gene
can be stated with one of the possible solutions and then
the genetic flow is operated to get the best solution. Defi-
nitely with the later approach, the numbers of iterations
to converge to a solution are decreased. This approach is
applied with the improved pluck operation [9].
The genetic operations are defined as follows:
1) Crossover: Crossover [10] is the most important func-
tion in GA [10], which produces children as new chro-
mosomes from the process of combining two chromo-
somes (parents). The operation of crossover may gives
the children (new chromosomes) with better fitness as it
takes best attributes from both the parents [10] we have
used single point crossover as shown below.
First matrix
| cut point
1 2 3 4 5 6 7 8 9 5
2 3 4 3 4 8 7 3 5 8
5 6 7 5 6 4 3 1 4 9
9 9 3 7 2 6 1 5 3 2
Second matrix
| cut point
4 5 6 5 7 4 6 5 7 9
4 3 5 8 7 9 4 7 9 8
1 2 2 6 6 3 4 8 9 3
5 3 9 3 7 6 6 2 8 1
The offspring or children [10] generated from two pa-
rental matrices are:
offspring 1
4 5 6 5 7 6 5 8 5 9
4 3 5 8 7 4 8 3 5 8
1 2 2 6 6 3 4 1 4 9
5 3 9 3 6 2 2 5 3 1
offspring 2
1 2 3 4 5 4 6 7 7 9
2 3 4 3 4 9 8 7 9 7
5 6 7 5 6 4 3 8 3 9
9 9 3 7 7 2 6 6 8 1
This crossover function facilitates not to copy entire
second half of the second matrix to the first matrix, while
it retains the common elements of the parental matrix,
which is essential for borrowing policy to preserve neces-
sary information or the parent characteristics of the chil-
dren. However if row wise second half of the parental
matrices are entirely different then this crossover function
serves as normal crossover function discussed initially.
2) Mutation: The mutation [10] operator is responsible
for diversity into a population. In the first phase of re-
sults, we are using the normal mutation operator with
high probability to provide diversity in genes as we have
started with initial gene zero in seeding. However, before
swapping the child to next population, we check the

K. SRIDHARA ET AL. 921
needy technologies that require borrowing with correct
mutations applied by means of a simple probability.
3.1. GA Application in the Simulation
In this simulation, whenever the users find that their re-
spective allocated spectrum are fully occupied, then they
try to grab empty spectrum in the other allocated tech-
nology spectrum bands. This is where GA is employed to
select an empty space among the many available by re-
peated iterations based on crossovers and seeding so that
due to the borrowing of spectrum by another technology
user i.e., the secondary users [6] quality of service (QoS)
of primary users [6] are not affected. The allocation is
done by first randomly seeding the population with all
possible solutions and then applying fitness function to
select only the healthy children [10] or fit solutions. The
crossover and mutation is applied to further consolidate
the results and possible scheme of allocation of unutilized
spectrum based on the fitness function mentioned earlier,
which keeps in mind the possible traffic pattern and future
trends to evolve an optimal solution. Also before the appli-
cation of GA, it is checked if a specific technology spec-
rum which has been full and cannot accommodate the pri-
mary users, a spectrum search scanning [6] has been per-
formed to find out the secondary users residing in this spe-
cific spectrum slot and terminates them to create space for
the primary users before the application of GA. The termi-
nation is done on the basis of least number of calls to be
terminated and in order of time of residence in the spectrum.
The flowchart shown in Figure 2 describes the steps
employed to obtain the optimal solutions using GA [10]
for DSA by using its basic operators. As shown in the
flowchart, the results are obtained by following a number
of iterations which is fixed as a constant named MAX
ITERATION. In each of the iterations, all the basic steps
are followed until we obtain an optimal solution.
4. Results
This section shows the comparison of GoS among fixed
frequency spectrum allocation (FSA) which is existing
regulation, FSA random (FSA_RAND), total randomized
allocation (TOT_RAND) and GA-based DSA scheme which
is proposed regulation for GSM, CDMA, UMTS and Wi-
MAX along with mixed traffic of all these technologies.
Figure 3 shows the case of GoS of GSM for 120 MHz
common spectrum for all the technologies during a span
t
Figure 2. Flow chart of GA applicable to DSA in the present case.

K. SRIDHARA ET AL. 922
GSM BLOCKING
0.0000
0.2000
0.4000
0.6000
0.8000
1.0000
1.2000
00:00:00 04:48:0009:36:00 14:24:00 19:12:0000:00:00 04:48:00
TIME
GOS
GSM_FSA
GSM_FSA_RAND
GSM_GA
G SM_TOT_RAND
Figure 3. Comparison of GoS for GSM (0.2 MHz) users among FSA, FSA_RAND, GA-based DSA scheme and TOT_RAND
over duration of 24 hours of the day.
of one day. The GSM_FSA graph shows GoS for GSM
when different technologies users grab the channel slots
in the allocated frequency spectrum slots only by using
FSA which is existing regulation. The GSM_TOT_RAND
graph shows the GoS for GSM user when there is no
regulation and any technology user can grab the channel
at random position if it is available. The GSM_FSA_
RAND graph is of the randomized channel grabbing FSA
where the spectrum is divided among the various tech-
nologies but the channel grab is at any random position
provided it is vacant. The GSM_GA graph shows the
GoS for GSM user with GA optimized DSA algorithm
which is proposed regulation. It is seen that GA optimized
DSA algorithm is always better than all other cases. Also
it has approximately up to 72% improvement in com-
parison to FSA which is the best performing algorithm
amongst the three. If the mean GoS is taken for the given
time then a maximum of 67% improvement is noticed
neglecting the cases where DSA gives zero blocking.
Figure 4 shows the case of GoS of CDMA for 120
MHz common spectrum for all the technologies during a
span of one day. The CDMA_FSA graph shows GoS for
CDMA when different technologies users grab the channel
slots in the allocated frequency spectrum slots only by
using FSA which is existing regulation. The CDMA_TOT_
RAND graph shows the GoS for CDMA user when there
is no regulation and any technology user can grab the
channel at random position if it is available. The CDMA_
FSA_RAND graph is of the randomized channel grabbing
FSA where the spectrum is divided among the various
technologies but the channel grab is at any random posi-
tion provided it is vacant. The CDMA_GA graph shows
the GoS for CDMA user with GA optimized DSA algo-
rithm which is proposed regulation. It is seen that GA
optimized DSA algorithm is always better than all other
cases. Also it has approximately up to 72% improvement
in comparison with FSA which is the best performing algo-
rithm amongst the three. If the mean GoS is taken for the

K. SRIDHARA ET AL. 923
given time then a maximum of 95% improvement is no-
ticed neglecting the cases where DSA gives zero blocking.
Figure 5 shows the case of GoS of UMTS for 120 MHz
common spectrum for all the technologies during a span
of one day. The UMTS_FSA graph shows GoS for UMTS
when different technologies users grab the channel slots
in the allocated frequency spectrum slots only by using
FSA which is existing regulation. The UMTS_TOT_RAND
graph shows the GoS for UMTS user when there is no
regulation and any technology user can grab the channel
at random position if it is available. The UMTS_FSA_RAND
graph is of the randomized channel grabbing FSA where
the spectrum is divided among the various technologies
but the channel grab is at any random position provided
it is vacant. The UMTS_GA graph shows the GoS for
UMTS user with GA optimized DSA algorithm which is
proposed regulation. It is seen that GA optimized DSA
algorithm is always better than all other cases. Also it has
approximately up to 30% improvement in comparison to
FSA which is the best performing algorithm amongst the
three. If the mean GoS is taken for the given time then a
maximum of 60% improvement is noticed neglecting the
cases where DSA gives zero blocking.
Figure 6 shows the case of GoS of WiMAX for 120
MHz common spectrum for all the technologies during a
span of one day. The WiMAX_FSA graph shows GoS
for UMTS when different technologies users grab the
channel slots in the allocated frequency spectrum slots
only by using FSA which is existing regulation. The
WiMAX_TOT_RAND graph shows the GoS for Wi-
MAX user when there is no regulation and any technol-
ogy user can grab the channel at random position if it is
available. The WiMAX_FSA_RAND graph is of the ran-
domized channel grabbing FSA where the spectrum is
divided among the various technologies but the channel
grab is at any random position provided it is vacant.
The WiMAX_GA graph shows the GoS for WiMAX
user with GA optimized DSA algorithm which is pro-
posed regulation. It is seen that GA optimized DSA al-
gorithm is always better than all other cases. Also it has
approximately up to 15% improvement in comparison with
FSA which is the best performing algorithm amongst the
three. If the mean GoS is taken for the given time then a
maximum of 70% improvement is noticed neglecting the
cases where DSA gives zero blocking.
Figure 7 shows the case of GoS of all four technologies
for 120 MHz common spectrum for all the technologies
CDMA BLOCKING
0.0000
0.2000
0.4000
0.6000
0.8000
1.0000
1.2000
00:00:00 04:48:0009:36:00 14:24:00 19:12:0000:00:00 04:48:00
TIME
GOS
CDMA_FSA
CDMA_FSA_RAND
CDMA_GA
CDMA_TOT _RAND
Figure 4. Comparison of GoS for CDMA (1.25 MHz) users among FSA, FSA_RAND, GA-based DSA scheme and TOT_
RAND over duration of 24 hours of the day.

K. SRIDHARA ET AL. 924
UMTS BLOCKING
0.0000
0.2000
0.4000
0.6000
0.8000
1.0000
1.2000
00:00:00 04:48:00 09:36:00 14:24:0019:12:00 00:00:00 04:48:00
TIME
GOS
UMTS_FSA
UMTS_FSA_RAND
UMTS_GA
UMTS_TOT_RAND
Figure 5. Comparison of GoS for UMTS (5 MHz) users among FSA, FSA_RAND, GA-based DSA scheme and TOT_RAND
over duration of 24 hours of the day.
WIMAX BLOCKING
0. 0000
0. 2000
0. 4000
0. 6000
0. 8000
1. 0000
1. 2000
00:00:00 04:48:00 09:36:0014:24:00 19:12:00 00:00:00 04:48:00
TIME
GOS
WIMAX_FSA
WIMAX_FSA_ R AN D
WIMAX_GA
WIMAX_TOT_RAND
Figure 6. Comparison of GoS for WiMAX (10 MHz) users among FSA, FSA_RAND, GA-based DSA scheme and TOT_
RAND over duration of 24 hours of the day.

K. SRIDHARA ET AL. 925
TOTAL TRAFFIC BLOCKING
0.0000
0.2000
0.4000
0.6000
0.8000
1.0000
1.2000
00:00:00 04:48:0009:36:00 14:24:00 19:12:0000:00:00 04:48:00
TIME
GOS
BLOC KED _F SA
BLOC KED _F SA_ R AN D
BLOCKE D _GA
BLOC KED _TOT_RAND
Figure 7. Comparison of GoS for all technologies mixed traffic users among FSA, FSA_RAND, GA-based DSA scheme and
OT_RAND over duration of 24 hours of the day. T
during a span of one day. The BLOCKED_FSA graph
shows GoS for UMTS when different technologies users
grab the channel slots in the allocated frequency spec-
trum slots only by using FSA which is existing regulation.
The BLOCKED_TOT_RAND graph shows the GoS for
the combined user traffic when there is no regulation and
any technology user can grab the channel at random posi-
tion if it is available. The BLOCKED_FSA_RAND graph
is of the randomized channel grabbing FSA where the
spectrum is divided among the various technologies but
the channel grab is at any random position provided it is
vacant. The BLOCKED_GA graph shows the GoS for
combined user traffic with GA optimized DSA algorithm
which is proposed regulation. It is seen that GA optimized
DSA algorithm is always better than all other cases. Also
it has approximately up to 55% improvement in com-
parison to FSA which is the best performing algorithm
amongst the three. If the mean GoS is taken for the given
time then a maximum of 80% improvement is noticed
neglecting the cases where DSA gives zero blocking.
5. Conclusions
In this study, an attempt is made to analyze the impact of
our proposed GA based optimized DSA mechanism
for spectrum utilization by comparing GoS for existing
FSA based regulation with proposed mechanism. Four
prominent commonly used technologies are used for the
simulations which occupy different bandwidths to ana-
lyze the impact of our proposed Genetic Algorithm based
solution which uses aspects of cognitive radio for im-
proving the overall GoS of a shared cellular spectrum
scenario. The simulation results can be utilized to justify
the need of regulatory approach in case of liberalized
spectrum sharing in the present cellular network spec-
trum in the preview of cognitive radio. As it can be seen
from results in Figures 3 to 7 that GA based algorithm
enhances the GoS when compared to the present alloca-
tion scheme. The maximum value of improvement in
GoS for GSM, CDMA, UMTS, WiMAX and mixed traf-
fic are 72%, 72%, 30%, 15% and 55% respectively. The
improvement in GoS for GSM, CDMA, UMTS, Wi-
MAX and mixed traffic around mean values are 67%,
95%, 60%, 70% and 80% respectively. This study shows
some sort of regulatory based approach within liberalized
spectrum sharing concept is required to allow all users to
get equal chance of utilizing the spectrum and getting a
better GoS for the overall traffic.
6. References
[1] M. P. Olivieri, G. Barnett, A. Lackpour, A. Davis, and P.

K. SRIDHARA ET AL. 926
Ngo, “A scalable dynamic spectrum allocation system
with interference mitigation for teams of spectrally agile
software defined radios,” New Frontiers in Dynamic Spec-
trum Access Networks, pp. 170–179, November 2005.
[2] J. Mitola III, “Cognitive radio an integrated agent archi-
tecture for software defined radio,” Dissertation, Doctor
of Technology, Royal Institute of Technology (KTH),
Sweden, May 2000.
[3] Paul Burns, “Software defined radio for 3G,” Artech House,
Inc., 2003.
[4] J. Hwang and H. Yoon, “Dynamic spectrum management
policy for cognitive radio: An analysis of implementation
feasibility issues,” New Frontiers in Dynamic Spectrum
Access Networks, 3rd IEEE Symposium, Digital Object
Identifier, pp. 1–9, October 2008.
[5] S. Haykin, “Cognitive radio: Brain-Empowered wireless
communications,” IEEE Journal on Selected Areas in Com-
munications, Vol. 23, No. 2, February 2005.
[6] Bruce A. Fette., editor, “Cognitive radio technology,” El-
sevier Inc., 2006.
[7] S. Almeida, J. Queijo, and L. M. Correia, “Spatial and tem-
poral traffic distribution models for GSM,” Vehicular Tech-
nology Conference, Vol. 1, pp. 131–135, September 1999.
[8] A. Y. Zomaya, Senior Member, IEEE, and Michael Wright,
“Observations on using genetic-algorithms for channel
allocation in mobile computing,” IEEE Transactions on
Parallel and Distributed Systems, Vol. 13, No. 9, Sep-
tember 2002.
[9] S. S. M. Patra, K. Roy, S. Banerjee, and D. P. Vidyarthi,
“Improved genetic algorithm for channel allocation with
channel borrowing in mobile computing,” IEEE Transac-
tions on Mobile Computing, Vol. 5, No. 7, July 2006.
[10] M. Melanie, “An introduction to genetic algorithm,” MIT
press, Cambridge, 1999.
[11] D. Maldonado, B. Le, A. Hugine, T. W. Rondeau, and C.
W. Bostian, “Cognitive radio applications to dynamic spec-
trum allocation: A discussion and an illustrative exam-
ple,” New Frontiers in Dynamic Spectrum Access Networks,
First IEEE International Symposium, pp. 597–600, No-
vember 2005.
[12] P. Leaves, S. Ghaheri-Niri, R. Tafazolli, L. Christodoulides,
T. Sammut, W. Staht, and J. Huschke, “Dynamic spectrum
allocation in a multi-radio environment: Concept and al-
gorithm,” 3G Mobile Communication Technologies, Second
International Conference on (Conference Publication No.
477), pp. 53–57, March, 2001.
[13] J. Zhao, H. T. Zheng, and G. H. Yang, “Distributed coordina-
tion in dynamic spectrum allocation networks,” New Fron-
tiers in Dynamic Spectrum Access Networks, pp. 259–268,
November 2005.
[14] Viswanathan and Thiagarajan, “Telecommunication switch-
ing systems and networks,” Prentice-Hall, New Delhi,
1992.
[15] D. Thilakawardana, K.Moessner, and R.Tafazolli, “Dar-
winian approach for dynamic spectrum allocation in next
generation systems,” IET Communications, Centre for
Communication Systems Research, University of Surrey,
Guildford, UK, 2008.