Content uploaded by Jay Prakash
Author content
All content in this area was uploaded by Jay Prakash on Jun 01, 2021
Content may be subject to copyright.
Received: April 12, 2017 235
International Journal of Intelligent Engineering and Systems, Vol.10, No.4, 2017 DOI: 10.22266/ijies2017.0831.25
Path Load Balancing Adaptive Gateway Discovery in MANET-Internet
Integration Using PSO
Jay Prakash 1*, Rakesh Kumar 2, Jai Prakash Saini 3
1,2Department of Computer Science & Engineering,
Madan Mohan Malaviya University of Technology Gorakhpur, India
3Department of Electronics and Communication and Engineering,
Bundelkhand Institute of Engineering and Technology, Jhansi, India
* Corresponding author’s Email: jpr_1998@yahoo.co.in
Abstract: Mobile Ad-hoc Networks (MANETs) find numerous applications. Its utility can be extended by
integrating it with Internet. Some real-life applications where such integrated networks can be used are disaster
situation viz., earthquake, battle fields etc. MANETs are dynamic topology, infrastructure less and standalone
networks of wireless mobile nodes. To overcome this limitation, MANET needs to connect to the Internet. In such
networks, mobile nodes do not have wired connection, so they do not follow any particular route to transfer their
packets. Thus, every time when topology changes, routing paths changes too. However, every node selects the
shortest path to route their packet. In that case, this path will be congested due to overload. We propose a load
balancing adaptive Internet gateway discovery approach which focuses on solving the problem of overload and
congestion in mobile ad-hoc domain. Particle Swarm Optimization is used for selecting an optimal path among all
paths available. Our work is to optimize this load balancing problem. We are using Particle Swarm Optimization
(PSO) concept to overcome this problem. In our approach, we use soft computing technique viz., working procedure
of PSO algorithm which searches the optimal paths using some mathematical function. Then, we combine shortest
path algorithm with the gateway discovery algorithm. Thus, the algorithm first selects the shortest path stored in the
memory and check whether it is free for routing packets or not, if the route is free or lightly loaded, then the source
node will transfer packets otherwise if the node is overloaded, then PSO searches for the next optimal neighbour or
node to route the packet. The new path is selected to route packets and stored in the place of current available path.
Proposed algorithm has been implemented in MATLAB. Our approach outperforms the existing approaches.
Keywords: Mobile ad hoc network, Path load, Swarm optimization, Shortest path, Performance.
1. Introduction
Ad-hoc networks [1] can be defined as the
network infrastructure where the mobile nodes are
connected to each other wirelessly. The mobile
nodes such as laptops, mobile phones, digital
devices and all the personal assistance devices.
These mobile devices communicate through internet
gateways for transferring data packets to each other
via wireless connection. Whenever a mobile node
wants to communicate or transfer data to another
mobile node it searches for gateways. When the
gateway responses for their request they transfer
their data via routers. Routers are also mobile device
which only perform the data transformation coming
from one node to another node. A router does not
only help to transfer the data packet from source
node to destination node. They always select the
shortest path to route the packet so that time delay
will be less and bandwidth will be more utilized.
These mobile nodes move their position frequently
thus the topology change very often in mobile Ad-
hoc networks. So, the path which is selected once to
route the packet can change even though the second-
time source and destination nodes are same or
different. Topology change is one of the main
Received: April 12, 2017 236
International Journal of Intelligent Engineering and Systems, Vol.10, No.4, 2017 DOI: 10.22266/ijies2017.0831.25
Figure.1 Architecture of Integrated Internet-MANET
characteristics of the Ad-hoc network. Fig. 1 shows
the architecture of mobile Ad-hoc network. For
transferring data packets, first it searches for all the
available paths, then select the shortest path from
source to destination node. As every node wants to
route their data from the shortest path only, thus the
path becomes congested one.
Our main concern is to reduce the load on the
path and balance the load on the Ad-hoc network.
We are using Particle Swarm Optimization [2] to
minimize the path load. Particle Swarm optimization
(PSO) is a pretending optimization technique which
is based on population based problem. Here, we are
using shortest path algorithm with the combination
of PSO.
Organisation of research paper is as follows. In
current section an overview of working culture is
presented. In section 2 literature review of existing
works is described. Basic concept of particle swarm
optimization is presented in section 3. Section 4
includes proposed work that contains algorithm for
path selection using PSO and also algorithm for
gateway discovery. In section 5 simulations and
result analysis is presented and finally in section 6
conclusion and future work are described.
2. Related work
There are various issues of load balancing in an
ad-hoc networks. In the following section, we
analyze many survey and research papers where
authors discussed and gave various solutions to deal
with load balancing problem.
The technique given by Zhou et al. [3] consists
of a dynamic GW (DGW) and a foreign agent with a
distance of one hop in them and the DSDV protocol
is used for data transmission. DGW solves the
problem of load balancing. The registration is done
in two ways proactive in this registration is between
DGW and FA and reactive in this in between DGW
and mobile nodes. The registration contains current
queue length and no. of nodes registered. The
selection of DGW and FA is emphasized by the load
balancing. This approach consists of load balancing
DGW selection. The DGW though improves the
process but also complicates it.
The technique proposed by Trivino-Cabrera et al.
[4], consists of an access router AR and the nodes in
its range are default gateway(DLW) and the nodes
away from it are the candidate gateway
nodes(CGW) along with these the reactive protocol
is AODV. The DGW broadcast MRA message for
advertisement and receives the MRS message from
other nodes who wants to connect. Whenever the
DLW becomes congested mobile nodes becomes the
CGW and is used to remove the congestion problem.
The drawback of this approach is single point failure.
The technique proposed by the Hsu et al. [5],
consists of two tier architectures. In first tier is
MANET provided with services from DCHP server
which acts as a gateway, this forms the lower level
of DSDV. The mobile nodes can also get the valid
IP address by DCHP server. The DCHP server
might not be in the MANET. There are two
categories based on DCHP server presence and
absence. The DCHP server may be present or absent
in MANET. The first is sub-MANET with DCHP
server and other is disconnected component without
DCHP. The second tier have heterogeneous cellular
infrastructure. The GW is connected with one base
station and DCHP acts as GW and registration is by
DCHP protocol. The GW selection is by minimum
load index routing (MLIR). This protocol decides
the gateway and mobile node based on load and
traffic and proves that range may expand or shrink
on traffic basis. The limitation is that the handoff
mechanism is not well supported.
The technique proposed by Shin et al. [6]
classifies the traffic in external and internal traffic.
If the both source and destination are in ad-hoc
network then it is internal traffic else external traffic.
This external traffic has two types- incoming and
outing so for it two GW are maintained. The GW
having least distance is default GW but congestion
may occur so registration is done to avoid it with
load balancing. The ad-hoc network uses reactive
routing protocol in which the node sends request and
GW sends the proxy route reply message (pRREP).
Received: April 12, 2017 237
International Journal of Intelligent Engineering and Systems, Vol.10, No.4, 2017 DOI: 10.22266/ijies2017.0831.25
The load balancing mechanism is combined with
delay of sending the p-RREP. The mobile node will
receive p-RREP of the GW. For receiving the
message reactive method is used to send solicit
message to GW and registration to it is done. For
measuring the load of GW, the
average_packet_queue_size (APQS) is used.
The technique proposed by Le-Trang et al. [7]
divides network into MANET domains each having
a GW which manages the network topology. GW
registration is necessarily done and GW broadcast
advertisement like in proactive scheme and the
selection of GW is on the basis of load on gateway.
Limitation is that, during the performance
evaluation based on average delay transmission
factor, AODV routing protocol should be used for
decrementing delay ratio.
In paper [8], authors have presented a congestion
adaptive multipath routing protocol for enhancing
the throughput and eliminates congestion in
MANETs. Here, the average load of an existing link
increases beyond a defined threshold. The available
bandwidth and residual battery energy decreases
below a defined threshold, traffic is distributed over
fail-safe multiple routes to reduce the traffic load on
a congested link. Demerit of this approach is that,
author estimate each threshold value individually for
link presence in network, on behalf of selecting a
single threshold for all link in network.
Pham et al. [9] suggested a wireless multi hop
network. In this, the Internet Gateway (IGW)
mechanism is adopted to obtain Internet
connectivity, linking the wireless network with the
global Internet. The efficiency of the protocol can
be enhanced if it balances the traffic among
available IGWs by which the network performance
can be optimized. In this proposed approach three
type of overhead faced such as tunneling overhead,
additional cost and last is selection of gateway,
which are not be consider by author. This is major
drawback of this approach.
In paper [10], authors have proposed a modified
version of AODV where a new gateway load
balancing strategy is based on a load balanced
version of AODV routing protocol. The network
provides Internet connectivity to mobile nodes
which form a mobile ad-hoc network by using a set
of Internet Gateways (IGW). With the aid of IGW,
mobile nodes can connect to the outside world and
wired node like correspondent Node (CN). Thus,
two protocols are run by IGW, IP for wired network
and modified-AODV for the wireless ad-hoc
network. Throughput and load balancing factor are
increased but due to considering a very small size
buffer the packet drop ratio is dramatically increased,
which is the drawback of this modified work.
In paper [11], authors have proposed a load
balancing scheme in ad-hoc networks. The proposed
mechanism is designed which is compatible with ad-
hoc on-demand distance vector routing (AODV)
protocol. In this mechanism, each nodes within a
MANET checks its queue occupancy continuously
in order to determine whether or not it should
respond against received route requests (RREQ).
This decision is totally relying on a threshold value
and this value for threshold is adjusted adaptively as
per of network load conditions. The limitation of
this scheme is that, there are variations among the
threshold value during the forward registration and
backward reply process.
The technique given by Gunes et al. [12] uses
Ant-Colony based Routing Algorithm that uses
route discovery by flooding as in case of AODV.
The duplicate packets are identified using sequence
number. When we found a route to the destination,
and we found backward ant like RREP in AODV.
The ant follows backward path with shortest trip
time by detecting the forward ant. The pheromone
amount deposited by ant is a function of route length.
The route maintenance is responsible for managing
the routing information.
Hu et al. [13] presents a review paper which
describes Particle Swarm Optimization Algorithm in
detail. It discusses the basic concept of PSO and all
the parameters used in it. Author details the
algorithm used in PSO and explain how the
algorithm searches the optimize value in the
dimensional space.
Zaman et al. [14] presents an integration of ad-
hoc network with the Internet through Internet
Gateway (IG). The IG offers Internet services to the
mobile nodes within ad-hoc domain. Prior to
connecting with Internet, there are two steps
involved. Firstly IG discovery and then registration.
Numerous strategies are defined for gateway
registration such as proactive, reactive and hybrid.
Now when mobile nodes communicate with the
gateway then due to actual routing of the data path
towards gateway result in overloaded. To overcome
this problem, the author’s proposed PLB-MSC (Path
Load Balancing Scheme for Maximal Source
Coverage). In this paper, PLB-MSC algorithm is
considered that helps in adjusting the load of the
network by deviating network traffic towards less
congested route and at the same time it dynamically
alter the range of the gateway advertisement
message, so as try to eliminate the unnecessary
control message in the ad hoc network. This
combined system present maximal source coverage
Received: April 12, 2017 238
International Journal of Intelligent Engineering and Systems, Vol.10, No.4, 2017 DOI: 10.22266/ijies2017.0831.25
strategy in which path load balancing and the
adaptive gateway discovery both issues are
discussed simultaneously. Maximal source coverage
strategy truly adaptive path load balancing
mechanism for that there is need to enhance the
range of the Gateway advertisement message
dynamically.
Khan et al. [15] presents a method for selection
of an efficient gateway and routes to other mobile
nodes two factors were consider in this approach
first is length of the routing queue and other is
minimum hop count. This is based on the proactive
gateway discovery approach. It is a novel approach
implemented by calculating the path load and hence
updating the routing table entry whenever there is a
route request from one node to other. The newly
updated routes can be used without any prior
waiting for advertisement. There is reduction in the
delay along the path for packet transmission. As
here there is selection of less congested path it will
help in increasing the throughput. There is always a
relative overhead of managing routing queue and
pre-advertisement route maintenance.
The above-mentioned schemes have used the
different schemes like the candidate gateway, multi-
tier, network division and some others also and
various protocols like the DSDV and AODV were
used. In these the gateway discovery process was
improved by improving the congestion control
tactics but overloading is observed in these
approaches. So, to remove this and develop a more
efficient scheme the proposed scheme has been
developed based on the swarm optimization method.
3. Particle swarm optimization basics
Particle Swarm Optimization algorithm is
developed by Kennedy and Eberhart in 1995 [16].
This new concept was given to solve the
optimization problem and was adapted very vastly
and researchers focus more about it and interest
gradually increase in last few years. This
optimization technique is based on the social
behaviour of bird flock or fish school. The PSO
finds optimal solution of a problem by updating
velocity and position of particle in search space
iteratively until the termination criteria is not
achieved. In Fig.2, we can see that every cluster
having population of various particle searches some
optimal particle with whom it can communicate.
The working procedure of PSO is that a particle
roams in the search space and interacts with other
particles to find the optimal solution, in addition to
this every particle has their own position and
Figure.2 Search model
velocity vector. They also have their global vector to
present global optimal solution. Here ‘i’ represents
the particular particle. Each particle stored some
memory which contains particle best position and
particle global best position.
Particle Swarm Optimization optimizes the
function by finding the minimal optimal solution.
Here Eq. (1) performs the velocity update and
position update. Vid denotes the present velocity of
particle with ID number id, c_(1 )and c_2 denote the
acceleration constant. Normally we take this
constant close to 2.
(1)
(2)
where, id = 1, 2,... n
and are the position best and global best
value of the particle respectively. Further, in Eq. (2),
represents the position of a particle. These both
equations are used in optimization process and
velocity and position is adjusted according to this.
Algorithm for PSO is discussed below.
Algorithm for Particle Swarm Optimization
1: Randomly set the position and velocity for
each particle.
2: Calculate fitness value of each particle.
Received: April 12, 2017 239
International Journal of Intelligent Engineering and Systems, Vol.10, No.4, 2017 DOI: 10.22266/ijies2017.0831.25
3: Calculate Pi, BEST and Gi,BEST for each
particle.
4: Iterate following process until termination
criteria is achieved
a. Update velocity of each particle by
using Eq. (1)
b. Update position of each particle by
using Eq. (2)
c. Calculate fitness value of each
particle
d. Update
Pi, BEST and Gi, BEST of
each particle using following
equations-
Where Pj is the position of particle with maximum
fitness value and f(x) is the fitness function
Once the iteration will over, the algorithm gives the
global best and position best value for every particle
in the search space. Now the algorithm compares the
value with previous stored global best and position
best value in history. If it gets more optimal values
then set this current value as present best optimal
value for each particle.
To implement the PSO we need an objective
function or fitness function which will iterate the
value of particle best position. Here, we are
assuming a graph G and the vertices of the graph are
represented by (v1, v2, v3,… vn) and set of node in
the graph G are represented by (n1, n2, n3, .... nn) and
edges are represented by (e1, e2, e3,.....em) and the load
on the edges or the weight on the edges are
represented by W = (w1,w2,w3,…wm). Swarm
optimization and the right path selection is given
above in algorithm. Thus, to enhance the
performance and increase the services in the
network the shortest path between the source to
destination is calculated by using this objective
function in PSO which is given by this Eq. (3).
(3)
and
In the above equation PPi is the sequential set of
nodes in path of particle ith particle and Ni is the
number of nodes in path of particle i. Cyz represents
the load in path connecting node y and z.
Now we are defining the procedure, how we are
implementing the particle PSO and the right path
selection algorithm. Particle Swarm Optimization is
used in MANET for selecting appropriate path
among all paths available. We discuss the algorithm
for path selection using PSO below. In this,
CurNode is number of current nodes, M is the
number of neighbours of a particular particle, Ni is
the ith neighbour.
4. Proposed work
Here we are developing two different
approaches in which first one is used for path
selection and second one is used in discovering
Internet gateways.
4.1 Algorithm for path selection using PSO
Particle Swarm Optimization is used in MANET
for selecting appropriate path among all paths
available. We have discussed the algorithm for path
selection using PSO below.
Optimum Minimum Load Finding Algorithm: N:
Number of Mobile Nodes, Ni: ith Neighbour of a
node
1: Define N Number of mobile Nodes in the
network with specific parameters in terms of
energy, transmission rate etc.
2: Define the Source and the Destination node
over the network
3: Set CurNode as the current node i.e. Source
Node
CurNode ← ID of current node
4. Find M Neighbor Nodes of Nodes CurNode
and maintains the respective Information.
5. for i=1 to M calculate
a. Distance (Neighbor (i))
b. Load (Neighbor (i))
c. Delay (Neighbor (i))
end for
6. if Load (Neighbor (i)) < Threshold and
Delay < Threshold1
/* Reliable node identify*/
Received: April 12, 2017 240
International Journal of Intelligent Engineering and Systems, Vol.10, No.4, 2017 DOI: 10.22266/ijies2017.0831.25
for i=1 to Mi
CollectInformation(Ni,
Neighbor(Ni))
end for
end if
7. Implement Forward SWARM to find the
alternate path in each Direction of
Neighbour (N (i)).
8. Set the Pheromone on Each Hop and
Identify the possible Path
9. Implement Backward SWARM to inform
neighbour nodes about backup path
10. Trace the pheromone and communicate to
new path.
11. Perform normal communication
The description of the Swarm concept is presented
here
At regular interval, any node s (source) is
selected to send data to some destination node d.
Each forward Swarm selects the next hop node
using the routing table information. The next
node selected depends on some random scheme.
If all nodes already visited a uniform selection
will be performed.
If the selected node is some attack or damage
node or it is not currently available. The forward
Swarm waits to turn in the low priority node
from the queue.
It will identify any of the next non-visited nodes
and pay some delay on it.
If some cycle detected, the Swarm is forced to
turn on the visited node.
When the Swarm reaches the destination node a
backward Swarm is generated to transfer all its
memory.
Backward Swarm uses same path generated by
forward SWARM.
Every time to send the packet, a root is selected
on the basis of path selection formula and the
criteria of selecting the best path is on the basis of
which path has lowest load and minimum energy.
So that we always select the less loaded path. That
criteria will ensure that we always select the path
with minimum load. If we get any problem in
communicating with next node or selecting the next
node for example if all routes are busy in routing or
some links are broken then we apply the Particle
Swarm Algorithm. The purpose of PSO is to select
the same previous path until no other path is
available to route the packet. Thus, we achieve load
balancing in the ad-hoc network by applying path
selection whereas Swarm Optimization algorithm
gives the required reliability. Thus, above algorithm
will find out the optimal group of node by which we
can route data without load creating on path and this
path will be more efficient and reliable.
4.2 Algorithm for gateway discovery using PSO
In our gateway discovery approach, we use the
discovery process based on the optimal route
selection. In this a mobile node discovers the
gateway and updates its route after selecting the
optimal path. For this purpose, we propose the
following information for gateway discovery. For
discovering the gateway, a node relies on gateway
advertisement message (GW_ADV). If not, it
receives an advertisement message; it has to send a
gateway solicitation message (GW_SOL) to Internet
gateway via multi-hop neighbours.
Algorithm for Gateway Discovery using PSO
1. Gateway broadcasts GW_ADV message to
the entire mobile node in its proactive zone.
2. if mobile nodes receive a single GW_ADV
Updates its routing table of mobile node
else
Performs PSO for getting the optimal
route.
end if
3. if mobile node has a default gateway
if no another GW_ADV received
Continue operation
else
Compare the paths based on PSO
and update routing table accordingly.
end if
4. if nodes do not receive a GW_ADV
Mobile nodes send a multi-hop
GW_SOL to gateway.
end if
5. if gateway reply with a single GW_ADV to
source node
Update route according to this message.
else
Compare optimal route using PSO and
update route accordingly.
6. Repeat Step 1
Received: April 12, 2017 241
International Journal of Intelligent Engineering and Systems, Vol.10, No.4, 2017 DOI: 10.22266/ijies2017.0831.25
Figure.3 Comparison Graph between Local Best and
Global Best Value after Executing Proposed Algorithm
5. Simulation and result analysis
We used MATLAB tool to implement Swarm
optimization algorithm. First, we execute the
Particle Swarm Optimization using proposed fitness
function and shown how the particle gets the best
value for local best and global best position. Then
we have shown two scenarios, where we have first
implemented AODV routing protocol to route the
packet by shortest path. Then we have compared it
to our proposed path selection protocol.
We have shown the graph between our proposed
objective and number of iteration in Fig. 3. It shows
the particle best position and global best position
attained in different number of iterations. The PSO
has optimal result, which consist of set nodes
between source and destination node to route data
packet, when it terminates.
We have considered two scenarios to illustrate
the working of our proposed work, which are
discussed below. In first scenario, we took 30
number of nodes to find shortest path between
source and destination node. Parameters used in first
scenario are given in Table 1.
Fig. 4 shows the distribution of nodes in ad-hoc
network. As we can discern the nodes are numbered
from 1 to 30. Blue dots are displaying source node
and the destination node is red. All supplementary
nodes are the intermediate nodes. On this web, we
have early requested the shortest trail algorithm
utilized by our proposed work.
Table 1. Simulation parameters in scenario 1
Parameter
Value
Number of Nodes
30
Topography Dimension
11mx11 m
Traffic Type
CBR
Topology
Random
Initial Node
1
Destination Node
30
Distance
2.2678 (meters)
Energy Consumed
1.8020e+003 (joules)
Network Delay
596.9317ms
Elapsed time
0.000846 s
Path followed
[1 24 30]
Figure.4 Random number of nodes generated in an Ad-
Hoc area
Figure.5 Generated path with 30 nodes using proposed
scheme
-80
-60
-40
-20
0
0 5 10 15 20 25 30
Objective
Iteration
Mean (F_best)
Mean (F)
Global best
012345678910 11
0
1
2
3
4
5
6
7
8
9
10
11
1
2
3
4
5
6
7
8
9
10
11 12
13
14
15
16 17
18
19 20
21
22 23
24
25
26
27
28
29
30
0 1 2 3 4 5 6 7 8 9 10 11
0
1
2
3
4
5
6
7
8
9
10
11
1
2
34
5
67
8
9
10
11
12
13 14
15
16
17
18
19
20
21
22 23
24
25
26
27
28 29
30
Received: April 12, 2017 242
International Journal of Intelligent Engineering and Systems, Vol.10, No.4, 2017 DOI: 10.22266/ijies2017.0831.25
We applied our proposed algorithm to find out
shortest and load balanced path between source
node 1 and destination node 30. Fig. 5 shows that
generated path with 30 Nodes with existing scheme.
Fig. 5 shows that generated path with 30 nodes
using our proposed work, which shows the shortest
path is 1=>24=>30. We took 40 numbers of nodes
for second scenario The Parameters used in Second
scenario are given in Table 2.
Fig. 6 shows the distribution of nodes in ad-hoc
network. As we can discern the nodes are numbered
from 1 to 40. Blue dots are displaying source node
and the destination node is red. All supplementary
nodes are the intermediate nodes.
Table 2. Simulation parameters in scenario 2
Parameter
Value
Number of Nodes
40
Topography Dimension
11mx11 m
Traffic Type
CBR
Topology
Random
Initial Node
1
Destination Node
30
Distance
3.2354 (meters)
Energy Consumed
1.4478e+003(joules)
Network Delay
1.6904e+003 ms
Elapsed time
0.000892 s
Path followed
[1-37-38 -11- 30]
Figure.6 Random number of node generated in Ad-Hoc
area
Figure.7 Generated path after executing proposed
approach
Figure.8 Normalized routing overhead v/s speed of nodes
We can see in Fig.7 how the path is generated
when we took 40 numbers of nodes and run the
algorithm. The shortest path generated is
1=>37=>38 =>11=> 30. We have compared our
proposed techniques with other approaches like
PLB-MSC [14], PLB [15] and AODV [16] protocols.
The result is shown in Fig. 8, which shown the
simulation result of our proposed work and PLB,
PLB-MSC and AODV protocols. Here it is clearly
shown that our proposed graph will give better result
by selecting more optimal path to route the data
packet. Fig.8 shows the comparison of normalized
routing overhead with existing approaches.
Graph shown in Fig. 9 is drawn which compares
the energy graph of our proposed work and PLB-
MSC [14], PLB [15] and AODV [16] protocols.
Initially, the energy of each node in all the cases was
same but as time passes, energy gets depleted.
0 1 2 3 4 5 6 7 8 9 10 11
0
1
2
3
4
5
6
7
8
9
10
11
12
34
5
67
89
10
11
12
13
14 15
16
17
18
19
20
21
22
23
24
25 26
27
28 29
30
31
32
33
34
35
36
37
38
39
40
0 1 2 3 4 5 6 7 8 9 10 11
0
1
2
3
4
5
6
7
8
9
10
11
12
34
5
67
89
10
11
12
13
14 15
16
17
18
19
20
21
22
23
24
25 26
27
28 29
30
31
32
33
34
35
36
37
38
39
40
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
1 2 3 4 5 6 7
Normalized Routing Load
Speed of Node (Mts/Sec)
AODV
PLB
PLB-MSC
Proposed work
Received: April 12, 2017 243
International Journal of Intelligent Engineering and Systems, Vol.10, No.4, 2017 DOI: 10.22266/ijies2017.0831.25
Figure.9 Energy graph
Figure.10 Lifetime
But one can easily interpret that the consumption
is fast in case of AODV, PLB and PLB-MSC based
approaches in comparison with our proposed one.
We have considered a simulation scenario which is
run till 300 sec and initial energy in all the cases are
20 joules.
Another comparison is made and represented
using graph shown in Fig. 10 for comparing the
lifetime of nodes with our proposed work and other
existing works like PLB-MSC[14], PLB[15] and
AODV[16]. Comparison is done in a varying
environment in which the number of low energy
nodes is varying in different situations. Outcome
reflects that our proposed approach achieves more
lifetime than existing approaches.
6. Conclusion and future scope
In this paper, we gave our main concern to
decrease the overload and remove or solve the
problem of congestion which occurs due to path
load in ad-hoc networks. So, we surveyed in depth
about the many existing packet routing algorithms
which select the shortest paths to transfer the data
reliably. As ad-hoc networks are dynamic than
wired networks where nodes are mobile, thus our
challenge was not only to balance the load on the
network but also to provide reliable path to the
entire mobile nodes. In our proposed work, we have
combined PSO with routing algorithm to find the
optimized route between source and destination
node if the previous selected path gets overloaded or
the link is broken, by default it will always select the
minimum distance path. Our proposed techniques
have been compared with other approaches like
PLB-MSC, PLB and AODV protocols. Our
approach shows enhancement over existing ones by
selecting more optimal path to route the data packet.
It is also observed through simulation that our
proposed approach achieves more lifetime than
existing approaches due to appropriate load
balancing technique used as Particle Swarm
Optimization is used for selecting an optimal path
among all paths available.
In PSO the whole working depends on how we
are taking the objective function or fitness function.
If one wants more optimized value, then one has to
select the more precise function to evaluate the
fitness value. To select the shortest path, there are
many algorithms which can be applied for mobile
ad-hoc network, some of them are already
implemented and some are still not considered.
References
[1] G. Aggelou, “Mobile Ad Hoc Networks: From
Wireless LANs to 4G Networks”, McGraw-Hill
Professional, 2004.
[2] R. Eberhart and J. Kennedy, “A new optimizer
using particle swarm theory”. In Micro Machine
and Human Science MHS'95, In: Proc. of the
Sixth International Symposium, pp. 39-43, 1995.
[3] J.H. Zhao, X.Z. Yang, and H.W. Liu, “Load-
balancing Strategy of Multi-gateway for Ad hoc
Internet Connectivity”, In: Proc. of the
International Conference on Information
Technology: Coding and Computing (ITCC'05) –
Vol. 2, pp.592-596, 2005.
[4] A. Trivino-Cabrera, E. Casilari, D. Bartolome,
and A. Ariza, “Traffic Load Distribution in Ad
Hoc Networks through Mobile Internet
Gateways”, In: Proc. of Fourth International
Working Conference on Performance Modelling
0
5
10
15
20
25
30 90 120 150 180 210 240 270 300
Node Energy in Joules
Time (Sec)
Proposed Work
PLB-MSC
PLB
AODV
0
10
20
30
40
50
60
10 20 30 40 50 60 70 80 90
Percentage of Energy Used
No. of Low Energy Nodes
AODV
PLB
PLB-MSC
Proposed Work
Received: April 12, 2017 244
International Journal of Intelligent Engineering and Systems, Vol.10, No.4, 2017 DOI: 10.22266/ijies2017.0831.25
and Evaluation of Heterogeneous Networks,
2006.
[5] Y.Y. Hsu, Y.C. Tseng, C.C. Tseng, C.F. Huang,
J.H. Fan, and H.L. Wu, “Design and
Implementation of Two-Tier Mobile Ad Hoc
Networks with Seamless Roaming and Load-
Balancing Routing Capability”, In: Proc. of the
First International Conference on Quality of
Service in Heterogeneous Wired/Wireless
Networks, pp.52-58, 2004.
[6] J. Shin, H. Lee, J. Na, A. Park, and S. Kim,
“Load Balancing among Internet Gateways in Ad
Hoc Networks”, In: Proc. of 62nd IEEE
Vehicular Technology Conference, pp.1677-
1680, 2005.
[7] Q. Le-Trung, P.E. Engelstad, T. Skeie, and A.
Taherkordi, “LoadBalance of Intra/Inter-
MANET Traffic over Multiple Internet
Gateways”, In: Proc. of the 6th International
Conference on Advances in Mobile Computing
and Multimedia, pp. 50-57, 2008.
[8] M. Ali, B. G. Stewart, A. Shahrabi, and A.
Vallavaraj. "Congestion adaptive multipath
routing for load balancing in Mobile Ad hoc
Networks”, In: Proc. of Innovations in
Information Technology (IIT), 2012 International
Conference on, pp. 305-309. IEEE, 2012.
[9] V. Pham, E. Larsen, P. E. Engelstad, and Ø.
Kure. "Performance analysis of gateway load
balancing in ad hoc networks with random
topologies." In: Proceedings of the 7th ACM
international symposium on Mobility
management and wireless access, pp. 66-74.
ACM, 2009.
[10] A. Rani and M.Dave, “Performance Evaluation
of Modified AODV for Load Balancing”,
Journal of Computer Science, Vol 3, issue 11,
pp 863-868, 2007.
[11] Y.H. Yuan, H.M. Chen, and M. Jia, “Adaptive
load balancing scheme in ad hoc
networks”, Journal of Shanghai University
(English Edition), Vol. 11, No.3, pp.296-299,
2007.
[12] A.K Gupta, A.K Verma, and H.Sadawarti,
“Analysis of various Swarm-Based & Ant-
based Algorithm”, In: Proc. of ACAI’11, pp.
39-43, 2011.
[13] X. Hu, Y.Shi, and R. Eberhart, “Recent
Advances in Particle Swarm”, Evolutionary
Computation, Vol.1, pp. 90-97, 2004.
[14] R.U. Zaman, K.U.R. Khan and A.V. Reddy,
“Path load balanced adaptive gateway discovery
in integrated internet-MANET”, In: Proc. of
Communication Systems and Network
Technologies (CSNT), 2014 Fourth
International Conference on, pp. 203-206, 2014.
[15] K.U.R. Khan, M. Ehthesham, M. Kumar, and
R.U. Zaman, "An effective gateway discovery
mechanism in an Integrated Internet-MANET
(IIM)", In: Proc. of Ultra Modern
Telecommunications & Workshops, 2009.
ICUMT'09. International Conference on, pp. 1-
7, 2009.
[16] C. Perkins, E. Belding-Royer and S. Das, "Ad
hoc On-Demand Distance Vector (AODV)
Routing", RFC 3561, DOI 10.17487/RFC3561,
July 2003,
<http://www.rfc-editor.org/info/rfc3561>.