Zone Based Ant Colony Routing in MANETS

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Zone Based Ant Colony Routing In Mobile

Ad-hoc Network
Maumita Bandyopadhyay, maumita.bandyopadhyay@tcs.com. Parama Bhaumik,
parama@it.jusl.ac.in. Dept of Information Technology. Jadavpur University. Salt Lake Campus.
Block-LB, Plot-8, sector-III, Kolkata-700098, India.

forwards packets towards the packet’s destination node.


Abstract— Ant colony optimization (ACO) is a stochastic Wireless sensor networks are characterized by frequently
approach for solving combinatorial optimization problems like changing network topology, multi-hop wireless connection
routing in computer networks. The idea of this optimization is and the need of dynamic, efficient routing protocols.
based on the food accumulation methodology of the ant In this paper we propose Zone based Ant Colony routing
community. Zone based routing algorithms is build on the algorithm using Cluster in mobile ad-hoc network. This
concept of individual node’s position for routing of packets in
algorithm uses the concept of clustering in mobile ad-hoc
mobile ad-hoc networks. Here the nodes’ position can be further
utilized to discover routes by the Ants in optimized way. Position network. Those clusters will be controlled by cluster heads. In
based routing algorithms (POSANT) had some significant this paper we show zone based ant colony routing algorithm
loopholes to find route (source to destination) like it never using cluster is more efficient than POSANT routing
guarantees the route would be the shortest one, in cases while it is algorithm by comparative Overhead study of POSANT and
able to find it. On the contrary, routing algorithms which are Zone based Ant using clustering concept with respect to
based on ant colony optimization find routing paths that are close varying Node Number, Zone Size and Mobility.
in length to the shortest paths. The drawback of these algorithms
is the large number of control messages that needs to be sent or
the long delay before the routes are established from a source to a
destination. II. OVERVIEW OF ANT COLONY ROUTING ALGORITHM
Here In this paper we have used Zone based ANT colony using Ant colony algorithm (ACO) is distributed and adaptive in
Clustering which assures to find shortest route using the DIR finding shortest paths from source to destination nodes while
principle (In this principle, the source or intermediate node also constructing a low cost overlay routing network. ACO is
transmits message to several neighbors and the node whose a stochastic approach for solving combinatorial optimization
direction is closest to the direction of destination gets selected as
problems like routing in computer networks. The idea of this
the next hop forwarding node.) together with minimum overhead
for route discovery and mobility management. Unlike other Zone
optimization is based on the observation of how ants optimize
based approach, in clustering it is not required to consider zone food gathering in the nature. Each node can take itself some
related information of each node while finding shortest path. stochastic decision based on some specific attributes supplied
Here, it is being proposed a new routing algorithm for mobile ad to the node [7].
hoc network by combining the concept of Ant Colony approach
and Zone based routing approach using clustering to get shortest Ant colony algorithms get there inspiration from the
path with small number of control messages to minimize the behavior of real ants gathering for food. Individual ants act as
overhead. Simulations show that Zone Based ant colony routing simple agents that hunt for food by following pheromone trails
algorithm has relatively short route establishment overhead than and depositing pheromone along the path taken. An ant is
other zone based ant colony algorithms in highly mobile more likely to follow a trail with a high concentration of
scenarios. pheromone. Pheromone on shorter paths gets increasingly
reinforced as more ants follow the higher concentration of
Index Terms— Ad-hoc Network; Ant Colony; Zone; Cluster; pheromones. This is due to more ants being able to travel a
Routing; POSANT; Max-Min D. shorter path than a longer path over a given period of time [1].

A. POSANT Routing Algorithm


I. INTRODUCTION
In position based routing algorithm the destination node is

W ireless Sensor Network is a collection of wireless,


uniquely addressable sensor devices which dynamically
form a temporary network, without using any existing
known and addressed by means of its location. Routing is
performed by a scheme that is based on this information,
which is generally classified as position-based scheme [2] [8].
network infrastructure or centralized administration. Each
node in the network effectively becomes a router, and POSANT is ant colony optimization based routing
algorithm which uses location information to improve its
Copyright Clearance Center (CCC). 978-1-4244-5489-1/10/ $26.00 © 2010
efficiency. POSANT is able to find optimum or nearly
IEEE optimum routes when a given network contains nodes.
Zones in POSANT: of pheromone trail of its outgoing link. As discussed above,
Consider a destination node D and a network graph G. For value of pheromone trail of zone 1 will be greater than that of
each node S (i.e. S is not necessarily the source node) we other zone. So ant will follow the zone 1 go get the shortest
partition its neighbors into 3 zones called zone1, zone2 and path. Similarly, ants of zone 2 and zone 3 will follow the path
zone3. Consider a line segment between S to D. For a of 2nd and 3rd priority.
neighbors H of S, angle θH is defined as the angle between line In most cases, the ant of zone 1 will establish the shortest
segments SH and SD. Node H belongs to zone1 if θH ≤ π/4, path in less overhead. But sometimes may happen that due to
zone2 if π/4 < θH < 3π/4, and zone3 if 3π/4 ≤ θH ≤ π, see Fig 1 some breakage in the path, ant of zone 1 may not find the
[3]. destination. At that time zone 2 and / or zone 3 ant may get the
path. As an example, in the Figure below, according to the
POSANT concept, ant of zone 1 chooses A as next hop and it
follows path Source -> A -> B -> L. But the path comes to a
dead end and ant 1 lost because it already traverse node B and
no other path is there from node l. But the ant of zone 2 gets
the path from source to destination.

Fig. 1. (a) θH is the angle between SH and SD. (b) Different zones of network
N for destination node D

Zone Selection Algorithm of POSANT:


if (abs(θH) >= z5)
θH = abs(θH) - z5;
if(((abs(θH) < z1) && (abs(θH) >= 0.0)) || ((abs(θH) > z4) && Fig. 2. Ant of zone 1 fails to find the destination. Ant of Zone 2 finds the
destination node. So we need to send ants in all three zones to increase the
(abs(θH) <= z5))) probability of finding out the destination node
ret = 1;
elsif (((abs(θH) < z2) && (abs(θH) >= z1)) || ((abs(θH) > z3)
In addition to the pheromone trail table discussed before,
&& (abs(θH) <= z4)))
each node maintains another table which we call Back Routing
ret = 2;
(BR) table. Whenever a forward ant enters a node from one of
elsif((abs(θH) <= z3) && (abs(θH) >= z2))
its neighbors, an entry in the BR table will be created that
ret = 3;
stores the identifier of the neighbor which the forward ant is
return ret;
coming from, the sequence number of the ant and the
identifier of the destination. Repeated forward ants will be
Where, z5 = 2π, z4 = 7π/4, z3 = 5π/4, z2 = 3π/4, z1 = π/4.
destroyed. When a forward ant reaches the destination, it is
destroyed and a backward ant is sent back to the source. This
Route establishment in POSANT: backward ant has the same sequence number as the
Consider a set of data packets coming to source node S
corresponding forward ant and traverses the same path to the
where the destination address is D. To establish a route, S
source using the information stored in BR tables. Moving
launches n forward ants with unique sequence numbers for
from node B to node A, the backward ant increases the
each zone (3n ants). Assigning very large value to n increases
amount of pheromone stored in AB. An evaporation process
the overhead of the algorithm without any significant
causes the amount of pheromone deposited in each link to
improvement. Similar to other ACO routing algorithms, at
decrease as the time passes on [3].
each node a forward ant makes a stochastic decision which is
based on the values of pheromone trails to select the next hop.
Limitations of POSANT:
The values of pheromone trails are stored in a table called
• Robustness is a problem in position based ant colony
pheromone trail table at each node according to the zone
algorithm. The use of position of destination node causes
value of the particular node. Suppose that a forward ant is
problem in terms of reliability. The accuracy of destination
currently residing in node S and this node has k neighbors H1,
position is another important problem to consider. For static
H2,...,Hk, Among them let H1, H2,..., Hi are in Zone 1, Hi1,
network the problem is straightforward, while the problem of
Hi2,...,Hj are in Zone 2 and Hj1, Hj2…. Hk are in Zone 3. S
designing location updates schemes to enable efficient routing
launches 3 ants in each zone. The ant in zone 1 will take
in mobile ad hoc network appears to be more difficult than
stochastic decision to find the next hop depending upon value
routing itself.
• In this type of algorithm methods should be incorporated those clusters. To create static zones two phase of the
to find out the positioning error which may cause mistake to algorithms are being followed, they are:
identify the destination node [5]. Phase-I: Max-Min D–Cluster Formation algorithm:
Step I: At some common epoch each node initiates 2d
B. Zone based Routing Algorithm using Cluster rounds of flooding of information exchange (node id) where d
Concept of clustering needs grouping of nodes in the is the given heuristic. In this algorithm, flooding occurs once,
network. This grouping depends upon transmission range and at the time of initial cluster formation. Each node maintains a
number of hop in a group. Each node group will have a group logged entry of two arrays, WINNER and SENDER to store
head called Cluster head having the responsibility of the results of each flooding round.
communication among its member nodes and other cluster Step II: Initially each node sets its winner to be equal to its
heads. Cluster head should contain address of its member own node id.
nodes as well as that of other cluster heads. Member nodes Step III: This is the phase for FLOODMAX where a node
need to store address information of their cluster head and chooses the largest value among its own WINNER array and
neighbor nodes. When information needs to pass from one this process continues for d rounds.
node to another, member node sends this information to its Step IV: This FLOODMIN phase follows FLOODMAX
corresponding cluster head, which decides whether the where a node chooses the smallest rather than the largest value
destination is a member or not. If yes, it directly sends the as its new WINNER to reclaim some of their territory.
Step V: After these two d rounds of information exchange a
information to destination. If no, it sends the information alone
node is able to determine its cluster-head [6].
with the destination node to other cluster heads which then
Phase-II: Zone Formation algorithm:
start to search in their own cluster [4].
Step I: The cluster head broadcasts get_Position_forAll ( )
request message along with its own GPS to get percolated
within d hop.
Step II: All member nodes in turn unicast back the message
node_GPS ( ) to the cluster-head using geographical routing.
Step III: Cluster-head receives all the GPS values of its
members and calculates the maximum limiting coordinates for
Left, Right, Up and Down values to define its boundary.
Step IV: The cluster head then broadcasts the message
get_Boundary_values ( ) within the d hop transmission range
to notify all the member nodes about the cluster boundary.
Step IV: All the member nodes become alerted about the
rectangular cluster boundary information, which can be
verified while changing their positions [6] and these
geographical formed boundaries are considered as zones.
Fig. 3. Network after phase-I Clustering. Here the whole network has been
divided into four clusters each with a cluster head
Phase-III: Mobility Management of Cluster Heads
through Selection of Surrogate Heads:
C. Ant Colony Routing Algorithm with Zones In ad hoc networks, because node can move arbitrarily, any
Concept of Ant Colony algorithm is merged with zone node can leave its cluster boundary. The problem of node
based (clustering) algorithm to form ant colony routing mobility is much more alarming when the cluster head itself
algorithm with zones. This algorithm will provide advantage becomes mobile and thereby generating the necessity of
of both ant colony and zone based algorithm. Like ant colony periodic re-clustering.
algorithm, here we need not store large routing tables in In the cluster management protocol any node including the
nodes, we need to store only neighboring node information cluster head automatically gets alarmed while crossing the
and previous traversed node information. As nodes in mobile geographical boundary of a cluster using the cluster
ad-hoc network will have memory of small storage capacity, it management protocol. Thus it is quite easy for a departing
would be tough to store large routing table inside each node. cluster head to make a timely arrangement for a surrogate
In our zone based ant colony algorithm there will be cluster head. The cluster head can select any of its current cluster
head available within each zone. As clustering concept, only members for delegating the cluster headship and thus can
cluster head node need to store path information of its member handover the entire cluster head responsibility to facilitate the
nodes as well as zone boundary information for all other process of data communication. This entire process of using
clusters in the network. surrogate head highly reduces the need for cluster head re-
election and there by decreasing the network traffic load
Initial Cluster formation: involved.
At the time of initialization of this algorithm, transmission In the Zone based Ant colony routing algorithm flooding
range and number of hop (d) should be mentioned so that occurs only once that is, at the time of initial cluster formation.
cluster can be formed and cluster heads can be selected for As the geographical cluster boundary once defined is static
information and is kept inside the cluster heads there is no
need to redefine the cluster boundaries again during the life
time of the network. When any cluster head crosses its
boundary, surrogate head will be chosen [15] to take the
responsibility of the departing cluster head. Thus there is no
requirement of repetition of flooding for the purpose of re-
clustering. So, at the time of overhead calculation we need to
consider flooding overhead once that is, only during the initial
cluster formation.
D. Route Establishment Procedure of Zone based ANT
Colony:
Consider the source node S in Figure 4, has a set of data
packets which should be forwarded to the destination node D.
Every member node keeps positional information of its
neighboring nodes, cluster boundary and cluster heads. S will
Fig. 4. Zone based Ant Colony Routing using Clustering concept. Here,
contact with its Cluster head to confirm whether the
source and Destination nodes are situated in separate clusters.
destination node is within the same cluster. Here, we need to
mention that, each cluster head will maintain a table
containing information of its member nodes. At the step-V of E. Route establishment Algorithm of Zone based ANT
the algorithm Max-Min d-cluster formation algorithm, Colony:
member node information comes to each cluster head and at Let us assume that, Source S has message M to send the
the same time each member nodes become aware of its cluster destination node D. Then the Route establishment Algorithm
head. So, when a set of data packet comes to a source node, steps are:
source node contacts with its cluster head to consult if the 1. S unicast data-packet DP to ClusterHead(S) [Data
destination node is in the same cluster. If yes, then data packet Packet DP consists of Destination Node address + Message M
is directed towards the destination node through the cluster which need to be delivered to D].
head. 2. CH(S) searches its Member List Table to see if D is a
If the destination node is not in the same cluster of the member of CH(S). [Member List Table is maintained by each
source node, then cluster head of the source node creates n - 1 cluster head which consists of its member nodes’ id and
number of forward ants with unique sequence number and address information].
send those ants to each and every reachable cluster heads. 3. If CH(S) finds that D is its own member, send
Here n is the number of reachable clusters from the source’s feedback to S and deliver the message M to D.
cluster head including its own cluster. As source cluster head 4. If CH(S) finds that D is not its own member, CH(S)
send the forward ants after conforming that the destination generates (n-1) forward ants with unique sequence numbers to
node is not in its cluster, total number of forward ants are n – send them to (n-1) reachable Cluster heads through multi-hop
1, that is, excluding source cluster itself. Whenever a forward paths. [n is the number of reachable clusters from CH(S). As
ant enters a node from one of its neighbors, an entry in the CH(S) should not send forward Ant to itself, so (n-1) Ants
Backward Routing table will be created that stores the should be generated].
identifier of the neighbor which the forward ant is coming 5. Each forward Ant, which consists of unique sequence
from, the sequence number of the ant and the identifier of the number, destination Id and message M, is forwarded towards
destination. boundary nodes of Cluster(S).
After getting destination id from the forward ants, each 6. From these boundary nodes, forward Ants are sent to
cluster head start searching the destination id within its those neighbor nodes, which are boundary nodes of other
member list. If any of the cluster heads finds that the clusters.
destination id is its member node, then it destroys the forward 7. From this boundary nodes of other clusters, forward
ant and creates backward ant. This backward ant has the same ants travels to the Cluster head of other Clusters.
sequence number as the corresponding forward ant and 8. When forward Ant comes to a Cluster head, sequence
traverses the same path to the source using the information number of this forward ant is stored in this Cluster head to
stored in Backward Routing tables. Moving from node B to keep track of duplicate entry.
node A, the backward ant increases the amount of pheromone 9. Each Cluster head checks the sequence number of the
stored in AB. An evaporation process causes the amount of forward ant and see if it already has the sequence stored in it.
pheromone deposited in each link to decrease as the time If no, it searches its Member List Table to see if D is its
passes on. member. Otherwise it kills the forward and as it has already
searches for this ant. Thus this algorithm avoids loops.
10. When any of the Cluster heads finds that D is its
member, it kills the forward Ant and generates Backward Ant
with same sequence number.
11. Cluster head which has D as its member node,
delivers the message M to D.
12. Then sends back the Backward Ant to the Source
node S as a feedback.
13. When a Cluster head finds that D is not its member, G. Loop Avoidance of Zone based ANT Colony Algorithm:
it sends the forward ant to its neighbor cluster to search. In Zone based ANT Colony Algorithm, each node (Cluster
heads as well as member nodes) stores the sequence number
F. Route establishment Flow Chart of Zone based ANT of the forward ant, when the ant traverses through these nodes.
Colony: This sequence number remains stored in this node until the
algorithm terminates. When ever the same forward ant again
comes to the same node, the node recognizes the ant by
checking the sequence number and kills the forward ant to
avoid redundancy. Thus Zone based ANT colony Algorithm
avoids looping around of the ants in the network.

III. COMPARATIVE STUDY OF OVERHEAD OF POSANT WITH


THAT OF ZONE BASED ANT COLONY ROUTING
In POSANT algorithm the overhead is coming from three
paths that have been considered there between source to
destination node. As per our previous discussion, path of Zone
1 should be shortest path, path of Zone 2 will be longer than
that of Zone 1 but will be shorter than path of Zone 3. Total
overhead in POSANT should be sum of overhead of all three
paths.
In Zone based Ant Colony, if source and destination node
are not in the same zone then cluster head of the source node
will send ants to all other cluster heads to find the destination
node. So, if n number of zones formed in the network, n
number of cluster heads will be there and (n – 1) number of
ants will be sent by source cluster head to all other cluster
heads, so total over head will be sum of overhead of all (n-1)
paths. Here we can decrease the number of paths by
decreasing the number of zones that is, by increasing the
number of hops in cluster formation algorithm (d), where d is
an input parameter (as transmission range of each node is
fixed, we can not decrease number of clusters by increasing
the transmission range of nodes.) Decreasing zone (cluster)
number may cause less overhead by reducing number of paths,
but that will cause maintenance of zones more difficult. So, to
get the best performance and less overhead from zone based
ant routing, number of zone-forming-hops should be
optimized.

A. Overhead Comparison between POSANT and Zone based


ANT colony using Clustering with varying Zone size:
In Zone based ANT colony using clustering, we can increate
the Zone size by increasing the HOP value of cluster value.
Thus, by increasing zone size we can reduce overhead of Zone
Fig. 5. Route establishment flow chart based ANT colony. Because, for a fixed network area, when
we are increasing each zone size, number of zones in this
network area be reduced. By reducing number of zones we can
Abbreviation used in the Flow Chart: reduce the number of forward ants which will be sent from
DP: Data Packet. source cluster head to all other reachable cluster head to find
CH(S): Cluster head of Source Node. out the destination node’s cluster head. So, in other word we
FAnt: Forward Ant. can reduce overhead of forwarding ants in zone based ANT
Seq no: Unique sequence number of ant. colony by increasing HOP value. Though it will not affect
Cnt: Counter. When forward Ant comes to a CH, this overhead of POSANT as it does not depends upon HOP value.
counter should be increased by 1. Fig. 6 shows the comparison table and graph between
BN: Boundary Notes. POSAND and Zone based ANT colony with varying HOP
T: Total number of cluster heads. value.
C. Overhead Comparison between POSANT and Zone based
4000
3500
ANT colony using Clustering with varying Mobility:
3000 In the mobile ad hoc network, all nodes are mobile. Here we
ZONE BASED
2500 have considered random mobility in four directions. Some
Overhead

OVERHEAD
2000 nodes are moving towards east, some towards west, some
1500 POSANT
OVERHEAD towards north and some towards south. At the time of
1000
500
overhead calculation, mobility will increase the overhead of
0 nodes as there will be number of route fails when the node
2 3 4 5 6 7 mobility increases considerably in the network. As we
HOP Value increase the mobility of nodes, overhead will be increased
accordingly. For both POSANT and Zone based ANT colony,
Fig. 6. Zone size Vs Overhead: Zone based Overhead decreases when Zone overhead will be increased according to the increment of
size increased i.e. HOP value increases
mobility. But in case of POSANT, the increment of overhead
due to increment of mobility will be much greater than of
B. Overhead Comparison between POSANT and Zone based Zone based ANT colony.
ANT colony with varying Node Number: In case of Zone based ANT colony, when the destination
In a specific range of a network, we can increase the number node will leave its zone due to mobility and enter into a new
of nodes which will increase the density of node in the zone, previous cluster head will reply back to the source’s
considered region. When we increase number of nodes, cluster head that destination node is not its member node.
overhead will also be increased because to traverse from one After getting this message, source cluster head will forward
side to another side of this network, ant needs to traverse more ants towards all reachable cluster heads to know the current
nodes. In Zone based ANT colony when number of nodes will zone of destination node. Where as in POSANT, source node
be increased, more number of nodes will be there in each zone will come to know the location change information of the
and will increase zone concentration and will be taken care by destination node (or any other node in the path from source to
corresponding cluster head. As here HOP value D is fixed, destination) after a certain period (time out period) when the
number of zones and zone size will not be increased. So, sender of the source node will not get any acknowledgement
number of ants traversing from one cluster head to another message from the receiver of destination node. After realizing
cluster head will not be increased. Only concentration of node that, source node again have to start POSANT algorithm by
within each zone will be creased. It may increase sending ants in three zones as mentioned above. As POSANT
responsibility of each cluster head but at the same time algorithm needs more overhead to execute than of Zone based
increment of overhead will be less than that of POSANT. In ANT colony, mobility factor affect POSANT much more than
POSANT when number of nodes is increased ANT needs to Zone based ANT colony.
traverse more number of nodes to reach the destination node. Fig. 8 shows the graph representing the comparison between
Fig. 7 shows the graph representing the comparison between POSAND and Zone based ANT colony with varying node
POSANT and Zone based ANT colony with varying node number. Simulation shows that increment of mobility
number. Here the simulation shows that, POSANT Overhead increases overhead. Rate of increment of POSANT overhead
increases when Number of Nodes increases. Zone based is much more than that of Zone based overhead.
overhead also increases with increment of Number of Nodes,
but rate of increment is less than of POSANT. 40000
35000
30000
7000 25000 ZONE BASED
Overhead

OVERHEAD
6000 20000
POSANT
5000 15000
ZONE BASED OVERHEAD
10000
Overhead

4000 OVERHEAD
5000
3000
POSANT 0
2000 OVERHEAD 4 8 12 16 20 24
1000 M obility
0
50 100 150 200 250 300 Fig. 8. Mobility Vs Overhead: POSANT Overhead increases when Mobility
Number of Nodes increases. Zone based overhead also increases with increment of Mobility, but
rate of increment is less than of POSANT

Fig. 7. Number of Node Vs Overhead: POSANT Overhead increases when


Number of Nodes increases. Zone based overhead also increases with D. HOP Count Comparison between POSANT and Zone
increment of Number of Nodes, but rate of increment is less than of POSANT based ANT colony with varying Number of Nodes:

When we increase number of nodes in a specific range of


network, number of HOP will also be increased because to
traverse from source to destination, ant has to traverse more
nodes. In Zone based ANT colony when we increase number
50
of nodes, only concentration of zone will be increased. As 45
here HOP value is fixed, zone size will not be increased. 40
35
Number of zones will also be unchanged. So, number of ants ZONE BASED

HOP Count
30 HOP COUNT
traversing from one cluster head to another cluster head will 25
POSANT HOP
20
not be increased. As a result HOP count will not be increased 15 COUNT
for entire network, only for the specific zone where the 10
5
destination node exists, HOP count will be increased. As zone 0
concentration will be increased, ant has to traverse more 4 8 12 16 20 24
number of nodes within the zone of destination node. All M obility
cluster heads other than destination’s cluster head will not
forward ants to any of their member nodes because cluster Fig. 10. Mobility Vs HOP Count: HOP Count increases with increment of
head will have the information that destination node is not Mobility. But the increment rate is greater in POSANT than Zone based Ant
present in its zone. Whereas in POSANT, increment of
number of nodes will result increment of HOP count for the
F. HOP Count Comparison between POSANT and Zone
entire network. As a result ant has to traverse more number of
based ANT colony with varying HOP value:
HOP. So it is very obvious that, increment of number of nodes
in a network will cause much more HOP count increment in In Zone based ANT colony, zone size can be increased by
case of POSANT than of Zone based ANT colony. incrementing HOP value. In a fixed boundary network,
Fig. 9 shows the graph representing the comparison between increment of zone size means lesser number of zones in this
POSANT and Zone based ANT colony with varying node network. So, when we increase HOP value of a network,
number. number of zones decreases. Lesser number of ants needs to
traverse to each zone to find out the destination nodes and thus
HOP count will be decreased. But in case of POSANT
35
algorithm, increment of HOP value does not have any effect
30
on HOP count because POSANT algorithm does not use HOP
25
ZONE BASED value.
HOP Count

20 HOP COUNT Fig. 11 shows the comparison table and graph between
15
POSANT HOP POSANT and Zone based ANT colony with varying Zone
10 COUNT size. Simulation shows that HOP value does not affect HOP
5
Count of POSANT, but by incrementing HOP value, HOP
0
count of zone based ant colony can be decreased.
50 100 150 200 250 300
Number of Nodes
30
Fig. 9. Number of Node Vs HOP Count: HOP Count increases with
25
increment of Number of Nodes. But the increment rate is greater in POSANT
than Zone based Ant 20 ZONE BASED HOP
HOP Count

COUNT
15
POSANT HOP
E. HOP Count Comparison between POSANT and Zone 10 COUNT

based ANT colony with varying Number of Nodes: 5

0
In mobile ad-hoc network, position of nodes is changed due 2 3 4 5 6 7
to mobility. This will increase HOP count as source, HOP Value
destination and intermediate nodes’ location will be changed
due to mobility. When mobility will be increased, position Fig. 11. HOP value Vs HOP Count: For Zone based ant colony, HOP Count
change also will be more and that will force ants and data decreases with increment of HOP Value, i.e. Zone size. But HOP Count of
POSANT does not vary with increment of HOP value
packets to traverse more HOP to reach the destination node.
In Zone based ANT colony, while location of destination
node or any other member node changes, Cluster head gets G. Delay Comparison between POSANT and Zone based
this information and guide ants to go to right direction. Where ANT colony Algorithm:
as in case of POSANT, when position of node changes due to In a network, if the diameter of the network increases or the
mobility, ants need to start searching in the three zones again density of node increases, ant or data packet needs to traverse
from the scratch. So, increment of mobility causes much more more number of HOP to reach the destination and that causes
increment of HOP count in case of POSANT than Zone based increase in processing delay at each node. Thus the overall
ANT colony. delay in delivering the data packets gets increased. So we can
Fig. 10 shows the comparison table and graph between say that delay is proportional to HOP Count of a network.
POSANT and Zone based ANT colony with varying mobility. Simulation results of HOP count can throw light on the
variation of delay in the network. If number of node increases,
delay will also be increased accordingly. Delay also increases
with increment of Mobility. We have shown that using our
Zone based Ant colony routing algorithm we can optimize this
delay by making the zone size larger. But Delay of POSANT
105
will not be affected much as the zone size or number of hops
can not be customized. 100

Throughput (%)
ZONE BASED
95 Throughput

H. Throughput Comparison between POSANT and Zone 90 POSANT


Throughput
based ANT colony with varying Number of Nodes:
85
In a network of fixed boundary, number of node increment
80
causes increment of density of the network. When density of
4 8 12 16 20 24
network increases, more nodes come within 1-hop position of
M obility
source node or any node in the source to destination path. This
decreases probability of data packet loss as data packet will be Fig. 13: Mobility Vs Throughput: Throughput decreases with increment of
carried forward by more number of nodes while node number Mobility. Decrement rate of POSANT is greater than Zone based Algorithm
increases. In case of POSANT, when number of node
increases, concentration of the whole network increases. In J. Throughput Comparison between POSANT Zone based
case of Zone based Ant colony, when number of node ANT colony with varying HOP value:
increases, number of zones and zone size will remains same.
But concentration of each zone will be increased. So when Ant In Zone based ANT colony, zone size can be increased by
traverses in a zone with higher concentration, the probability incrementing clustering HOP value. In a fixed boundary
of getting lost will be decreases. That increases throughput. network, increment of zone size means lesser number of zones
Fig. 12 shows the comparison table and graph between in this network. So, when we increase Clustering HOP value
POSANT and Zone based ANT colony with varying node of a network, number of zones decreases. Data packets need to
number. Simulation shows that throughput of both POSANT traverse lesser number of zones. But inside single zone they
and zone based ant colony increases with increment of node need to traverse more nodes as zone size increases. So
number. throughput will not change much by changing clustering HOP
value. In case of POSANT algorithm, increment of clustering
HOP value does not have any effect on throughput because
101.0 POSANT algorithm does not use clustering HOP value.
99.0
Fig. 14 shows the comparison table and graph between
97.0
POSANT and Zone based ANT colony with varying
Throughput (%)

95.0 ZONE BASED

93.0
Throughput clustering HOP value.
91.0 POSANT
89.0 Throughput
110
87.0
108
85.0
106
50 100 150 200 250 300 104
Throughput (%)

ZONE BASED
Number of Nodes 102
Throughput
100
98
Fig. 12. Number of Node Vs Throughput: Throughput increases with POSANT
96 Throughput
increment of Number of Nodes 94
92
I. Throughput Comparison between POSANT and Zone 90
based ANT colony with varying Mobility: 2 3 4 5 6 7

In mobile ad-hoc network, position of nodes is changed due HOP Value

to mobility. This will decrease throughput. Location of source, Fig. 14. HOP value Vs Throughput: Throughput does not vary much with
destination and intermediate nodes will be changed due to increment of HOP Count , i.e. Zone size.
mobility. Position change of node is directly proportional with
mobility of the network. When mobility increases, position
K. Advantage of Zone based ANT colony over POSANT:
change also increases. Due to position change of nodes, data
packets need to discover the direction of destination again and In POSANT routing, like all other position based routing,
that decreases throughput. source node need to know the position of the destination node.
In Zone based ANT colony, while location of destination Robustness is a problem in position based ant colony
node or any other member node changes, Cluster head gets algorithm. The use of position of destination node causes
this information and guide ants to go to right direction. Where problem in terms of reliability. The accuracy of destination
as in case of POSANT, when position of node changes due to position is an important problem to consider. For static
mobility, ants need to start searching in the three zones again network the problem is straightforward, while the problem of
from the scratch which causes more data loss. So, increment designing location updates schemes to enable efficient routing
of mobility causes much more decrement of throughput in in mobile ad hoc network appears to be more difficult than
case of POSANT than Zone based ANT colony. routing itself.
Fig. 13 shows the comparison table and graph between Overhead of Zone based ANT colony is much less than that
POSANT and Zone based ANT colony with varying mobility. of POSANT, as we can see in the previous section, Increment
of overhead due to increment of node number and mobility is V. CONCLUSION AND FUTURE WORK
much more in POSANT than that of Zone based Ant system. Zone based routing using ant colony optimization aims to
Decrement of overhead due to increment of HOP value is support zone based routing with minimum routing overhead
much more in Zone based ant system than that of POSANT. for mobile ad hoc networks. The idea behind this zone based
HOP Count of Zone based ANT colony is less than of routing is to allow the nodes within each zone to get referred
POSANT, as we can see from the results of simulations on with the zone boundaries and control the node. Ant Colony
varying Number of nodes, Mobility and Zone Size (HOP technique merged with Zone based technique to optimize
value). memory utilization, overhead and throughput.
Throughput of Zone based ANT colony is better than of
POSANT, because in Zone Based Algorithm, data packets In zone based routing, each zone member is aware of its
need not traverse to each individual node. Cluster heads will mobility using the zone boundary values. It allows the node to
decide whether data packets should traverse to a particular inform about their mobility and thereby enabling the protocol
zone or not. So chances of loss are less than of POSANT. to find the mobile destination quickly with minimum
overhead. A cluster head is available within each zone to
perform data transmission and routing. The performance
IV. RELATED WORK analysis shows that, in highly mobile network, Zone based ant
Inspired by insect societies' biological models, ACO colony reduces overhead, Hop value and increases throughput
provides a simple and efficient routing solution. than POSANT in all three cases (varying node number,
Fernando [9] presented a MANET routing protocol, SARA mobility and Zone size). So in conclusion, we can say that
(Simple Ant Routing Algorithm), which used a controlled Zone based ANT colony is better than POSANT.
neighbor broadcast to find route. Compared the performance In Zone Based ANT Colony routing algorithm, we are
of ant-routing algorithm with Dijkstra's OSPF, Dhillon [10] always using shortest path for routing the data packets. Ants
illustrated that the performance of AntNet is better than OSPF.
always traverse the shortest path from source to destination
The more packets transmitting in the network, the more
through the source cluster head and destination cluster head
packet-loss and congestion will occur. These two sides are a
(using geographical shortest path forwarding technique). So
difficult dilemma for choosing.
Ant routing [11] can address the dilemma more efficiently here always the shortest path is selected for routing and load
than the other, which can make the best use of the network balancing concept has not been incorporated in this routing
resources and make little packet-loss and congestion. A algorithm. We may introduce load balancing by selecting
`swarm [12]' is a relatively simple agent, which can interact second shortest path sometimes in case we have large number
through simple mechanisms and solve complex problems. It of data packets to send so that the same set of nodes should
belongs to the subset of larger class of `Swarm Intelligence' not remain selected in the route path always. Thus load
algorithms. balancing can also be handled in Zone based Ant colony
Based on prediction model, a new routing algorithm called algorithm. This concept and the performance of the network
ant-based energy aware disjoint multi path routing algorithm will be evaluated in our future work.
[13] (AEADMRA) is proposed. Simulation results indicate
that performance of AEADMRA is much better than the REFERENCES
others. Based on load balancing of routing protocol, Zheng [1] Benjamin McBride, Caterina Scoglio, Sanjoy Das. “Ant Colony
[14] presented a cross-layer design and ant-colony Algorithm for Low Cost Overlay Network Routing”. 2061,
optimization (CALRA). In CALRA, ants deposit simulated Kansas State University. Distributed Biobjective
pheromones as a function of multiple parameters information [2] Silvia Giordano, Ivan Stojmenovic, Ljubica Blazevic. “Position
collected by each layer of each node visited. Based Routing Algorithms For AD HOC Networks”. ICA-
DSC-EPFL CH-1015 Lausanne (Switzerland), SITE, University
GLS (GRID Location Service) [16] provides locations of of Ottawa, Ottawa, Ontario K1N 6N5, Canada and DISCA,
nodes throughout a network in a distributed manner. GLS can IIMAS, UNAM, Direccion Circuito Escolar s/n, Ciudad
be queried for the location of different nodes in the network Universitaria, Coyoacan, Mexico D.F., Mexico, ICA-DSC-
using a unique ID of a node in the network. It is known that EPFL CH-1015 Lausanne (Switzerland).
there is a limitation of GLS to serve a metropolitan area, and [3] Shahab Kamali, Jaroslav Opatrny. “A Position Based Ant
Colony Routing Algorithm for Mobile Ad-hoc Networks”.
question whether any distributed location service can achieve Department of Computer Science and Software Engineering,
accuracy universally across a large area. Concordia University, Montreal, Canada.
Weak state [17] is a generalization of soft state that is [4] Alan D, Amis Ravi, Prakash Thai, H.P. Vuong, Dung T. Huynh.
characterized by probabilistic semantics and local updates. It “Max-Min D-Cluster Formation in Wireless Ad Hoc Networks”.
is interpreted as a probabilistic hint and not absolute truth. It Department of Computer Science University of Texas at Dallas
contains a measure of confidence in the state value, which is a Richardson, Texas 75083-0688.
measure of the probability that the state is valid. Nodes only [5] Silvia Giordano, Ivan Stojmenovic, Ljubica Blazevic. “Position
Based Routing Algorithms For AD HOC Networks”. ICA-
have partial information about the region a destination node is DSC-EPFL CH-1015 Lausanne (Switzerland), SITE, University
likely to be [18]. Even though the information weak state of Ottawa, Ottawa, Ontario K1N 6N5, Canada and DISCA,
contains is more consistent than that of strong state, it is not IIMAS, UNAM, Direccion Circuito Escolar s/n, Ciudad
exact. Universitaria, Coyoacan, Mexico D.F., Mexico, ICA-DSC-
EPFL CH-1015 Lausanne (Switzerland).
In this paper, I have a different approach to achieve goals
similar to GLS and WSR.
[6] Alan D, Amis Ravi, Prakash Thai, H.P. Vuong, Dung T. Huynh.
“Max-Min D-Cluster Formation in Wireless Ad Hoc Networks”.
Department of Computer Science University of Texas at Dallas
Richardson, Texas 75083-0688.
[7] Payman Arabshahi, Andrew Gray, Ioannis Kassabalidis,
Arindam Das, Sreeram Narayanan, Mohamed El-Sharkawi, and
Robert J. “Adaptive Routing in Wireless Communication
Networks using Swarm Intelligence”. Marks II, Jet Propulsion
Laboratory, University of Washington.
[8] S. Kamali, J. Opatrny. “Ant-Colony Routing in Position-aware
Mobile Ad-Hoc Networks”. Department of Computer Science
and Software Engineering, Concordia University, Montreal,
Canada.
[9] Fernando, C. and V. “Simple ant routing algorithm”. Teresa,
Proceeding of International Con-ference on Information
Networking, 486{493, Busan, South Korea, 2008.
[10] Dhillon, S. S. and P. Mieghem. “Performance analysis of the
AntNet algorithm. ACM Com-puter Networks”. The
International Journal of Computer and Telecommunications
Networking,Vol. 51, No. 8, 2104{2125, 2007.
[11] Manuel, T. A. and E. V. Carina. “Design of ant colony based
algorithm ant route for solve the OSPF problem”. Proceeding of
Fourth Congress of Electronics, Robotics and Automotive
Mechanics, 386{394, Sept. 2007.
[12] Purkayastha, P. and J. S. Baras. “Convergence results for ant
routing algorithms via stochas-tic approximation and
optimization”. Proceeding of 46th IEEE Conference on Decision
and Control, 340{345, Dec. 2007.
[13] Wu, Z. Y., X. J. Dong, and H. T. Song. “Ant-based energy
aware disjoint multipath routing algorithm in MANETs”.
Proceeding of 1st International Symposium on Pervasive
Computingand Applications, 752{757, Aug. 2006.
[14] Zheng, X. Q., W. Guo, and L. J. Ge. “A cross-layer design and
ant colony optimization based load balancing routing protocol
for ad hoc networks”. Journal of Software, Vol. 34, No. 7,
1199{1208, 2006.
[15] Parama Bhaumik, Somprakash Bandyopadhyay. “A Mobility
Tolerant Cluster Management Protocol with Dynamic Surrogate
Cluster-heads for A Large Ad Hoc Network”. Dept. of
Information Technology, Jadavpur. MIS group, Indian Institute
of Management, Calcutta, India.
[16] Aaron Beach, “GLS (Grid Location System). Performance
Observations & Summary”. Northwestern University.
[17] Utku G¨unay Acer, Alhussein A. Abouzeid, Shivkumar
Kalyanaraman. “An Evaluation of Weak State Mechanism
Design for Indirection in Dynamic Networks”. Rensselaer
Polytechnic Institute, Troy, NY, USA. IBM India Research
Laboratory, Bangalore, India.
[18] Chi-Han Lin, "Weak State Routing for Large Scale Dynamic
Networks". MobiCom2007.

You might also like