International Journal of Computer Applications (0975 – 8887)

Volume 176 – No. 32, June 2020

Reengineering MANET Routing using Ant Colony

Optimization: Modelling and Performance Study

Bright Selorm K. Anibrika Michael Asante Benjamin Hayfron-Acquah

Comp. Science Department, Computer Science Dept., Computer Science Dept.,
Koforidua Technical University Kwame Nkrumah Univ. of Kwame Nkrumah Univ. of
Box KF 981, KTU, Koforidua Science and Technology Science and Technology
Private Mail Bag, Kumasi Private Mail Bag, Kumasi

Ebenezer K. Gavua
Comp. Science Department, Koforidua Technical University
Box KF 981, KTU, Koforidua

ABSTRACT
A wireless network topology that comprise heterogeneous
routing paths without the supported of any preliminary
network infrastructure is known as a Mobile Ad Hoc
Network. Therefore, the nodes have the capacity to forward
packets to nodes. Furthermore, MANET properties, namely;
dynamic topology, nodal mobility, provide a large degree of
freedom and the ability to self-organize give them an urge
over other network architectures and topologies. For that
matter, designing and developing secure routing algorithms
becomes a daunting task for the researcher. To this end, this
paper seek to compare preexisting and proposed routing
algorithm for MANET based on the mechanism of the ant
system , hence Ant Colony Optimization frame would be
adopted. It is notable that MANET bandwidth, radio
propagation, energy supply, etc. Different MAC protocols
have proposed for adhoc networks. In this research, a new
algorithm based on the Ant Colony Optimization algorithm
framework is proposed. The proposed Ant Colony Fig. 1: A Pictorial Representation of Ant Colony
Optimization algorithm known as Optimized Multicast Optimization Algorithms (Anibrika et al., 2020; Dorigo et
Routing Algorithm was implemented with the aid of MANET al., 2006)
simulation in Mathlab and a performance comparison was
carried out. The study compared, the propriety protocol 1.2 Problem Statement
AODV, with the proposed algorithm. It was observed the Mobile devices engaged in MANET communication have the
proposed protocol outperformed AODV under the quantitative potential to connect to wireless networks. The tendency of
metrics used. such mobile devices to connect to wireless networks is termed
as Mobile Ad-hoc NETworks (Jha, 2010; Kumar et al, 2013).
General Terms These networks are infrastructureless in nature. Because they
Routing Protocols, Energy-Efficiency, Wireless Routing. Ant lack infrastructure as their “building block”, a MANET could
Colony Optimization. be deployed with lower power costs. Due to the nature
MANET, routing in such networks are bedeviled with
Keywords numerous challenges. These challenges are worsened the
Ants, Algorithm, Nodes, Pheromone, Routing. unpredictable nature of nodal communication especially in
finding the shortest routes from source to destination. Several
1. INTRODUCTION algorithm were proposed to address such challenges. One
Applications such as emergency disaster relief to networking class of such algorithms is termed Ant Colony Optimization
laptops in a conference room rely on Mobile Ad Hoc (ACO) algorithms, (Dorigo et al., 2006). ACOs aside their
Networks for operability. Despite MANET challenges, several ability to address routing challenges in MANET are also
optimization models to address these routing algorithms were energy-efficient as they turn conserve the energy of the nodes
proposed in the past, each (routing algorithm) is primarily engaged in routing within MANET. The ant algorithm is
responsible for determining the minimum routes to reach the iterative, probabilistic meta-heuristic which are intended to
destination from the source as described in (Perkins and provide optimal solutions to combinatorial optimization
Bhagwat, 1994; Jacquet et al., 2001). Also, extensive research problems most importantly in MANETs (Tonk, 2012). In
were out in the area of energy-efficient routing algorithms. searching for food, ants exchange information through
Find below the conceptual view of using ACO algorithms in pheromone deposits, which is a mark of their respective route
modelling routing problems (Royer et al., 2001; Patel and from their nests to a food source and that also attracts more
Kamboj, 2015), see Fig. 1. ants to access the route. In some cases, a dynamic problem

38
 International Journal of Computer Applications (0975 – 8887)
Volume 176 – No. 32, June 2020

might occur, demanding a change over time where the routing. Furthermore, each pheromone table has a routing
algorithm need to monitor by keeping track of the critical entry for each neighbor node, demonstrating the probability of
updates and consistently optimize the algorithm so as to using that neighbor’s node as the next hop. The choice of a
generate efficient solutions. The ant algorithm has several node choice is probabilistic while route creation is
striking characteristics such as its dynamism, robustness, deterministic Dorigo et al., 2006 ; Nand and Shama, 2011). A
adaption and its decentralized nature during node processing further explanation in Fig. 3 is found below:
(Tonk, 2012; Singh, 2013). These characteristics for dynamic
routing in modern wireless networks as in the case of
MANETs. To this end, this paper investigates Ant Colony
Optimization algorithms and their application to MANETs as
well as design a new multicast MANET routing algorithm that
is energy-efficient. Finally, a comparative study through
simulation would be carried out between AODV and the
proposed algorithm for MANET.

2. LITERATURE REVIEW
2.1 Ant-Based Control Algorithms
Research work in the past had focused on Swarm Intelligent
routing algorithms as described (Dorigo et al., 2006). Initially,
this algorithm was applied to circuit switched networks, like
the telephony network. However, the ABC algorithm was
later modeled analytically in (Singh, 2013). Again, a critical
component of this design (ABC) is the congestion-free
mechanism introduced stop calls from dropping
intermittently. In (Subramani A and Krishnan A, 2011), the
network topology and its routing updates are theoretically
modeled by means of a graph. Furthermore, each available
node is indicated by a total capacity Ci, a spare capacity Si and
an equally important routing table R𝑖. A link in the form of
(i,j) exemplifies, a vector quantity of unique pheromone
values Tij,d , which implies that, each destination d, represents
the tendency of a node traversing from i to j (Nand and
Shama, 2011). For this reasoning, the probabilistic values in
the routing table are rationalized as the ant traverses the node.
This is influenced by the pheromone trails and its intensities. Fig. 3. ABC model for pheromone update
This is furthermore explained in Fig 2.0 below. 2.3 AntNet Algorithm
According to (DiCaro and Dorigo, 2006; Shukla et al., 2013),
it is important to develop an improved algorithm influenced
by pheromone updates whereby an ant colony move
systematically in a food search process. It involves, the use of
stochastic decision making with local policy, which is once an
ant develops a possible solution to the problem. The ants
evaluate possibilities of the solution and leaves pheromone
trails (memory buffer) on the path about the connection the
ants established. Finally, this specialized information would
now be used to determine the forward-bias movements of
future ants (Rafiq et al., 2013). In the same vein (Atsushi
Iwata et al., 1999) described the traditional ACO algorithm
employed using a fixed amount of pheromone as means of
updating the pheromone table. (Atsushi Iwata et al., 1999),
however developed an adaptive adjustment strategy for the
pheromone introduced, making a relatively uniform
pheromone distribution based on probabilities. This strategy
can deal effectively with the confusion of expanding the ants’
search and finding optimal solution algorithms, for finding the
local optimal solution (Dorigo, 2006).

2.4 Updating Routing Tables

Fig. 2. Pheromone Table Update (Anibrika et al., 2020)  T , C  1,if CT  1
r    1C,  O therwise
2.2 Ant Based Control Routing with 
Probabilistic Update
During the routing mechanism, routing tables get replaced by The trip time of the ant is represented by T. The scaling factor
pheromone table entries. Also, each node within Ant Based is denoted by C, where 2 is the default value. The mean and
Control network keeps a copy of the pheromone table for variance also denote the round trip time. Furthermore, = 𝜎 𝜇
every other node that form an integral part of the network , which is a ratio of the variance to the mean, denotes a

39
 International Journal of Computer Applications (0975 – 8887)
Volume 176 – No. 32, June 2020

quantitative measurement (Dorigo, 2006; Anibrika et al., 3.3 A Scenario of the Simulation
2020).
Algorithm
2.5 Mobile Ad-Hoc Networks
The topographic structure is completely infrastructureless.
This phenomenon has created numerous challenges for
MANET research, since the ultimate aim of the routing
mechanism is unclear due to several challenges such as
limited bandwidth, power, and latency (Atsushi Iwata et al.,
1999; M. Bergamo; Jha, 2010; Chlamtac et al., 2003). The
diagram in Fig. 4 below, explains further.

Fig. 4. A Mobile Adhoc Network to Solve Rescue

Operation Challenges (Veerasamy, 2016; Anibrika et al.,
2020)

3. SIMULATION STUDY
3.1 Introduction
The use of real life measurements is certainly costly (Chiang
et al., 1997). Therefore, the best approach to adopt is a
simulated environment.

3.2 Mobility
Nodal mobility are explained by the models known as
Mobility Models. Also, simulating mobility model is intended
to reveal the discrete nature of the routing algorithm. It shows
how well the algorithm can perform when subjected to certain Fig. 5. Flowchart of the simulation OMRA algorithm
conditions. It can also be used for inter-group communication
during the simulation stage (Arun and Shweta, 2016; Anibrika
et al., 2020; V. Davies; Tan et al., 2002).

40
 International Journal of Computer Applications (0975 – 8887)
Volume 176 – No. 32, June 2020

3.4 Flow Chart of Ant Colony

Optimization

Fig. 7. RREQ Message to Destination Node. (Anibrika et

al., 2020) The Algorithm for AODV

Fig.6. ACO Flowchart

3.5 The AODV Framework

Routers within the network move indiscriminately and
connect to themselves subjectively. The AODV protocol was
developed with the following objectives (Arun and Shweta,
2016): Fig.8. AODV Flowchart
i. To broadcast routes discovery packets. 3.6 The Proposed Algorithm (Optimized
ii. To differentiate among local node connectivity Multicast Routing Algorithm)
management. It has the following properties:
iii. General topology maintenance i.It is proactive
iv. To circulate updates in the routing table in a ii.The propagation technique employed are; Anycast,
localized fashion to the neighboring nodes. Unicast and Multicast.
Each node in AODV is able to maintain two major counters: iii.The proposed algorithm can create routing table
entries in each node, Network Table (NT),
i. A Broadcast-ID: It is increased at the source node Optimum Routes Table (ORT). It saves all optimum
by generating a new Route Request (RREQ) routes to each node in the network.
message.
i. The algorithm manages route repair and
ii. A Sequence Number: It is a counter which is maintenance.
increased sequentially so as to maintain current
updates. ii. Two hops a registered from the new node.
i. The nodes maintain localized information. iii. It will initiate a route discovery by building and
saving routes the nodes.
The two major mechanisms that characterize AODV routing
protocol are; route discovery, route maintenance. Find further iv. The nodes acquire information from any route that
explanation in the diagram below. passes through. The diagram below gives further
explanation, Fig. 9.

41
 International Journal of Computer Applications (0975 – 8887)
Volume 176 – No. 32, June 2020

3.7.4 Timer Initialization

The timers are initialized at beginning of the simulation. The
ant timer 𝑇𝑎𝑛𝑡 denotes the termination condition for ant
algorithm. The simulation timer 𝑇𝑠𝑖𝑚𝑢𝑙 denotes time the ant
algorithm will be repeated each 𝑇𝑎𝑛𝑡 . For instance, if 𝑇𝑠𝑖𝑚𝑢𝑙
=180 secs. and 𝑇𝑎𝑛𝑡 =30 secs., the ant algorithm 6 repetitions
on 6 separate topologies.

3.7.5 Network Parameters and Assumptions

i. End-to-End Delay: It is the total estimated time
taken by a data packet during packet transmission
destination. Mathematically calculated as: E2E
Delay =
Fig. 9. Optimum and No Route between 3 Neighbours
ii. Σ =
3.7 Local Environment Setup 𝐴𝑟𝑟𝑖𝑣𝑎𝑙 𝑡𝑖𝑚𝑒 − 𝑠𝑒𝑛𝑑 𝑡𝑖𝑚𝑒 /
Several discrete event simulation systems have been proposed Σ (𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑐𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑜𝑛𝑠)
for MANET, namely: NS2, Opnet, Simulink, Qualnet, iii. Packet Delivery Ratio: The acute failure of the
OMNet++, NetSim, etc. Upon evaluation of the simulators router to transfer and deliver packets of data
above based on the objectives of this paper, Mathlab was because of loss of connections. Calculated as:
used. Mathlab describes an environment for multi paradigm 𝑃𝐷𝑅 = 𝑆𝑢𝑚 𝑜𝑓 𝑛𝑜. 𝑜𝑓 𝑝𝑎𝑐𝑘𝑒𝑡𝑠 𝑟𝑒𝑐𝑖𝑒𝑣𝑒𝑑/
and numerical computing environment with programming 𝑆𝑢𝑚 𝑜𝑓 𝑛𝑜. 𝑜𝑓 𝑝𝑎𝑐𝑘𝑒𝑡𝑠 𝑠𝑒𝑛𝑡
languages integrated. It allows for the manipulation of
matrices, plotting functions, algorithms, implementation and iv. Throughput: It is the ratio of output with respect
creation of user interfaces. Mathlab is written in C/C++ and to input signals, due to different load of users
Java. sharing network resources, known as bit rate.

3.7.1 Network Configuration v. Simulation Model A simulation area 1000×800 m2

The topology of MANET is plotted on to a fixed graph with is set for this simulation based on the random
nodes (N) and links (M). The links could be described as waypoint model. Find below in Table 1, details of
pipes, indicating two parameters. These parameters are the simulation parameters and the corresponding
bandwidth in bits per secs and a delay in transmission in values. The diagram Table 1, outlines the
seconds. In this simulation, every node is able to store and configurations and its parameters for the simulation.
forward routing information holds a buffer space (queue) as
Table 1. Simulation Parameters
described in Fig.10 below.
Parameter Value
Number of nodes 100, 200, 300……1000
i. Rate of Arrival 150 Kbps
ii. Range of 250 m
Transmission
iii. Velocity and 10 m/sec & 45 degree
Direction
iv. Size of Packet 64 byte
v. Pause time 10, 20, …, 100 sec.
(Ant time)

Fig.10. The Node Model vi. Time of 180 sec.

Simulation
Packets are put in a queue and transmitted on First-In-First-
Out (FIFO) procedure. vii. bandwidth 1 Mbps
Link
3.7.2 Routing Table Initialization
viii. area of 1000 m × 800 m2
The probability value Pin expressing the quality of selecting
Simulation
n as the next node. It is influenced by the constraint:
ix. Model RWM
𝑃𝑖𝑛 = 1 ∈ 1, 𝑁𝑘 , 𝑁𝑘 = 𝑛𝑒𝑖𝑔𝑕𝑏𝑜𝑟𝑠 𝑘 ;
𝑛∈𝑁𝑘 x. Protocols AODV, OMRA

3.7.3 Parameter Initialization

The simulation process must begin by defining λ denoting the
arrival time. The service rate µij also denotes the bandwidth 3.8 Mobility Modelling for MANET
of two nodes, i and j within the network. These parameters ( i This paper explores the following mobility models for the
and j) are used to determine the average packets serviced in simulation (V. Davies, 2002):
each unit of time.

42
 International Journal of Computer Applications (0975 – 8887)
Volume 176 – No. 32, June 2020

i. Random Way Point: ii. End to End Delay for Mobility Models
A node selects a random position with the parameters (x, y).
This point becomes the destination point. Velocity (V) is
chosen by computing the distributed range of parameters
[minspeed, maxspeed]. This process is repeated based on the
destination and speed.
ii. Random Walk
It enables the nodes to freely move about selecting a speed
and a direction value in constant time intervals (Δt).
iii. Random Direction
The nodes choose an arbitrary direction within the specific
range [0, 2∏]. Again, the nodes’ velocity is chosen
consistently from the specified range [minspeed, maxspeed].

4. EXPERIMENTAL RESULTS AND

ANALYSIS
These graphs are labeled and interpreted.
The newly proposed ant algorithm was simulated using the Fig. 12. OMRA with End-to-End Delay
three mobility models. The Packet Delay values generated by OMRA protocol with
i. Graph of three Mobility Models the three simulation mobility models under consideration for
the purposes of this research work is demonstrated in fig. 12.
When the OMRA protocol was simulated with End-to-End
Delay (1 m/s), it exhibited the least delay value. Whereas,
under Random Walk Mobility model (1.3 m/s), the delay is
greater compared to Random Waypoint. The value is better
than the one obtained for Boundless Simulation Area Mobility
model (3 m/s), exhibiting high values of End-to-End Delay as
there is an increment in the speed of the nodes.
iii. Packet Delivery Ration in OMRA

Fig. 13. The Packet Delivery Ration in OMRA with

Mobility Models.
Fig.13, presents Packet Delivery Ratio graph for OMRA
Fig. 11. Simulation of the Mobility Models algorithm. It is realized that, the Boundless Simulation Area
Mobility Model (0.98 sec) is improved providing a better
In Fig.11 the results for Throughput are presented. The nodes’ performance value with regards to the Packet Delivery Ratio
speed vary from 10 m/s to 50 m/s. From fig. 11, it is realized when compared to the remaining models. This improvement,
that, the performance of OMRA with respect to Throughput however comes at the cost of low Throughput and high End-
under Random Waypoint and Boundless models have certain to-End Delay. With regards to the remaining two models,
similarities with very little differences. However, an Random Walk (0.89 secs) provided better and enhanced
increment in the speed of the node, affects the Throughput Packet Delivery Ratio which is better than the Random
value as it is seen to have been steady. Random Walk Waypoint value (0.859). Also, the performance of Random
however, demonstrates a gradual decline in the throughput Waypoint Mobility model concerning Packet Delivery Ratio
value possibly due pace at which the nodes were greatly improved due to gradual increase in speed node but
communicating. In terms of Boundless Simulation Area the Random Walk Mobility exhibited unreliable and different
Mobility, as the speed of the node increases a considerable values for Packet Delivery Ratio with regards to speed
drop in the throughput is observed. variations.

43
 International Journal of Computer Applications (0975 – 8887)
Volume 176 – No. 32, June 2020

iv. Throughput for Mobility Models with Time vi. Packet Delivery Ratio for Mobility Models with Time

Fig. 16. Packet Delivery Ratio for Mobility Models

4.1 Conclusions/Recommendations
The three mobility models that is Random Waypoint Model,
Random Walk Model and Boundless Simulation Area
OMRA’s performance with the three mobility models over a Mobility Model and their impact was demonstrated through
certain time produced that were obtained for node speed of simulation using
50m/s (50 meters per second) and Mean Delay is
demonstrated in fig. 14,15 and 16 respectively. Throughput, End-to-End Delay, Packet Delivery Ratio and
Mean Delay are the major quantitative metrics that
In fig. 14 above, it is evident that, the Random Waypoint characterize network latency within a network. These network
(4330000 bps) and Random Walk (4430000) are relatively metrics have been examined through simulation.
similar with specific reference to Throughput whereas the
Throughput for Random Waypoint is comparatively stable. The output of the simulation results showed clearly that each
The Throughput for Random Walk portrays quite a steady mobility model was able to outperform the other two in
decline for a time period. However, the End-to-End Delay, as respect of one of the quantitative metrics mentioned above.
explained in fig. 15 below, is the lowest in terms of Random By considering the three parameters as a whole, it was clear
Waypoint, but the highest for Boundless Simulation Area that, the performance of Boundless Simulation Area seems
(3950000), especially when it is being used for the simulation. not to perform better than the remaining two mobility models.
This is because, it provided a better packet delivery, at the
From the results shown in fig. 16, random waypoint (0.8 secs) expense of lower Throughput and higher End-to-End Delay
seems to outperform the other two mobility models under and so far as Random Walk and Random Waypoint is
consideration. It rather it exhibited low throughput and high considered, OMRA especially with regards to Random
end-to-end delay. The Random walk value of 1 secs provided Waypoint Mobility is able to provide better Throughput with
better Packet Delivery Ratio than random waypoint mobility low End-to-End Delay.
model.
However, concerning Packet Delivery Ratio, the Random
v. End to End Delay for Mobility Models Walk is able to outperform that of the Random Waypoint
Mobility Model .It is also evident from the simulation results
that OMRA’s performance under the various quantitative
metrics differ from one type of mobility model to another.

5. REFERENCES
[1]. Perkins and Bhagwat, 1994] Charles E. Perkins and Pravin
Bhagwat, Highly dynamic destination-sequenced
distance-vector routing (DSDV) for mobile computers.
In ACM SIGCOMM'94 Conference on Communications
Architectures, Protocols and Applications, pp. 234-244,
1994.
[2]. Royer at al., 2001] Elizabeth M. Royer, P. M. Melliar-
Smith, Louise E. Moser. An Analysis of the Optimum
Node Density for Ad hoc Mobile Networks. Proceedings
of the IEEE International Conference on
Communications, Helsinki, Finland, 2001.
[3]. Patel, R.; Kamboj, P. Investigation of Network Simulation
Tools and Comparison Study: NS3 vs NS2. J. Netw.
Commun. Emerg. Technol. JNCET 2015, 5, 137–142
[4]. Jha RK, Limkar SV, Dr. Dalal UD, A Performance
Fig. 15. End to End Delay for Mobility Model with Time
Comparison of Routing Protocols (DSR and TORA) for

44
 International Journal of Computer Applications (0975 – 8887)
Volume 176 – No. 32, June 2020

Security Issue in MANET (Mobile Ad Hoc Networks), [11]. Rafiq M, Manoj Kumar, Ahmed S, Comparative
International Journal of Computer Applications, Special Analysis of Energy Efficient Routing Protocols in
Issue on “Mobile Ad-hoc Networks” MANETs, 2010, MANET, International Journal of Computer
pp. 78-83. Applications, 2013, 82(5), pp. 1-6.
[5]. Sunil Kumar, Mishra R, Krishna Kumar, Singh BM, An [12]. [Atsushi Iwata et al., 1999] Atsushi Iwata, Ching-Chuan
Inspection of Mobility Models Routine on Routing Chiang, Guangyu Pei, Mario Gerla, and Tsu Wei Chen.
Protocols in Ad Hoc Network, International Journal of Scalable routing strategies for ad hoc wireless networks.
Engineering and Innovative Technology (IJEIT), 2013, Tech. Rep., Department of Computer Science University
2(8), pp. 269-274. of California, Los Angeles, 1999.
[6]. Tonk G, Kashyap I, Tyagi SS, Performance Comparison [13]. M. Bergamo. System design specification for mobile
of Ad-Hoc Network Routing Protocols using NS-2, multimedia wireless network (MMWN). Technical
International Journal of Innovative Technology and report, DARPA project, 1996.
Exploring Engineering (IJITEE), 2012 1(1), pp. 53-57.
[14]. [Chiang et al., 1997] Ching-Chuan Chiang, Hsiao-Kuang
[7]. Singh P, Bhola A, Jha CK, Simulation based Behavioral Wu, Winston Liu, and Mario Gerla. Routing in clustered
Study of AODV, DSR, OLSR and TORA Routing multihop, mobile wireless networks with fading channel.
Protocols in Manet, International Journal of Computer Tech. Rep., University of California at Los Angeles
Applications, 2013, 67(23), pp. 23-26. Computer Science Department, 1997.
[8]. Nand P, Dr. Sharma SC, Performance Study of Broadcast [15]. [Chlamtac et al., 2003] Chlamtac, M. Conti and J. J.-N.
based Mobile Adhoc Routing Protocols AODV, DSR Liu. Mobile ad hoc networking: imperatives and
and DYMO, International Journal of Security and its challenges. Ad Hoc Networks, Vol. (1), pages 13–64,
Applications, 2011, 5(1), pp. 53-64. 2003.
[9]. Subramani A, Krishnan A, Node Mobility Tracking in [16]. V. Davies. Evaluating mobility models within an ad hoc
Mobile Ad-Hoc Networks in their Geographical Position network. Master’s thesis, Colorado School of Mines,
(Dynamic Networks), International Journal of Soft 2000.
Computing and Engineering (IJSCE), ISSN: 2231-2307,
2011, 1(5), pp. 36-40. [17]. D.S.Tan, S. Zhou, J. Ho, J.S. Mehta, H. Tanabe. Design
and Evaluation of an Individually Simulated Mobility
[10]. Shukla AK, Jha CK, Sharma D, An Estimation of Model in Wireless Ad Hoc Networks. Communication
Routing Protocols in Mobility Models Used for Ad Hoc Networks and Distributed Systems Modeling and
Networks: Simulation Study, International Conference Simulation Conference, San Antonio, 2002.
on Advances in Computer Application (ICACA-2013),
Proceedings Published in International Journal of [18]. M. Dorigo, M. Birattari and T. Stutzle, "Ant colony
Computer Applications (IJCA) (975-8887), pp. 21-27. optimization," in IEEE Computational Intelligence
Magazine, vol. 1, no. 4, pp. 28-39, Nov. 2006, DOI:
10.1109/MCI.2006.329691

IJCATM : www.ijcaonline.org 45