You are on page 1of 13

INTERNATIONAL JOURNAL OF COMMUNICATION SYSTEMS

Int. J. Commun. Syst. (2015)


Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/dac.2989

An energy-efcient history-based routing scheme for opportunistic


networks

Sanjay K. Dhurandher1, Deepak Kumar Sharma1, Isaac Woungang2,*,


and Aakanksha Saini1
1
CAITFS, Division of Information Technology, NSIT, University of Delhi, New Delhi, India
2
Department of Computer Science, Ryerson University, Toronto, Ontario, Canada

SUMMARY
In opportunistic networks (Oppnets), nodes rely on contact opportunities between them to exchange infor-
mation with each other. Routing and forwarding in Oppnets remains a challenging task because of the lim-
ited energy and bandwidth constraints. Various routing protocols for Oppnets have been proposed in the
literature, but only few of them have explicitly investigated the energy issue. In this paper, some improve-
ments in the already existing history-based prediction for routing protocol for infrastructure-less Oppnets
(so-called HBPR) is suggested so as to make it energy efcient. The proposed energy-efcient HBPR pro-
tocol (EHBPR) addresses the energy constraints in HBPR and reduces the number of packets transferred in
the network, which in turn results to a reduction in the nodes energy consumption. Through simulations,
the performance of EHBPR in terms of energy consumption is compared against the HBPR and the
energy-efcient n-epidemic routing protocol. The results show that (1) EHBPR consumes 14.66% less
energy than HBPR (respectively 13.14% less energy than n-epidemic); (2) EHBPR generates 67.4% less
dead nodes compared with HBPR (resp. 66.33% less dead nodes compared to n-epidemic); and (3) EHBPR
yields 77.86% less overhead ratio compared with HBPR (resp. 84.49% less overhead ratio compared with
n-epidemic). Copyright 2015 John Wiley & Sons, Ltd.

Received 31 December 2014; Revised 17 April 2015; Accepted 17 April 2015

KEY WORDS: Opportunistic networks; Opportunistic routing; ONE (Opportunistic Network Environment)
simulator; HBPR Protocol

1. INTRODUCTION

Opportunistic networks (Oppnets) [1] are considered as the most recent expansion of mobile ad hoc
networks (MANETs) [2]. In an Oppnet, an end-to-end path from source node to destination node is
not necessarily guaranteed most of the time. Due to their self-congurable property, Oppnets have
become an interesting and important topic for research among the research communities working in
the areas of wireless communication. The nodes in Oppnets are highly mobile and do not have a
xed infrastructure, which results in an extremely dynamic network topology. Frequent disconnec-
tions, sparse connectivity, intermittent links, and limited resources are considered as the basic char-
acteristics of these networks [3]. So, the communication opportunities are very occasional and
discontinuous in such networks. Further, node mobility patterns are uncertain and hard to predict
before time that makes the message delivery in Oppnets a very compelling task. In extreme cases,
an Oppnet may start functioning with a single node [1] called the seed Oppnet, and then by
employing several foreign helper nodes, it can grow into an expanded Oppnet. These helper nodes

*Correspondence to: Isaac Woungang, Department of Computer Science, Ryerson University, Toronto, Ontario, Canada.

E-mail: iwoungan@scs.ryerson.ca

Copyright 2015 John Wiley & Sons, Ltd.


S. K. DHURANDHER ET AL.

contribute in the routing and forwarding of the messages. Such type of Oppnet is employed for
emergency preparedness and response activities in disaster scenarios [4].
Opportunistic networks have many issues and challenges that are quite different from MANETs.
In traditional MANET routing protocols, an end-to-end path between the source and destination
nodes is rst established before the start of message passing. In Oppnet, the end-to-end path require-
ment may not necessarily be satised because of short range wireless networks, nodes power fail-
ure, and nodes mobility, which may lead to message dropping. Further in MANETs, for multiple
message transfer between a pair of nodes, all nodes through which message has propagated in the
past participate like intermediate nodes for the next message. While in Oppnets, due to high mobil-
ity, some or all the nodes may not be a part of the nal path between a pair of nodes for the next
message even if they had forwarded the message at some point of time in the past. Thus, MANET
routing protocols will fail to work in this kind of environment. Due to long propagation and variable
queuing delays, Internet protocols, which are designed to assume quick return of acknowledge-
ments and information, can fail to work in such networks. Oppnets resolve this issue by exploiting
the node mobility and local forwarding in order to transfer data. Data can be stored and carried by
taking advantage of node mobility and then forwarded during opportunistic contacts between the
nodes [4, 5].
Routing and forwarding in Oppnets is based on contacts between the nodes, their desire to par-
ticipate in forwarding, and store-carry-forward paradigm [5]. The contacts are hardly predictable
because of the node mobility or the dynamics of wireless channel; they must be exploited opportu-
nistically for exchanging messages between some nodes that can move in the remote fragments of
the network. During forwarding process, it is possible that the source or any intermediate node may
not nd a good candidate next hop node that can carry the message as closer as possible to the des-
tination. In such case, the message has to be kept in the buffer of the nodes. Thus, to avoid the
dropping of packets, nodes in the network are required to have enough buffer space to store all mes-
sages for an unpredictable period of time until next contact occurs and a suitable node is found. This
required storage space increases a function of the number of messages in the network. The afore-
mentioned store-carry-forward technique increases the probability of successful message delivery
but at the same time introduces longer delivery delays. This is why Oppnets are also known as
sub-class of delay tolerant networks [6, 7].
One of the most critical issues in Oppnets is the limited availability of energy within network
nodes. Because of the fact that nodes have limited batteries, the energy management of the batteries
is a key challenge in Oppnets. A lot of battery power gets consumed in forwarding and next hop
selection, which reduces the network lifetime. Further, if a node sends each packet many times, it
will drain its battery quickly and cannot relay other packets. Thus, to maximize the Oppnet lifetime,
packets should be sent via a route that can avoid nodes with low power so that their power can be
saved and they can remain available in the network for longer period of time. Hence, there is a need
to design a proper energy-efcient routing protocol for message passing in these types of networks
in order to increase the overall network capacity. Although HBPR [8, 9] works better in human mo-
bility scenario, it consumes a considerable amount of energy. This motivated the authors to devise a
new technique to make the HBPR energy efcient so as to maximize the lifetime of nodes in the
network.
The rest of the paper is organized as follows. Section 2 presents the background and related work
in the area of energy awareness in routing protocols on Oppnets. Section 3 describes the proposed
EHBPR protocol in detail. Section 4 is devoted to the simulation setup. Section 5 presents the sim-
ulation results, where the performance of the EHBPR is discussed and compared with the HBPR
and energy-efcient n-epidemic protocol [10]. Finally, Section 6 concludes this paper.

2. RELATED WORK

Most of the works for the design of energy aware routing protocols have been carried out in ad
hoc networks [11, 12] and sensor networks [13, 14]. Only few energy-efcient routing protocols
[10, 1518] exist in the literature for Oppnets. In this section, a brief overview of the HBPR proto-
col along with some existing energy aware routing protocols for Oppnets has been discussed.

Copyright 2015 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2015)
DOI: 10.1002/dac
ENERGY EFFICIENCY

2.1. HBPR protocol


This protocol [8, 9] uses the behavioral information of the nodes to make predictions about their
movements in the network. It uses the past history about the movement of nodes in the network
and predicts the places they visit very frequently or very rarely. This predicted information is used
to nd and select the best next hop node that can contribute to route the message to the destination.
In HBPR, it is assumed that nodes move according to the custom human mobility pattern [8]. The
working of HBPR protocol is divided into three phases: the initialization of the home locations, the
message generation and home location updating, and the next hop selection. These three phases are
described as follows:
Initialization of home locations: nodes that follow the human mobility pattern visit some loca-
tions more frequently than the other. These frequently visited locations are named as home
locations of the nodes. HBPR uses the Home Location table to store the information about such
locations of nodes. The nodes ood their home location information in the network so that every
node knows the home location of other nodes.
Message generation and home location updating: New messages are generated at few nodes of
the network. If a node changes its behavior and visits some other location more frequently, then
it oods the details of its current home location in the network. A history table is used to main-
tain the list of various locations visited by a node in the past. After a xed time interval, the
history of node gets updated in the table.
Next hop selection: For sending the message further, the next hop is decided using a utility metric.
This utility metric is calculated using three parameters, namely, the stability of a nodes move-
ment, prediction of direction of movement using Markov predictor, and the perpendicular dis-
tance of the neighboring nodes from the line of sight of source and destination (SD line). The
stability is computed by checking the change in the speed of a node. As the nodes move in the
network, they store the coordinate of the locations visited by them with time to calculate their
speed. It is assumed that if the change in speed is very little, then the node is stable, else it is un-
stable. Markov models [19] are used to predict the next location based on the past histories with
the help of a table. This table has two elds that store the location and the frequencies of visit to
this location so as to predict the next location. The perpendicular distance of the nodes with SD
line is checked to select those nodes, which are closer to it. If it is small, it implies that a node
has to travel lesser distance to reach destination node. Larger perpendicular distance implies larger
distance to travel. Finally, the utility metric value for a node is calculated and the message is
forwarded to those nodes that have their utility metric value greater than the threshold (T).

2.2. Energy-efcient n-epidemic routing protocol


This protocol [10] is an energy-based improvement of the traditional epidemic routing protocol [20]
used in Oppnets. It limits the amount of ooding of packets in the network, which in turn reduces
the battery consumption of the nodes, as they have to forward lesser packets to other nodes. In this
protocol, a node forwards a packet only when it has at least n neighbors. It does not forward the
packet if it has lesser than n neighbors and, thus, optimizes the chances of packet forwarding in
the network. By n-epidemic routing, a node cannot forward a packet as casually as by basic epi-
demic routing. So the value of n is determined very carefully. If n is large, the probability of a node
having so many neighbors is low, and the chances that this packet is sent out are less. If a packet
cannot be spread out widely, the destination has low probability of receiving it. On the contrary,
if n is very small, nodes will have lot of chances to forward a packet, and ultimately, their batteries
will get drained very quickly. Therefore, n should be set accordingly based on how sparse the net-
work is and how much battery power of the nodes is available. Even though this protocol reduces
the ooding of packets, the nodes residual energy is not considered in its design.

2.3. Energy-aware spray and wait routing


In this protocol [15], binary spray and wait [21] protocol has been taken as the baseline scheme, in
which a node which has more than one message copy, forwards half of its copies to another

Copyright 2015 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2015)
DOI: 10.1002/dac
S. K. DHURANDHER ET AL.

neighboring node. To take care of the uncontrolled spraying and low efciency in data accumula-
tion, the residual energy and speed information of all nodes are considered during the spraying of
message copies. A node with high speed may visit more number of nodes than the one with low
speed during the same time interval. Therefore, after analysis of speed factor, it can be deduced that
the spray phase can be sped up if more message copies are carried by high speed nodes. Also, the
nodes with less residual energy will not be the best hops to forward messages as this may lead to
higher hazard of data accumulation. Therefore, a utility function comprising of speed and energy
information is used to forward messages in spray phase. The spray phase continues until there is
only one message copy left on a forwarding node. After that, the node commences the wait phase
to wait for direct communication opportunity with the destination node.

2.4. Vibrant energy aware spray and wait routing


This protocol as proposed in [16] is also the energy-based extension of binary spray and wait pro-
tocol. Such transmission may result in blind spraying as there is no knowledge about the network
topology and the behavior of nodes. In vibrant energy aware spray and wait protocol two factors,
the vibrancy and the remaining energy information of nodes are checked to improve the routing
efciency. The vibrancy represents the movement or activity of the nodes in a given period of time.
The node that meets the greater number of nodes will have greater vibrancy. Whenever an oppor-
tunity arises for two nodes to meet, they exchange this vibrancy information with each other and
updates accordingly. The sender or intermediate node then checks the remaining energy of the
neighboring node. If everything is ne, that is, its vibrancy and energy values are good enough, then
it sends the message copy until there is only one copy left in its buffer. After that, the sender node
starts the wait phase, in which it waits for direct contact with the destination node.

2.5. Energy-aware BUBBLE Rap


This protocol [17] is the combination of socially aware routing and optimization in terms of energy
consumption. In BUBBLE Rap [22], the forwarding of messages is performed with the help of a
utility function. The decision is taken by checking the global and local ranks of the two nodes that
meet each other. A node forwards its message to the neighboring node that is most popular among
other nodes, that is, which meets the maximum number of nodes in its community, by using the
local rank. Further, the messages are sent between different communities using the global rank.
When the community of destination node is reached, the message is locally forwarded until it
reaches the destination. The energy-aware BUBBLE Rap extends the utility function by allowing
it to decrease and increase according to the energy resources a node that has to support the message
transfer. Therefore, the chances for a node to be a successful carrier decrease with the decrease in
energy.

2.6. GAER: genetic algorithm-based energy-efcient routing


This protocol as proposed in [18] uses genetic algorithm to send message copies from source to des-
tination. Initially, a set of chromosomes is randomly generated either at the source or intermediate
node that wishes to forward the message to the neighboring nodes. Then the tness of every chro-
mosome of a directly connected node with the sender or intermediate node is calculated. A tness
function is determined, which is the metric to measure the capability of a node to be good enough to
get selected as the next hop. The tness function includes two factors, namely, the mean factor and
the place factor. The mean factor is the square of the distance between the mean home location of
destination node and the mean home location value of the neighboring node. On the generated chro-
mosomes, some genetic operations such as selection and crossover are applied, yielding new set of
chromosomes. The obtained chromosomes are sorted based on their respective tness values, and
then only the chromosomes with good enough value of the tness function are saved and the rest
are removed. This process of evolution gets repeated until the best possible results are obtained.
The quality of chromosomes obtained after each stage in terms of tness value also increases.
Afterwards, the node having the ttest chromosome is selected as the next hop.

Copyright 2015 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2015)
DOI: 10.1002/dac
ENERGY EFFICIENCY

3. PROPOSED PROTOCOL: ENERGY-EFFICIENT HISTORY-BASED PREDICTION FOR


ROUTING PROTOCOL

In this section, the EHBPR protocol is introduced. The same assumptions that were considered for
the HBPR protocol prevail in the case of the EHBPR protocol. For the design of EHBPR, the follow-
ing improvement factors, namely, perpendicular factor, transmission factor, sparse_constant factor,
and residual energy factor, have been added to the HBPR protocol. They are described as follows.

3.1. Perpendicular factor


This factor is used to choose those neighbors as next hop that are closer to the sender or the interme-
diate node. When a node forwards a message to a neighbor node closer to the sender, lesser energy will
be consumed. Thus, in the utility metric of HBPR, instead of checking the perpendicular distance be-
tween the neighboring nodes and the SD line, the actual distance between the sender and the neighbor
is checked. For example, suppose at a particular time, node S has two neighboring nodes N1 and N2 as
shown in Figure 1. As per HBPR, node N2 will be chosen as it is nearer to the SD line, but more energy
of sender will be consumed while sending messages to N2 rather than N1. Therefore, here in EHBPR,
N1 gets selected to decrease the consumption of energy while transferring the messages.

3.2. Transmission factor


This factor limits the number of message copies with a unique ID that a node can forward, which has
been generated by a particular source for a particular destination. It avoids the overloading of the net-
work resources with duplicate copies of the same message. A node will not forward a message if it
has already forwarded enough copies of this message. A transmission table is maintained by each node.
This table stores the number of messages forwarded by it that belongs to a particular source and desti-
nation pair. This table has four elds, namely, Source_ID, Destination_ID, Message_ID, and Count.
The Source_ID and Destination_ID elds contain the IDs of source and destination nodes, and the
Message_ID eld contains the ID of the message. The Count eld denotes the number of copies
forwarded by a node with this Message_ID between the source and destination. Every time a node
has a message to forward, this table will be updated. When a node receives a message to be forwarded,
it checks the transmission table. If there is no entry for the Source_ID, Destination_ID, and
Message_ID of the message in the transmission table, then a new row with Count = 1 is added to it.
If the entry is already present and the Count of message copies forwarded by it is less than or equal
to the transmission factor, then the message is forwarded; otherwise, it is dropped. Therefore, total num-
ber of messages owing in the network will be much less. This will lead to lesser overhead ratio and
lesser energy consumption in transmission of messages. Thus, overall residual energy of the network
will increase. Table I shows a snapshot of the transmission table for a node at a particular instant of time.

3.3. Sparse_constant factor


This factor is used to avoid the calculation of the Utility Metric done for next hop selection if the number
of nodes that are currently in the senders (or intermediate nodes) neighborhood is lesser than a certain
value called as Sparse_constant factor. Depending on how sparse the network is, the value of the
Sparse_constant factor is set. The Utility Metric value is calculated for the neighboring nodes only if

Figure 1. Example scenario for perpendicular factor.

Copyright 2015 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2015)
DOI: 10.1002/dac
S. K. DHURANDHER ET AL.

Table I. Transmission table.


Source_ID Destination_ID Message_ID Count
N1 N2 M1 8
N1 N2 M2 5
N3 N1 M1 7
N3 N4 M5 1
N5 N6 M3 10

the number of nodes that are currently in the neighborhood is greater than or equal to the Sparse_constant
factor. This will reduce the energy consumption as it will avoid big calculations and reduce the unnec-
essary transmission of messages. The value of the Sparse_constant factor should not be so large that the
nodes will keep on waiting for large interval of time. Also, it should not be so small that its impact be-
comes negligible. This value depends on the nodes mobility as well as the amount of sparseness in the
network. For example, let the number of neighboring nodes be n. If n is less than the Sparse_constant
factor i.e. if lesser number of nodes are there than the assumed constant, then no further calculations will
be done. The node will wait for a time t and it will again check the number of its neighboring nodes. After
time t, if the value of n is greater than or equal to that of the Sparse_constant factor, then further calcu-
lation of Utility Metric for next hop selection will be done. This checking to nd the number of neigh-
boring nodes is done for a maximum time period of X units. After waiting for time X, if the number of
neighboring nodes remains lesser than the Sparse_constant factor, then this node will give up. It will start
further calculations with whatever number of nodes it has in its neighborhood. This value of X is used to
avoid innite waiting or waiting for the whole simulation.

3.4. Residual_energy factor


This factor is used to avoid the low-power nodes from getting energyless in the network. As energy
keeps getting consumed in scanning of the nodes and forwarding of the messages, it may happen
that after some time, a nodes all energy gets consumed and it is unable to further participate in
any communication.
The residual_energy factor resolves this problem. First, the sender (or intermediate) node checks
the residual energy of the neighboring nodes before forwarding any message to them. Second, a
message is forwarded to its neighboring node if its residual energy is greater than or equal to the
residual_energy factor. The message transfer gets aborted if the residual energy of the neighboring
node at that particular time is lesser than the residual_energy factor. Thus, residual_energy factor
will avoid those nodes to be selected as next hop whose energy is very less. Hence, it reduces their
chances of getting drained. Figure 2 shows an example scenario for residual_energy factor. Assume
that node S would like to send a message. R1 and R2 are the neighboring nodes of node S. It can be
seen that R2 has better utility metric and is closer to node S compared with node R1, but the residual
energy of node R2 is less than the residual_energy factor. Therefore, although node R2 seems

Figure 2. Example scenario for residual_energy factor.

Copyright 2015 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2015)
DOI: 10.1002/dac
ENERGY EFFICIENCY

perfect, it will become energyless after some time. Therefore, selecting the node R2 is not a suitable
choice for next hop. Therefore, EHBPR will select node R1 as next hop.
With the inclusion of the aforementioned factors, the EHBPR can consume lesser energy of nodes
as compared with HBPR protocol. However, the delivery probability of EHBPR may decrease, and
the average latency and average buffer time may increase. This is due to the energy constraints of
EHBPR that are being imposed to save the energy of the nodes. These constraints limit the chances
of a node being selected as a next hop; hence, they affect the delivery probability, average latency,
and average buffer time. It has been observed (through simulations) that EHBPR works better in
terms these parameters when lesser number of nodes is considered that is, in sparse networks
because when a large number of nodes is considered, the energy constraints become a hindrance.

Algorithm of EHBPR

/*EHBPR executes the steps that follows to deliver the message to the destination. We assume
that all tables that are used in EHBPR are available within the nodes.*/

Step 1: The source node (SN) creates a new message or selects the next message in its message
queue till messages exist. Let this message be M.
Step 2: Repeat for each neighboring node (NN) of SN or any intermediate node until message
reaches the destination.
(a) If NN is the message destination, deliver the message to it and exit.
(b) If there is no entry for this message M in the transmission table of SN or any
intermediate node, then add it to the transmission table.
Else
Go to Step 2(c).
(c) If (value of count for message M at an intermediate node<=transmission factor)
{
This node can forward the message M to a suitable neighbor. Go to Step 2(d).
}
Else
{
This node drops the message M.
}
(d) For (i=1; i<=X; i++) // X denotes the time units.
{
If (n >= Sparse_constant factor) // n denotes the total number of neighboring nodes.
{
Calculate the utility metric of EHBPR for the next hop selection and go
to Step 5(e).
}
Else
{
Wait for time t.
}
} //End of loop.

Calculate the utility metric of EHBPR with whatever value of n.

(e) Select those nodes from n neighboring nodes as a candidate for next hop whose
value of utility metric > Threshold (T). Let this set of nodes be K.
(f) Out of this set K, select those nodes as the next hop whose residual
energy > Residual_energy factor. Let this set of nodes be L.
(g) Forward the message copy to these L nodes.

Copyright 2015 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2015)
DOI: 10.1002/dac
S. K. DHURANDHER ET AL.

4. SIMULATION SETUP

In this work, the simulation and energy-based performance evaluation of EHBPR with HBPR and
n-epidemic routing protocols have been performed using the ONE simulator [23]. The
MessageStatsReport and EnergyLevelReport available in ONE have been used to take the delivery
and energy-related results for all the protocols in this work. The detailed description of various en-
ergy related parameters for a node is given in [24]. The value of various parameters taken during the
simulation is shown in Table II.
In the simulation, the factors such as number of nodes, message size, and message generation in-
terval are varied to see their effect on the results obtained for EHBPR. These variations are
discussed as follows:
(a) Varying the number of nodes. During the simulation, the total number of nodes is changed from
40> 80> 120> 160, that is, an increment of 40 nodes is performed each time.
(b) Varying the message size. The size of message is altered from 0.00.5 MB to 1.52.0 MB, with
an increment of 0.5 MB each time.
(c) Varying the message generation interval. The message generation interval is varied from 0-10
seconds to 30-40 seconds, with an increment of 10 seconds each time.
While varying the aforementioned elds, the other parameters are kept xed to their default
values as specied in Table II. The following metrics are used for evaluating the performance of
EHBPR and comparing it with HBPR and n-epidemic protocols.
(a) Average residual energy: This metric denotes the average of all nodes remaining energy when
the simulation ends.
(b) Number of dead nodes: This metric denotes the count of nodes whose residual energy becomes
lesser than a cut-off value after the completion of the simulation. In this work, this cut-off value
is taken to be 900 units.

Table II. Simulation parameters values.


Parameter Value
Simulation area 4500 m * 3400 m
Number of node groups 4
Total number of nodes 80
Communication Interface Bluetooth
Transmission range 10m
Transmission speed 2 MB/s
Group1 nodes type, followed by velocity Pedestrians, 0.51.5 m/s
Group2 nodes type, followed by velocity Cyclists, 1.55 m/s
Group3 nodes type, followed by velocity Cars, 510 m/s
Group4 nodes type, followed by velocity Trams, 713 m/s
Buffer capacity of Group1 and Group2 nodes 5 Mb
Buffer capacity of Group3 and Group4 nodes 50 Mb
Initial energy of all nodes 18, 000 units
Energy consumed per node scanning 4 units
Energy consumed per node scanning response 4 units
Energy consumed in message transmission 6 units
Energy consumed when node is idle 0.4 units
Message size 500 Kb up to 1 Mb
Message generation interval 2535 s
Simulation time 43200 s
Message time-to-live 300 min
Nodes movement model Custom Human Mobility model [8]
t 1 min
X 10 min
Sparse_constant factor 3
Transmission factor 10
Residual_energy factor 900 units

Copyright 2015 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2015)
DOI: 10.1002/dac
ENERGY EFFICIENCY

(c) Overhead ratio: This metric denotes the average number of copies that are relayed per message,
that is, the amount of trafc that nodes had to manage so as to send the messages to their
destination.

5. RESULTS

In this section, the various results obtained through simulations in the ONE simulator are shown
with the help of the graphs.

5.1. Varying the number of nodes


The results showing the effect of the number of nodes on the performance of EHBPR and its com-
parison with other protocols are depicted in Figure 3(a)(c). It can be observed from Figure 3(a) that
with the increase in the number of nodes, the average residual energy decreases because of more
interactions between the nodes. These increased interactions results in greater number of device
scans and their responses, which eventually decrease the average residual energy of the nodes.
EHBPR has the highest value of average residual energy among three protocols because of the var-
ious factors incorporated in it for energy optimization. Figure 3(b) shows that EHBPR produces the
lowest number of dead nodes after the simulation as compared with the HBPR and n-epidemic pro-
tocols. This is due to the residual_energy factor used in EHBPR that avoids those nodes to be
selected as next hop whose energy is very less. Hence, it reduces the chances of low power nodes
from getting drained in the network. In Figure 3(c), the overhead ratio increases with an increase in
the number of nodes due to the increase in the number of messages owing in the network.
Compared with other protocols, EHBPR yields the least overhead ratio. This is justied by the fact
that EHBPR generates a lesser number of message copies because of its transmission factor that
restricts the number of message copies forwarded by a node. These results show that the application

Figure 3. Performance metrics versus number of nodes. HBPR, history-based prediction for routing;
EHBPR, energy-efcient history-based prediction for routing.

Copyright 2015 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2015)
DOI: 10.1002/dac
S. K. DHURANDHER ET AL.

of EHBPR protocol in Oppnets can decrease the power consumption of the nodes, hence increase
the network lifetime.
The average residual energy of EHBPR is 15.20% more than HBPR and 13.31% more than
n-epidemic protocol. The number of dead nodes in EHBPR is 75% lesser than HBPR and 72.5%
lesser than n-epidemic protocol. EHBPR has 78.745% and 84.5% lesser overhead than HBPR
and n-epidemic protocols, respectively.

5.2. Varying the message size


Figure 4(a)(c) depicts the effect of increase in message size on the performance of EHBPR and its
comparison with other protocols. From Figure 4(a), it can be seen that as the message size increases,
the average residual energy of nodes decreases. This is justied by the fact that an increase in
message size increases the number of packets to be transmitted, which in turn consumes energy
of those nodes that participate in message forwarding. EHBPR has the highest value of average
residual energy among three protocols. This is because of the transmission factor used in EHBPR
that limits the number of message copies forwarded by a node in the network. This results in lesser
power consumption of the participating nodes. Figure 4(b) shows that EHBPR has the lowest num-
ber of dead nodes as compared with the HBPR and n-epidemic protocols because of the
residual_energy factor. In Figure 4(c), the overhead ratio increases with increase in message size.
The rate of increase is higher in the cases of HBPR and n-epidemic protocols, while it is lesser in
the case of EHBPR because of its transmission factor. This shows that EHBPR is energy efcient
and can be used effectively in sparse network scenarios.
The average residual energy of EHBPR is 15.35% more than HBPR and 14.18% more than
n-Epidemic protocol. The number of dead nodes in EHBPR is 67.2% lesser than HBPR and
67.47% lesser than n-Epidemic protocol. EHBPR has 76.19% and 84.4% lesser overhead than
HBPR and n-Epidemic protocols respectively.

Figure 4. Performance metrics versus message size. HBPR, history-based prediction for routing; EHBPR,
energy-efcient history-based prediction for routing.

Copyright 2015 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2015)
DOI: 10.1002/dac
ENERGY EFFICIENCY

Figure 5. Performance metrics versus message generation interval. HBPR, history-based prediction for
routing; EHBPR, energy-efcient history-based prediction for routing.

5.3. Varying the message generation interval


Figure 5(a)(c) shows the effect of message generation interval on the performance of EHBPR and
its comparison with other protocols. It can be observed in Figure 5(a) that the value of average
residual energy increases with an increase in the message generation interval. This is due to the fact
that an increase in the message generation interval decreases the messages owing in the network.
This results in lesser number of scans and lesser number of messages transferred between the nodes;
and hence less energy gets consumed. EHBPR has the highest while HBPR has the lowest values of
average residual energy. Figure 5(b) shows that compared to HBPR and n-Epidemic protocols,
EHBPR yields less dead nodes. It checks the residual energy of a node before passing a message
to it using its Residual_energy factor. In Figure 5(a), it can be seen that the overhead ratio of
EHBPR is much lesser than HBPR and n-Epidemic protocol. This is due to the Transmission factor
that restricts the number of redundant copies of the same message owing in the network; therefore,
the overhead ratio is also reduced.
The average residual energy of EHBPR is 13.43% more than HBPR and 11.95% more than
n-epidemic protocol. The number of dead nodes in EHBPR is 60% lesser than HBPR and
59.07% lesser than n-epidemic protocol. EHBPR has 78.7% and 84.57% lesser overhead than
HBPR and n-epidemic protocols, respectively.

6. CONCLUSION

An energy-efcient routing protocol named EHBPR for infrastructure-less Oppnets has been pro-
posed in this paper. We have shown that EHBPR enhances the performance of HBPR in terms of
energy consumption. Four improvement factors, namely, perpendicular factor, transmission factor,
sparse_constant factor, and residual_energy factor have been incorporated in the HBPR so as to
make it an energy-efcient protocol for opportunistic networks. The EHBPR is compared with

Copyright 2015 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2015)
DOI: 10.1002/dac
S. K. DHURANDHER ET AL.

HBPR and n-epidemic protocols using various metrics such as average residual energy, number of
dead nodes, and overhead ratio by varying number of nodes, message size and message generation
interval. The simulation results obtained reveal that EHBPR outperforms HBPR and n-epidemic
protocols in terms of the aforementioned metrics. Thus, it reduces the energy consumption of nodes
and, hence, increases the network lifetime. The results obtained shows that EHBPR can increase the
average residual energy of nodes by 14.66% and 13.14% as compared with the HBPR and
n-epidemic protocols, respectively, with less overhead ratio.
In future, the performance of EHBPR can be compared with other energy-efcient routing pro-
tocols such as energy-aware spray and wait and Genetic algorithm-based energy-efcient routing,
to name a few. Few more energy constraints will be imposed to EHBPR so as to improve it further.
The EHBPR protocol can also be made secure by using a combination of cryptography techniques,
reputation-based techniques, and incentive-based techniques.

REFERENCES

1. Lilien L., Kamal Z. H., Bhuse V., Gupta A. Opportunistic networks: the concept and research challenges in privacy
and security. Proc. of NSF Intl. Workshop on Research Challenges in Security & Privacy for Mobile and Wireless
Networks, Miami, Mar. 2006, pp. 134147.
2. Toh C.-K. Ad hoc mobile wireless networks: protocols and systems. Prentice Hall, Englewood Cliffs, New Jersey,
2002.
3. Huang C.-M.., Lan K.-C., Tsai C.-Z. A survey of opportunistic networks. Proc. of the AINA 2008, Okinawa, Japan,
Mar. 25-28, 2008, pp. 16721677.
4. Dhurandher S. K., Sharma D. K., Woungang I., Chao H.-C. Performance evaluation of various routing protocols in
opportunistic networks. Proc. of IEEE GLOBECOM Workshop 2011, Houston, TX, USA, Dec. 59, 2011; pp.
10671071.
5. Pelusi L., Passarella A., Conti M. Opportunistic networking: data forwarding in disconnected mobile ad hoc net-
works. IEEE Communications Magazine, vol. 44, Issue 11, November 2006, pp. 134141.
6. Fall K. A delay-tolerant network architecture for challenged internets. in Proceedings of ACM SIGCOMM 2003,
Karlsruhe, Germany, 25-29 August, 2003, pp. 2734.
7. Jain S., Fall K., Patra R. Routing in a delay tolerant network. in Proceedings of ACM SIGCOMM 2004, 30 Aug.-3
Sept. 2004, pp. 145158.
8. Dhurandher S. K., Sharma D. K., Woungang I., Bhati S. HBPR: History based prediction for routing in
infrastructure-less opportunistic networks. Proc. of (AINA-2013), Barcelona, Spain, Mar. 25-28, 2013, pp. 931936.
9. Dhurandher S. K., Sharma D. K., Woungang I., Saini A. Efcient routing based on past information to predict the
future location for message passing in infrastructure-less opportunistic networks. Journal of Supercomputing,
Springer, USA, 2014, doi: 10.1007/s11227-014-1243-5, June 28, 2014.
10. Lu X., Hui P. "An energy-efcient n-epidemic routing protocol for delay tolerant networks", Procs of the 5th IEEE
Intl. Conference on Networking, Architecture, and Storage, Macau, 15-17 July 2010, pp. 341347.
11. Safa H., Mirza O. A load balancing energy efcient clustering algorithm for MANETs. International Journal of
Communication Systems, Wiley, vol. 23, Issue 4, Apr. 2010, pp. 463483.
12. Dhurandher S. K., Misra S., Obaidat M. S., Bansal V., Singh R. P., Punia V. EEAODR: an energy-efcient ad hoc
on-demand routing protocol for mobile ad hoc networks. International Journal of Communication Systems, Wiley,
vol. 22, Issue 7, July 2009, pp. 789817.
13. Mehmood A., Khan S., Shams B., Lloret J. Energy-efcient multi-level and distance-aware clustering mechanism
for WSNs. International Journal of Communication Systems, Wiley, DOI: 10.1002/dac.2720, online: 17 DEC 2013.
14. Tanwar S., Kumar N., Niu W.-J. EEMHR: energy-efcient multilevel heterogeneous routing protocol for wireless
sensor networks. Intl. Journal of Communication Systems, Wiley, vol. 27, Issue 9, Sept. 2014, pp. 12891318.
15. Gao S., Zhang L., Zhang H. Energy-aware Spray and Wait routing in mobile opportunistic sensor networks. Proc. of
the 3rd IEEE Intl. Conference on Broadband Network and Multimedia Technology, Beijing, China, Oct. 26-28,
2010, pp. 10581063.
16. Patel V. G., Oza T. K., Gohil D. M. Vibrant energy aware spray and wait routing in delay tolerant network. Journal
of Telematics and Informatics, March 2012, vol. 1, No. 1, 4347. ISSN: 2303-3703.
17. Chilipirea C., Petre A.-C., Dobre C. Energy-aware social-based routing in opportunistic networks. Proc. of WAINA
Workshop, Barcelona, Spain, Mar. 25-28, 2013, pp. 791796.
18. Dhurandher S. K., Sharma D. K., Woungang I., Gupta R., Garg S. GAER: Genetic algorithm based energy-efcient
routing protocol for infrastructure-less opportunistic networks. Journal of Supercomputing, Springer, vol. 69, Issue
3, Sept. 2014, pp 11831214.
19. Chen I. C. K., Coffey J. T., Mudge T. N. Analysis of branch prediction via data compression. Proc. of the 7th Intl.
Conference on Architectural Support for Programming Languages & Operating Systems (ASPLOS VII),
Cambridge, MA, USA, Oct. 1-4, 1996, pp. 128137.
20. Vahdat A., Becker D. Epidemic routing for partially connected ad hoc networks. Technical Rept. CS-2000-06,
Department of Computer Science, Duke University, Durham, NC, USA, 2000.

Copyright 2015 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2015)
DOI: 10.1002/dac
ENERGY EFFICIENCY

21. Spyropoulos T., Psounis K., Raghavendra C. S. Spray and wait: an efcient routing scheme for intermittently
connected mobile networks. Proc. of ACM SIGCOMM Workshop on Delay-Tolerant Networking (WDTN 05),
Philadelphia, Pennsylvania, USA, 22-26 Aug. 2005, pp. 252259.
22. Hui P., Crowcroft J., Yoneki E.. Bubble rap: social based forwarding in delay tolerant networks. Proc. of the 9th
ACM Intl. Symposium on Mobile ad hoc Networking and Computing (MobiHoc 2008), New York, USA, 27-30
May, 2008, pp 241250.
23. Keranen A. Opportunistic network environment simulator. Special Assignment Report, Helsinki University of Tech-
nology, Department of Communications and Networking, May 2008.
24. Dhurandher S. K., Sharma D. K., Woungang I. Energy-based performance evaluation of various routing protocols in
infrastructure-less opportunistic networks. Journal of Internet Services and Information Security (JISIS), ISEP/IPP,
vol. 3, Issue 1/2, No. 1, Feb. 2013, pp. 3748.

Copyright 2015 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2015)
DOI: 10.1002/dac

You might also like