Hindawi Publishing Corporation

International Journal of Distributed Sensor Networks

Volume 2015, Article ID 567269, 15 pages
http://dx.doi.org/10.1155/2015/567269

Research Article
A Clustering Algorithm Based on Mobility Properties in
Mobile Ad Hoc Networks

Young-jun Oh and Kang-whan Lee

Department of Computer Science Engineering, Korea University of Technology and Education, Cheonan 330708, Republic of Korea

Correspondence should be addressed to Kang-whan Lee; kwlee@koreatech.ac.kr

Received 19 December 2014; Revised 3 April 2015; Accepted 3 April 2015

Academic Editor: Neil Y. Yen

Copyright © 2015 Y.-j. Oh and K.-w. Lee. This is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly
cited.

We propose a context-awareness routing algorithm—the DDV- (Dynamic Direction Vector-) hop algorithm—in mobile ad hoc
networks. The existing algorithm in MANET has the limitations of the dynamic network topology and the absence of network
expandability of the mobility of nodes. The proposed algorithm performs cluster formation for the base station using the range of
direction and threshold of velocity. We calculate the exchange of the cluster head node probability using the direction and velocity
for maintaining cluster formation. The DDV-hop algorithm, a probabilistic routing protocol for such networks, is proposed and then
compared to the earlier presented algorithms through simulations. The simulations are conducted on a number of clusters, network
areas, transmission ranges, and velocity of nodes in mobile networks. Our results suggest that the DDV-hop algorithm demonstrates
efficiency of eventual delivery and maintains the proper number of clusters and cluster members regardless of topology changes
with a lower communication overhead in several interesting environments.

1. Introduction cannot be built, as in disaster areas or war zones [4, 5].

In recent years, with the development of wireless communi-
cation technologies and various sensor nodes which led to
the emergence of the low-power and low-cost RF technology,
wireless sensor networks have become the object of much
scholarly attention. Wireless sensor networks consist of wire-
less devices based on various protocols to facilitate wireless
transmission and a processing unit to process the data. Their
main characteristics include the network expandability, self-
recovery, and multicasting. Wireless sensor networks were
originally designed to build a temporary network for emer-
gencies, such as disasters, as well as for battlefield surveillance
or other uses of the military. Later on, the applications of
wireless sensor networks have become more versatile [1–3].
Mobile ad hoc network (MANET) consists of nodes that
have mobility. The nodes both send and receive messages
and can communicate with each other. Thus, the network
builds its own network structure that is not dependent on
infrastructure. Due to the characteristics of mobile ad hoc
networks, MANET has been used in the environments of
poor communication, such as those where infrastructure
Cluster networks using properties of the node have been
studied particularly extensively. A cluster network is divided
into mobile groups by setting the rules of communication
protocols. The mobile groups elect the cluster head node for
managing groups. The cluster head node aggregates sensing
data sent by a cluster member node and then sends the
aggregated data to the base station [6]. However, if the cluster
head node is discharged or does not move due to poor
environmental conditions, the routing path is disconnected
from the cluster head node. Therefore, the network cannot
communicate in a stable fashion. In that case, the cluster
network reelects the cluster head node to recover the routing
path. However, dynamic properties of a node cause frequent
disconnections and routing recovery. In addition, the nodes
have constraints, such as limited transmission bandwidth and
energy, as well as topology changes. To reduce the number
of disconnections and in order to set the routing path,
the network sends a control packet. Eventually, the load is
increased [7–11].
In this paper, we propose a context-awareness routing
algorithm: the DDV- (Dynamic Direction Vector-) hop
2 International Journal of Distributed Sensor Networks
algorithm. This algorithm considers dynamic properties of
nodes, such as velocity, direction, to form clusters. In order to
maintain clusters, the cluster head node finds cluster member
nodes with similar velocity and direction.
The rest of the paper is organized as follows. Section 2
describes relevant research related to the proposed algorithm.
Section 3 provides a detailed description of the DDV-hop
algorithm. This analyses the algorithm using network models
and optimization. The comparison of the performance of
the DDV-hop algorithm with that of the Lowest-ID and the
MOBIC algorithms is provided in Section 4. Finally, Section 5
shows the conclusions of the present paper.
2. Related Works
In MANET, Yan Zhang and Jim Mee Ng proposed the
distributed group mobility adaptive (DGMA) clustering algo-
rithm. In the DGMA, the node measures location informa-
tion of the node by GPS and calculates the variation of the
node location. Calculating variation of nodes is shown as
Δ𝑥𝑇𝑖 = (𝑥𝑡𝑖 − 𝑥𝑇𝑖 ) ,
(1)
Δ𝑦𝑇𝑖 = (𝑦𝑡𝑖 − 𝑦𝑇𝑖 ) ,
where Δ𝑥𝑇𝑖 and Δ𝑦𝑇𝑖 are variations of the node location. 𝑡 is
the current time, 𝑇 is previous time, 𝑥𝑡𝑖 , 𝑦𝑡𝑖 , 𝑥𝑇𝑖 , 𝑦𝑇𝑖 are the
coordinates of the node 𝑖 at time 𝑡 and 𝑇 respectively. Using
the difference in the node locations, it is possible to form and
maintain the clustering networks from the node analyses of
velocity and direction [12].
Similarly to the MANET environment, VANET (vehicu-
lar ad hoc network) has been reported to maintain routing
by frequently changing topology. MDBC (Moving Direction
Based Greedy Routing Algorithm) uses the direction prop-
erty of several properties for routing. To set the direction
of the node that has mobility, if the current coordinate 𝑋
(𝑋-position) is bigger than the former coordinate 𝑋, the
direction of the node is east. If the current coordinate 𝑋 is
smaller than the former coordinate 𝑋, then the direction of
the node is west. If the current coordinate 𝑌 (𝑌-position)
is bigger than the former coordinate 𝑌, the direction of the
node is north. If the current coordinate 𝑌 is smaller than the
former coordinate 𝑌, the direction of the node is south. When
the direction of the node is set, the node broadcasts hello
message and stores the location and direction of the neighbor
node in the table. To forward a packet to the destination
node, the node uses DREQ (Destination REQuest) and DREP
(Destination REPly) and finds a stable routing path [13].
Lili Hu proposed GPSR (Greedy Perimeter Stateless
Routing) considering the direction and velocity of a node
in VANET. In the GPSR algorithm, the node sends hello
message that periodically includes velocity and direction,
grasps current coordinates of neighbor nodes, and calculates
the current coordinates using velocity and direction. The
current coordinates of the node are shown in
𝑋𝑡 = 𝑋0 + (𝑡 − 𝑡0 ) ∗ 𝑉𝑥 ,
(2)
𝑌𝑡 = 𝑌0 + (𝑡 − 𝑡0 ) ∗ 𝑉𝑦 ,
where 𝑡 is current time and 𝑡0 is the time when hello message
was received. 𝑉𝑥 and 𝑉𝑦 refer to velocity. Then, the node
calculates the probability of transmission considering the
velocity and direction of the node and forwards the packet
[14].
The lowest ID cluster algorithm (LIC) is an initial clus-
tering algorithm in MANET. This algorithm performs cluster
formation using a unique ID of a node. To form a cluster, the
network chooses the cluster head node with the minimum ID.
The cluster head node chooses a cluster member node with
the ID higher than its own. When a node is located within
the transmission range of two or more cluster head nodes,
the node, called a gateway node, is used for routing between
clusters. However, since the node hears only the nodes with
IDs higher than that of the cluster head node, the network
consumes energy inefficiently and communication between
the nodes is frequently disconnected [15].
The Mobility Based Metric for Clustering (MOBIC)
algorithm proposes a local metric for cluster formation using
speed properties in MANET. For cluster formation, network
measures the speed of a node using its coordinates and the
interval time by hello message. The network chooses the
cluster head node that has a low speed variance. After the
cluster head node is selected, it joins a node as a cluster
member node in a 2-hop area. However, since the MOBIC
algorithm elects the node with a low speed as the cluster
head node, the network topology frequently changes and the
communication between nodes is disconnected [16].
3. System Modeling and Methods
In this paper, we propose the DDV-hop algorithm which per-
forms and maintains cluster formation taking into account
mobility properties, such as direction and velocity. The
existing clustering algorithm has the problem that the topol-
ogy gets frequently changed by node mobility. Therefore,
topology broadcasts change topology information, the net-
work load increases, and there is a lack of the network
expandability. According to the mobility of a node, the
proposed algorithm forms clusters from the nodes with
similar direction and velocity. Due to this method of cluster
formation, the topology change of the network is reduced,
packet delivery ratio increases, and the load of the network
decreases. To form the initial cluster, the proposed algorithm
sets a coverage as shown in
𝑅𝐶 = 𝑘 × 𝑅𝑇 , 𝑘 = 1, 2, 3, . . . , 𝑁, (3)
where 𝑅𝐶 is coverage, 𝑅𝑇 is transmission range, and 𝑘 refers to
the number of hops [17]. In this paper, we set the coverage that
confirms the information of the node by the number of hops
and the transmission range. To elect the initial cluster head
node, the base station uses coverage using the following.
First, the network elects the initial cluster head node by
coverage (see Figure 1).
In Figure 1, white circles mark cluster member nodes and
black circles mark initial cluster head nodes. The black circle
labelled “BS” refers to the base station and the broken-line
circle means the coverage of the base station by the number
International Journal of Distributed Sensor Networks
k=3
k=2
k=1
BS
Figure 1: Selection of the initial cluster head node by coverage.
of hop. The network builds the base station located in the
center of network. When the number of hops is 𝑘 = 𝑖, the base
station sets coverage. The base station confirms the direction
of the node and measures the number of nodes in the range
of direction set by the base station in each region. In each
region, the base station chooses the range of direction with
the highest number of nodes and elects a node with a similar
direction to that of the initial cluster head node. In each
region, the cluster head node cannot be elected by the base
station, the base station increases the number of hops, and
the procedure is repeated until the initial cluster head node is
elected by the direction in each region [18].
3.1. The DDV-Hop Algorithm Scenario. With regard to per-
forming cluster formation, the following two assumptions are
made in the paper.
First, the base station is located in the center of the
network.
Second, the network area is divided into four regions by
the base station.
Based on these assumptions, the network initializes
the network environment and the nodes randomly set the
location in the network. Next, the direction of the nodes
is randomly set by the base station. Furthermore, the node
arbitrarily establishes its properties, such as velocity. The
energy of the node is initialized.
In Figure 2, the cluster head node is elected by coverage
and direction. In this paper, we assume that the number
of hops is 1. Thus, the base station confirms all nodes in
the network and elects the initial cluster head node in each
region. Afterwards, the cluster head node searches for other
nodes with similar properties. The nodes located on the
boundary are included, based on their properties of direction
and velocity, into the same region; furthermore, the nodes
form a cluster in that region.
When the cluster head node is elected, it sends an
announcement message to the nodes nearby. A normal node
3
BS
Figure 2: The procedure of the initial cluster formation.
receives the announcement message and selects the cluster
head node that is closest to it and has the same direction
and velocity. Then, the nodes send join message to select
the cluster head node. The cluster head node receives that
join message and registers those nodes as its cluster member
nodes. At this point, cluster formation is completed. Now,
the network has a routing structure that allows the cluster
member nodes to send messages to the cluster head node and
the latter node to send a message to the base station. As each
node has mobility, the nodes get out of the cluster. In this case,
the node is included into the region where it finally belongs.
For example, when a node belongs to region 𝐴, but has moved
to region 𝐵, the node performs cluster formation again using
its direction and velocity. The cluster member node maintains
its routing path.
According to the procedure of the cluster formation, the
relationship of friendship with the cluster head node is as
follows (see Figure 3).
As shown in Figure 3, nodes 𝑋 and 𝑌 are included into the
transmission range of the cluster head nodes 𝐴 and 𝐵. This is a
simple example showing an efficient clustering method which
selects the cluster head node. In the existing algorithm, a node
selects the cluster head node that is closest to it, regardless of
direction and velocity. However, in the DDV-hop algorithm,
the node compares its own direction and velocity with those
of the cluster head node and selects the cluster head node
with similar direction and velocity. For example, node 𝑋
is included by the cluster head nodes 𝐴 and 𝐵. Node 𝑋
compares the direction and velocity of its own with those of
both cluster head nodes and then selects the cluster head node
𝐵, as this cluster head node has similar properties. Therefore,
the load of the network is reduced and a stable topology is
maintained with a lower communication overhead in mobile
ad hoc networks [19].
3.2. Network Models and Analysis of the DDV-Hop Algo-
rithm. In this paper, we propose the DDV-hop algorithm
which performs and maintains cluster formation considering
the properties of nodes such as direction and velocity.
Specifically, the DDV-hop algorithm considers the following
properties. A set of nodes defines 𝑁 in the network; each node
4 International Journal of Distributed Sensor Networks

j  ij
Dir
ij
Dirj
X
i 󳰀
Diri󳰀
i
Diri
A B.S B
i 󳰀 j󳰀
j󳰀

Diri󳰀 j󳰀 Dirj󳰀

Velocity of cluster head node Threshold of direction

Velocity of cluster member node Difference direction between nodes
Standard line of node direction Direction of cluster head node
Difference velocity between nodes Direction of cluster member node
Cluster head node Cluster member node

Figure 3: The procedure of join the cluster head node by the attributes (velocity, direction, location, energy, and transmission range).

has a unique ID. Furthermore, each node has direction (Dir𝑖 ) Step 1. For electing the initial cluster head node, the base
and velocity (V𝑖 ) (see (4)): station divides the network into four regions (see (6)):
network area
𝑁𝑖 = {Dir𝑖 , V𝑖 } , (4) (6)
= {Region𝑍 | Region𝑍 ∈ network area} ,
where Dir𝑖 is the direction of node 𝑖 and V𝑖 is the velocity of where Region𝑍 is the network area divided by the base
node 𝑖. The direction and velocity of each node are expressed station. When the network is divided by (6), the base station
in sets the range of direction. The range of direction is expressed
using rs, range of angle resolution. When rs is higher, then
Dir𝑖 = {0 ≤ Dir𝑖 < 2𝜋} , the number of cluster head nodes increases. The method of
(5) setting the range of direction is expressed in (7) and (8) as
V𝑖 = {0 ≤ V𝑖 < 𝑛} , follows:
rs = {0 < rs ≤ 𝑛} , (7)
where 𝑛 is a real number for setting the velocity of node 𝑖 and
node 𝑖 sets the velocity between 0 and 𝑛. { 𝛼𝜋 (𝛼 + 1) 𝜋 }
{ rs | [ , ]}
As the DDV-hop algorithm performs and maintains DirBS (𝑖) = { rs rs } , (8)
cluster formation by the direction and velocity of nodes, { }
{ 𝛼 = 0, 1, 2, . . . , 2rs − 1}
the network can communicate for a long time. For the
communication of the network, the proposed algorithm is where DirBS (𝑖) refers to set of the range of the direction
divided into two major procedures, that of cluster formation by the base station. 𝑛 means a real number for setting the
and the other of exchanging the cluster head node (see rs. The base station sets the rs between 0 and 𝑛. In the set
Sections 3.2.1 and 3.2.2 for further detail). range of direction by DirBS (𝑖), the number of the cluster head
nodes is determined. The cluster head node is elected by
the coverage and direction in each Region𝑍 ; the direction of
3.2.1. The Procedure of Cluster Formation. According to one node is included into the set range of direction. Equation (9)
of the assumptions specified above, the base station is located presents the method of electing a cluster head node:
in the center of the network area; the initial cluster is formed
according to the following procedure. CH = {𝑁𝑖 | 𝑁𝑖 ∈ Region𝑍 ∧ Dir𝑖 ∈ DirBS } . (9)
International Journal of Distributed Sensor Networks
In (9), CH refers to the cluster head node group in the
network. After selecting the cluster head node using (7) and
(8), the cluster head node measures the similarity of direction
for selecting cluster member nodes.
Step 2. The method of measuring similarity of direction
compares the differences in direction between the nodes
(Dir𝑖𝑗 ) and sets a threshold of direction (Dir𝑖 rs ) using a range
of direction. This method is presented in
󵄨 󵄨
Dir𝑖𝑗 = 󵄨󵄨󵄨󵄨Dir𝑖 − Dir𝑗 󵄨󵄨󵄨󵄨 ,
󵄨󵄨 𝜋 󵄨󵄨 (10)
Dir𝑖 rs = {Dir𝑖𝑗 | Dir𝑖𝑗 ∈ 󵄨󵄨󵄨󵄨 󵄨󵄨󵄨󵄨 ∧ 𝑁𝑗 ∈ Region𝑍 } ,
󵄨 rs 󵄨
where Dir𝑖𝑗 means the difference of direction between the
nodes, Dir𝑖 rs is the threshold of direction, and Region𝑍 is
the network area divided by the base station. When the
threshold of direction rs increases, the range of the threshold
of direction gets narrower.
Step 3. The method of measuring similarity of the velocity
between the nodes compares the differences of velocity
between the nodes and sets a threshold of velocity using a
deviation of velocity; the method is presented in (11) and (12)
as follows:
󵄨 󵄨
V𝑖𝑗 = 󵄨󵄨󵄨󵄨V𝑖 − V𝑗 󵄨󵄨󵄨󵄨 , (11)
󵄨 󵄨
ΔV𝑖 = {𝑗 | V𝑖𝑗 ≤ 󵄨󵄨󵄨󵄨𝜎V𝑖 󵄨󵄨󵄨󵄨 ∧ 𝑁𝑗 ∈ Region𝑍 } , (12)
where V𝑖𝑗 means the difference of velocity and ΔV𝑖 refers to a
threshold of the velocity. 𝜎V𝑖 is a factor that sets a threshold of
velocity.
Step 4. The cluster head node selects cluster member nodes
through measuring the similarity of direction and velocity in
Steps 2 and 3; the method is presented in
CM𝑖 = {𝑗 | Dir𝑖𝑗 ∈ Dir𝑖 rs ∧ V𝑖𝑗 ∈ V𝑖 } ,
(13)
HG𝑖 = {CM𝑖 | CM1 ∪ CM2 ∪ ⋅ ⋅ ⋅ ∪ CM4 } ,
where CM𝑖 is a group of the selected cluster member nodes
and HG𝑖 is a cluster that consists of the cluster member nodes.
3.2.2. The Procedure of Exchange of the Cluster Head Node.
To reduce the network load, the exchange of the cluster head
node is needed. After performing the cluster formation, the
procedure of maintaining cluster is as follows. The cluster
member nodes calculate successful probability by direction,
velocity, and residual energy. The exchange of the cluster head
follows the procedure specified below.
{ (𝑓 (𝐸𝑗 (𝑡)) × 𝜔𝑗 (𝑡) × 𝑓 (DV (𝑡)) × 𝜔𝑗 (DV (𝑡)))󵄨󵄨󵄨max ,
𝑃𝑗 (DDV (𝑡)) = {
{0,
5
Step 1. The cluster member nodes in the cluster calculate the
exchange probability (𝐸𝑗 (𝑡)) using the distance between the
nodes, the energy from transmitting data, and their average
residual energy. The method is shown in [20, 21]
𝑃 (𝐸𝑗 (𝑡))
(14)
𝐺 [𝐸𝑗 (𝑙) × 𝜔 (𝑙) + 𝐸𝑗 (dist) × 𝜔 (dist)]
= ∑ ,
𝑗∈CM𝑖 𝑁 − 𝐺 (𝑟 mod (𝑁/𝐺)) × (1/𝑁) ∑𝑛𝑗=1 𝐸𝑗 (𝑡)
where 𝑃𝑗 (𝐸𝑗 (𝑡)) is the probability of exchange using energy,
𝐸𝑗 (𝑙) is the energy by transmitting data, and 𝐸𝑗 (dist) is the
energy of the distance between nodes. 𝐺 is the cluster group
and 𝑟 is round. 𝜔(𝑙) and 𝜔(dist) refer to the weight factors
of transmitting coefficient value between the distance and
the residual energy of the nodes. When the cluster member
node finishes the calculation of the probability of exchange
using energy, the information forms a group in the cluster.
The method is shown in
𝑓 (𝐸𝑗 (𝑡)) = {𝑃𝑗 (𝐸𝑗 (𝑡)) | 𝑃𝑗 (𝐸𝑗 (𝑡)) , 𝑗 ∈ 𝑖} , (15)
where 𝑖 is the cluster head node 𝑖 and 𝑗 is the cluster member
node 𝑗 assigned to the cluster head node 𝑖. 𝑓(𝐸𝑗 (𝑡)) refers to
the group of the exchange probability using the energy of the
cluster member node 𝑗.
Step 2. The probability of exchange by velocity and the
direction is calculated using the similarity of direction and
velocity and their threshold; the method is shown in
Dir𝑖𝑗 V𝑖𝑗
𝑃𝑗 (DV (𝑡)) = ∑ ⋅ , (16)
𝑗∈CM𝑖 Dir𝑗
rs
ΔV
where 𝑃𝑗 (DV(𝑡)) is the probability of exchange by velocity and
direction. When the cluster member node finishes calculating
the probability of exchange by velocity and direction, this
information forms a group in the cluster. The method is
shown in
𝑓 (DV (𝑡)) = {𝑃𝑗 (DV (𝑡)) | 𝑃𝑗 (DV (𝑡)) , 𝑗 ∈ 𝑖} , (17)
where 𝑓(DV(𝑡)) is the group of the probability of exchange
by velocity and direction of the cluster member node 𝑗.
Step 3. Using the calculated probabilities of exchange by Steps
1 and 2, the cluster member node calculates final probabilities
using 𝑓(𝐸𝑗 (𝑡)) and 𝑓(DV(𝑡)). The method is shown in
󵄨󵄨
𝐶𝑗 (𝑡) = 0
(18)
𝐶𝑗 (𝑡) = 1,
6 International Journal of Distributed Sensor Networks

Table 1: Table of node’s properties (example).

RegionZ
ID Dir Velocity Region
DirBS (i) 0 1 2 3
𝑛 1 7 0

CHID — — — —

Table 2: Region table in the base station (example).

Region𝑍 Cut point

DirBS (𝑖) 0 1 2 3 RegionZ
CHID — — — —
DirBS (i) 0 1 2 3

where 𝑃𝑗 (DDV(𝑡)) means the final probability of exchange CHID 4 — — —

using energy, velocity, and direction of the node and 𝜔(𝐸𝑗 (𝑡))
and 𝜔(DV(𝑡)) refer to the weight of the probability of
exchange by the energy and the probability of exchange Cut point
by velocity and direction. 𝐶𝑗 (𝑡) is the function elected by
RegionZ
the cluster head node. When the cluster member node was
elected as the cluster head node, 𝐶𝑗 (𝑡) is 1. Thus, 𝑃𝑗 (DDV(𝑡)) DirBS (i)
is 0. When the cluster member node was not elected as 0 1 2 3
the cluster head node, then 𝐶𝑗 (𝑡) is 0. Thus, 𝑃𝑗 (DDV(𝑡)) is
calculated as shown in (18). CHID 4 5 — —

Step 4. The cluster head node elects the cluster member

node that has the maximum probability of 𝑃𝑗 (DDV(𝑡)) as Cut point
the exchange cluster head node in the cluster. The exchange
cluster head node compares its own direction and velocity Figure 4: The procedure of the selection of the cluster head node.
with those of the nodes in the cluster. When the nodes have
similar direction and velocity, they are joined to the cluster by
the exchange of the cluster head node.
Table 3: Table of the cluster head node (example).

3.3. The DDV-Hop Algorithm Modeling. In the DDV-hop CMID 1 3 13 15

algorithm, each node has a property table in the network (see
Table 1).
In Table 1, ID is the unique identification number of the
node, Dir is the direction of the node, Velocity is the velocity
of the node, and Region refers to the network area divided
by the base station. When the nodes are assigned to the
network, their properties are stored in the corresponding
property tables. Using the information from these property
tables, network performs cluster formation and changes the
cluster head node.
3.3.1. Cluster Formation Modeling. According to one of the
assumptions specified above, the base station is located in the
center of the network area and the network is divided into
regions by the base station. Then, the base station generates a
region table to store cluster information (see Table 2).
In Table 2, Region𝑍 is the network area divided by the
base station, DirBS (𝑖) is a range of direction by the base
station, and CHID is a head node from one of the selected
cluster nodes. Region tables store the cluster head node IDs.
The procedure of storing the cluster head node ID is as follows
(see Figure 4).
In Figure 4, the region table searches velocity and direc-
tion of the nodes and selects the cluster head node located
𝑃(DDV(𝑡)) — — — —
in the coverage by the base station. The direction is included
into the threshold of direction and the velocity is included
in the threshold of velocity. Finally, the selected cluster head
node is stored in the region table. The selected cluster head
node generates a table to store the cluster member node (see
Table 3).
In Table 3, CMID refers to the cluster member node ID
and 𝑃(DDV(𝑡)) is the probability of exchange of the cluster
head node. The cluster head node measures the similarity
of the direction and velocity of normal nodes located in
the region. When the value of similarity is included in the
threshold, a normal node is registered as a cluster member
node. The cluster head node stores the cluster member node
ID in the table and performs cluster formation. The procedure
of cluster formation is as follows (see Figure 5).
In Figure 5(a), the network area is divided into regions by
the base station. Figure 5(b) shows the selection of the cluster
head node by coverage of the base station, and the direction
of the node is included by the base station. In Figure 5(c),
the cluster head node selects cluster member nodes that are
similar to it in terms of direction and velocity. Figure 5(d)
International Journal of Distributed Sensor Networks 7

13 24 13 24

1 2 1 2
15 22 Dir4 ≤ DirBS (𝛼) Dir5 ≤ DirBS (𝛼)
22
15

3 4 5 6 3 4 5 6

7 8 9 10 7 8 9 10
21 21
16 11 12 16 Dir8 ≤ DirBS (𝛼)
11
Dir9 ≤ DirBS (𝛼)
12

18 19 18 19

B.S B.S

(a) Dividing the network area by the base station (b) Selection of the cluster head node by direction and velocity

Dir4,13 ≤ Dir4rs
4,13 ≤ Δ4

13 Dir4,1 ≤ Dir4rs24 13 24
Dir4,15 ≤ Dir4rs 4,1 ≤ Δ4
4,15 ≤ Δ4
1 2 1 2
22 15 22
15

3 4 5 6 3 4 5 6
4
Dir4,3 ≤ Dir rs
4,3 ≤ Δ4
7 8 9 10 7 8 9 10
21 21
16 16
11 12 11 12

18 19 18 19

B.S B.S

(c) Selection of the cluster member node by direction and velocity of the (d) Cluster formation network
cluster head node

Figure 5: The procedure of cluster formation by the DDV-hop algorithm.

shows the final result of cluster formation. Pseudocode 1
shows the pseudocode of cluster formation.
As shown in Pseudocode 1, 𝑁[𝑖] is the node that is
located in the network. 𝑖 refers to the initial cluster head.
For the initial cluster formation, the network depends on the
threshold of direction (DirBS (𝑖)) in (8) and to the threshold of
velocity (𝜎V𝑖 ) in (12). The node is selected including the cluster
head node when a node is compared to the threshold of
direction and refers to the threshold of velocity. The election
of the node is registered as role CH and cluster member node
groups. The selected cluster head node compares its velocity
and direction to those of other nodes in each region. When
the remaining normal nodes are similar to the cluster head
node in velocity and direction, the normal node is registered
as a cluster member node and is stored in CM𝑖 .
3.3.2. Maintaining Cluster Formation Modeling. The cluster
head node is exchanged to enhance the efficiency of energy
consumption and to ensure continuous communication. For
the exchange of the cluster head node, the cluster member
nodes calculate the exchange probability and send the respec-
tive message to their cluster head node. When a cluster head
node receives this message, it stores the exchange probability
in the cluster head node table (see Table 4).
8 International Journal of Distributed Sensor Networks

Procedure: initialize cluster formation of the DDV-hop algorithm

Input
Dir𝑖𝑗 : the difference of direction between nodes 𝑖 and 𝑗
Dir𝑖𝑗 rs : the threshold of direction
Output
Cluster head node group CH
Cluster member node group of all cluster head node 𝑖 CM
Begin
initialize CH
initialize Region𝑍
𝑖←0
for 𝑎 = 0 to 𝑧
for 𝑏 = 0 to 𝑁
if (𝑁[𝑏] is located Region𝑎 ) then
Region𝑎 ∋ 𝑁[𝑏]
if (𝑁[𝑏] is located in the coverage of the base station and has different direction of CH)
then
CH[𝑖] ← 𝑁[𝑏]
𝑖=𝑖+1
end
end
for 𝑎 = 0 to 𝑖
for 𝑏 = 0 to 𝑁 − 1
if (Region including CH[𝑎] ≡ Region including 𝑁[𝑏]) then
DirCH[𝑎]𝑁[𝑏] ← |DirCH[𝑎] − Dir𝑁[𝑏] |
]CH[𝑎]𝑁[𝑏] ← |]CH[𝑎] − ]𝑁[𝑏] |
if (0 ≤ (DirCH[𝑎] /|Dir𝑁[𝑏] rs |) ≤ 1) and (0 ≤ (]CH[𝑎] /|Δ]|) ≤ 1) then
CM𝑎 ∋ 𝑁[𝑏]
end
end
output the cluster head node group CH
output the cluster member node groups of all cluster head node 𝑖 of cluster head node 𝑖 CM𝑖
end

Pseudocode 1: The pseudocode of cluster formation.

Table 4: Storing the probability of exchange in the cluster head node

table CH(4) (example). CMID 1 3 13 15

CMID 1 3 13 15
P(DDV(t)) 0.87 0.92 0.54 0.49
𝑃(DDV(𝑡)) 0.87 0.92 0.54 0.49

Cut point

CMID 1 4 13 15
In the cluster head node table, the cluster head node com-
pares the probabilities of exchange; furthermore, the cluster
member node with the maximum probability of exchange P(DDV(t)) 0.87 0 0.54 0.49
becomes the next cluster head node. Then, for efficient
energy consumption and continuous communication, the Figure 6: The procedure of updating cluster information.
preceding cluster head node has “0” probability of exchange.
The preceding cluster head node also sends the cluster
information message making the cut point that permutates
a cluster member node into the maximum 𝑃(DDV(𝑡)) as a exchange CHID[𝑖]into maximum 𝑃(DDV(𝑡))node as shown
cluster head node in (18). The procedure of editing cluster in the region table (see Figure 7).
information is as shown in Figure 6. As shown in Figure 7, the base station receives the
The next permutation is that the cluster head node message that there is a cluster head node exchange and
receives cluster information and sends message that the changes the cluster head node information in the region table.
cluster head node is exchanged. When the base station The procedure of the exchange of the cluster head node is
receives the message making cut point.And the head node shown in Figure 8.
International Journal of Distributed Sensor Networks
RegionZ
DirBS (i) 0 1 2 3
CHID 4 5 8 9
Cut point
RegionZ
DirBS (i) 0 1 2 3
CHID 3 5 8 9
Figure 7: The procedure of updating the cluster head node in the
region table.
Figure 8(a) shows the network before the exchange of
the cluster head node. Figure 8(b) shows the procedure of
the exchange of the cluster head node by the probability
of exchange. Figure 8(c) shows the resulting network after
the exchange. Pseudocode 2 shows the pseudocode of the
exchange of the cluster head node.
In Pseudocode 2, 𝑁[𝑖] is the number of nodes located
in the networks. 𝐺 refers to the number of the cluster
head nodes and Gm is the number of cluster member
nodes. To maintain cluster formation, the cluster head node
calculates the probability of exchange of the cluster member
node considering energy, velocity, and direction in CM𝑖 .
When a cluster member node has the maximum probability
of exchange, it becomes the next cluster head node. The
exchanged cluster head node performs cluster formation.
4. Simulation and Results
In this paper, we demonstrate the efficient connection of
the proposed algorithm. We compare the LIC algorithm that
selects the cluster head node with the lowest ID algorithm
and the MOBIC algorithm which selects the cluster head
node with the slowest speed. The DDV-hop algorithm forms
a cluster where the nodes are similar in their direction and
velocity. Thus, we have a simulation by velocity, direction,
energy, and transmission. Equation (19) shows the packet
delivery ratio by velocity and direction:
PDR (𝑡)
flowBS (𝑡)
{
{ ⋅ arg max 𝑃𝑗 (DDV (𝑡)) , 𝐶𝑗 (𝑡) = 1
{
{ ∑𝑖∈CH ∑𝑗∈CM𝑗 flow𝑗 (𝑡)
={
{ flowBS (𝑡)
{
{ ⋅ arg max 𝑃𝑗 (DV (𝑡)) , 𝐶𝑗 (𝑡) = 0,
{ ∑𝑖∈CH ∑𝑗∈CM𝑗 flow𝑗 (𝑡)
where PDR(𝑡) is the packet delivery ratio of the network
[22, 23]. flowBS (𝑡) is the transmitted data received by the base
station in the proposed algorithm. CH refers to the group of
the cluster head nodes in the network and CM𝑖 is the group of
the cluster member nodes in the cluster head node 𝑖. flow𝑗 (𝑡)
is the data sent by the cluster member node to the base station.
9
Table 5: Simulation environments.
Parameter Value
1000 × 1000, 1200 × 1200, 1500 × 1500,
Network area (m2 )
1800 × 1800, and 2000 × 2000
Max velocity (m/s) 8, 12, 16, 20, and 24
Number of cluster head
16, 20, 24, 28, and 32
nodes
Number of nodes 2000
Transmission range (m) 400, 450, 500, 550, and 600
Transmission cycle (sec) 1
Simulation time (sec) 300
30, 60, 90, 120, 150, 180, 210, 240, and
Pause time (sec)
270
Weight of energy 0.5
Weight of direction and
0.5
velocity
Initial node energy (J) 100
arg max 𝑃𝑗 (DDV(𝑡)) refers to the base station that receives
the data by the proposed algorithm using direction, velocity,
and energy. arg max 𝑃𝑗 (DV(𝑡)) refers to the base station that
receives the data by the proposed algorithm using direction
and velocity. 𝐶𝑗 (𝑡) is the function whether the cluster member
node was elected or not. When the cluster head node was
exchanged, 𝐶𝑗 (𝑡) is 1. Thus, the PDR(𝑡) is calculated according
to the upper part of (19). When the cluster head node was
not exchanged, 𝐶𝑗 (𝑡) is 0. Thus, the PDR(𝑡) is calculated
according to the lower part of (19).
The moving model of the node has adopted a random
direction model that has set a new direction when the node
is located around the network area. Furthermore, we assume
that the base station is located in the center of the network and
that the network area is divided into four regions by the base
station. The simulation environments are shown in Table 5.
4.1. Number of Clusters. To confirm the packet delivery ratio,
we have simulated by the number of clusters. The number
of clusters depends on the direction and region. Therefore,
we assume that the regions imply a division into quadrants
and that the threshold of direction (Dir𝑖 rs ) is four to eight.
The network area was increased from 1000 × 1000, 1500 ×
1500 and 2000 × 2000, velocity as 1∼16 m/s. The number of
clusters ranged from 16 to 32 and transmission range was
450 m. Simulation results are shown in Figures 9–11.
The DDV-hop algorithm shows about 90% packet deliv-
ery ratio without a reference to the number of clusters when
(19)
the network area is 1000 × 1000 m2 (see Figure 9). As shown
in Figure 10, the DDV-hop algorithm yields about 78% packet
delivery ratio without a reference to the number of clusters
when the network area is 1500 × 1500 m2 . As shown in
Figure 11, the DDV-hop algorithm shows about 69% packet
delivery ratio without a reference to the number of clusters
when the network area is 2000 × 2000 m2 . By contrast, the
LIC algorithm generally shows 10% packet delivery ratio of
the number of clusters and the MOBIC algorithm shows
10 International Journal of Distributed Sensor Networks

13 24 13 24
0.54 0.54

1 2 1 2
0.87 22 0.87 22
15 15
0.49 0.49
3 4 5 6 3 4 5 6
0.92 0 0
0.92
7 8 9 10 7 8 9 10
21 21
16 11 12 16 11 12

18 19 18 19

B.S B.S

(a) Calculating exchange probability (b) Exchange of the cluster head node by the exchange probability

13 24

1 2
22
15

3 4 5 6

7 8 9 10
21
16 12
11

18 19

B.S

(c) The network cluster head node after the exchange

Figure 8: The procedure of the exchange of the cluster head node in the network.

20% packet delivery ratio. Therefore, the network area has
increased and the packet delivery ratio of the DDV-hop algo-
rithm has decreased and maintains a higher packet delivery
ratio than those of other cluster algorithms. These simulation
results suggest that the DDV-hop algorithm shows a more
continuous connection as compared to other algorithms.
4.2. Network Area. In this section, we proceed to the test to
see the packet delivery ratio according to the network area.
The network area is increased from 1000 × 1000 m2 to 2000 ×
2000 m2 . The velocity of node is 1∼16 m/s, transmission range
is 450 m, and the number of clusters is 16 to 32 (see Figures
12–14).
As shown in Figure 12, despite the increase of the network
area, the DDV-hop algorithm shows the packet delivery
ratio that has decreased from 91% to 61% when the number
of clusters is 16. Furthermore, as shown in Figure 13, the
DDV-hop algorithm shows the packet delivery ratio that has
decreased from 90% to 62% when the number of clusters is
24. Finally, the DDV-hop algorithm shows the packet delivery
ratio that has decreased from 88% to 63% when the number
of clusters is 32 (see Figure 14). The LIC algorithm shows
the packet delivery ratio that has decreased from 18% to 6%
when the network area is increased. The MOBIC algorithm
also shows the packet delivery ratio that has decreased from
33% to 13% when the network area is increased. In these
simulation results, the packet delivery ratio of the DDV-
hop, the LIC, and the MOBIC algorithms is decreased when
the network area is increased without a reference to the
number of clusters. However, the DDV-hop algorithm shows
International Journal of Distributed Sensor Networks 11

Procedure: exchange of the cluster head node of the DDV-hop algorithm

Input
Dir𝑖𝑗 : difference of direction between nodes 𝑖 and 𝑗
Dir𝑖𝑗 rs : threshold of direction
CH: Cluster head node group
CM𝑖 : Cluster member node group of allcluster head node 𝑖
𝐺: the number of cluster head node
Gm: the number of cluster member node
Output
Change cluster head node group CH󸀠
All change cluster member node group of cluster head node 𝑖 CM󸀠𝑖
Begin
initialize CH󸀠
nextCH ← 0
for 𝑎 = 0 to 𝐺
for 𝑏 = 0 to Gm
if (CH[𝑎] ∋ CMCH[𝑎] [𝑏]) then
calculate 𝑓(𝐸𝑏 (𝑡)) of CMCH[𝑎] [𝑏]
calculate 𝑓(DV(𝑡)) of CMCH[𝑎] [𝑏]
𝑃𝑏 (DDV(𝑡)) ← 𝑓(𝐸𝑏 (𝑡)) × 𝑓(DV(𝑡))
if (𝑃𝑏 (DDV(𝑡)) of CMCH[𝑎] [𝑏] is max value) then
nextCH ← CMCH[𝑎] [𝑏]
end
CH󸀠 [𝑎] ← nextCH
end
for 𝑎 = 0 to 𝐺
for 𝑏 = 0 to 𝑁 − 1
if (Region including CH󸀠 [𝑎] ≡ Region including 𝑁[𝑏]) then
DirCH󸀠 [𝑎]𝑁[𝑏] ← |DirCH󸀠 [𝑎] − Dir𝑁[𝑏] |
]CH󸀠 [𝑎]𝑁[𝑏] ← |]CH󸀠 [𝑎] − ]𝑁[𝑏] |
if (0 ≤ (DirCH󸀠 [𝑎] /|Dir𝑁[𝑏] rs |) ≤ 1) and (0 ≤ (]CH󸀠 [𝑎] /|Δ]|) ≤ 1) then
CM󸀠𝑎 ∋ 𝑁[𝑏]
end
end
output cluster head node group CH󸀠
output cluster member node group of all cluster head node 𝑖 CM󸀠𝑖
end

Pseudocode 2: The pseudocode of the exchange of the cluster head node.

the packet delivery ratio that is higher than that in the LIC
and MOBIC algorithms.
4.3. Transmission Range. Furthermore, we simulated the
packet delivery ratio by transmission range. In general, when
transmission range increases, the cluster head node can join
more cluster member nodes. Thus, we set transmission range
from 400 m to 600 m. The network area was increased from
1000 × 1000 m2 , 1500 × 1500 m2 and 2000 × 2000 m2 , the
number of clusters was set at 16, and the velocity was 1∼16.
The simulation results are shown in Figures 15–17.
As shown in Figure 15, when the transmission range is
increased, the proposed algorithm shows the packet delivery
ratio that has increased from 81% to 96% when the network
area is 1000 × 1000 m2 . Figure 16 illustrates that the DDV-
hop algorithm shows the packet delivery ratio that has
increased from 70% to 92% when the network area is 1500
× 1500 m2 . In Figure 17, the DDV-hop algorithm shows the
packet delivery ratio that has increased from 54% to 81%
when the network area is 2000 × 2000 m2 . According to the
increase of transmission range, the LIC algorithm maintains
the packet delivery ratio that is 2% to 20% without a reference
to the network area. Furthermore, the MOBIC algorithm
is shown in the packet delivery ratio that is 25% to 35%
when the network area is 1000 × 1000 m2 . However, the
network area is increased and the packet delivery ratio of
the MOBIC algorithm is decreased by 5%. These simulation
results suggest that, according to the increased transmission
range, the proposed algorithm offers a more continuous
connection as compared to other algorithms.
4.4. Velocity of Nodes. Furthermore, we simulated the packet
delivery ratio by velocity of a node. The max velocity of a node
was set at 8 m/s to 24 m/s. The network area was increased
from 1000 × 1000 m2 , 1500 × 1500 m2 and 2000 × 2000 m2 ,
12 International Journal of Distributed Sensor Networks

1.0 1.0

0.8
0.8
Packet delivery ratio (PDR)

Packet delivery ratio (PDR)

0.6
0.6
0.4
0.4
0.2

0.2
0.0

0.0
0 16 20 24 28 32 0 16 20 24 28 32
Number of clusters (EA) Number of clusters (EA)

DDV algorithm MOBIC algorithm DDV algorithm MOBIC algorithm

LIC algorithm LIC algorithm

Figure 9: The packet delivery ratio by the number of clusters and Figure 11: The packet delivery ratio by the number of clusters and
1000 × 1000 m2 . 2000 × 2000 m2 .

1.0 1.0

0.8 0.8
Packet delivery ratio (PDR)
Packet delivery ratio (PDR)

0.6
0.6

0.4
0.4
0.2
0.2
0.0

0.0
0

1000 × 1000

1200 × 1200

1500 × 1500

1800 × 1800

2000 × 2000
0 16 20 24 28 32
Number of clusters (EA)

DDV algorithm MOBIC algorithm

LIC algorithm Network area (m2 )

Figure 10: The packet delivery ratio by the number of clusters and DDV algorithm MOBIC algorithm
1500 × 1500 m2 . LIC algorithm

the number of clusters was set at 16, and transmission was
450 m. Simulation results are shown in Figures 18–20.
As shown in Figure 18, due to the increase the maximum
velocity of a node, the cluster was broken frequently. So the
packet delivery ratio of compared algorithm is lower than
the proposed algorithm. The proposed algorithm maintains
the packet delivery ratio higher than compared algorithm,
because this algorithm performs the cluster by the velocity
and the direction. In Figure 19, the DDV-hop algorithm
shows the packet delivery ratio that has decreased from 79%
to 74% when the network area is 1500 × 1500 m2 . In Figure 20,
the DDV-hop algorithm shows the packet delivery ratio that
has decreased from 65% to 59% when the network area is
2000 × 2000 m2 . The packet delivery ratio of the MOBIC
algorithm has remained stable (from 30% to 33%) when the
Figure 12: The packet delivery ratio by the network area and number
of clusters (16).
network area is 1000 × 1000 m2 . The network area is increased
and the packet delivery ratio of the MOBIC algorithm is
decreased by 7%. The LIC algorithm maintains the packet
delivery ratio of about 3% to 29% without a reference to the
network area. These simulation results suggest that, according
to the increased velocity and network area, the proposed
algorithm offers a more continuous connection as compared
to other algorithms.
5. Conclusions
MANET consists of nodes that have mobility properties such
as direction and velocity. Due to the characteristics of a node,
International Journal of Distributed Sensor Networks 13

1.0 1.0

0.8 0.8
Packet delivery ratio (PDR)

Packet delivery ratio (PDR)

0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0
0 400 450 500 550 600
0

1000 × 1000

1200 × 1200

1500 × 1500

1800 × 1800

2000 × 2000
Transmit range (m)

DDV algorithm MOBIC algorithm

Network area (m ) 2 LIC algorithm

DDV algorithm MOBIC algorithm Figure 15: The packet delivery ratio by transmission range and 1000
LIC algorithm × 1000 m2 .

Figure 13: The packet delivery ratio by the network area and the
number of clusters (24). 1.0

1.0 0.8
Packet delivery ratio (PDR)

0.8 0.6
Packet delivery ratio (PDR)

0.6 0.4

0.4 0.2

0.2 0.0
0 400 450 500 550 600
Transmit range (m)
0.0
0

1000 × 1000

1200 × 1200

1500 × 1500

1800 × 1800

2000 × 2000

DDV algorithm MOBIC algorithm

LIC algorithm

Figure 16: The packet delivery ratio by transmission range and 1500
Network area (m ) 2 × 1500 m2 .

DDV algorithm MOBIC algorithm

LIC algorithm
similar velocity and direction. Afterwards, to maintain cluster
Figure 14: The packet delivery ratio by the network area and the formation, the cluster head node periodically exchanges. For
number of clusters (32).
the exchange of the cluster head node, a cluster member
node calculates the exchange probability using the distance
between the cluster head node and the cluster member node,
MANET frequently exchanges topology. Due to the exchange direction, and velocity. Based on the calculated probability of
topology, MANET happens to experience overhead and cluster members, the cluster head node chooses the cluster
energy consumption is inefficient. To solve these problems, member node with the maximum probability and updates
this paper proposed the DDV-hop algorithm. The proposed cluster information. To prove the efficient connection of
algorithm is a cluster algorithm that uses direction and the proposed algorithm, we compared random cluster algo-
velocity as the components of cluster formation. rithms against the proposed algorithm. The simulation results
We proposed an algorithm for performing cluster forma- suggest that the DDV-hop algorithm is higher than related
tion where the base station establishes standards of direction cluster algorithms with regard to the packet delivery ratio.
and velocity. The cluster head node is determined by these Thus, we can conclude that the proposed DDV-hop algorithm
standards. The cluster head node joins cluster members with succeeded in its goal of maintaining communication with a
14 International Journal of Distributed Sensor Networks

1.0 1.0

0.8 0.8
Packet delivery ratio (PDR)

Packet delivery ratio (PDR)

0.6
0.6
0.4
0.4
0.2

0.2
0.0

0.0
0 400 450 500 550 600 0 8 12 16 20 24
Transmit range (m) Velocity of node (m/s)

DDV algorithm MOBIC algorithm DDV algorithm MOBIC algorithm

LIC algorithm LIC algorithm

Figure 17: The packet delivery ratio of transmission range and 2000 Figure 19: The packet delivery ratio max velocity and 1500 ×
× 2000 m2 . 1500 m2 .

1.0 1.0

0.8 0.8
Packet delivery ratio (PDR)
Packet delivery ratio (PDR)

0.6
0.6

0.4
0.4
0.2
0.2
0.0

0.0
0 8 12 16 20 24 0 8 12 16 20 24
Velocity of node (m/s) Velocity of node (m/s)

MOBIC algorithm DDV algorithm MOBIC algorithm

DDV algorithm
LIC algorithm LIC algorithm

Figure 18: The packet delivery ratio by max velocity and 1000 × Figure 20: The packet delivery ratio by max velocity and 2000 ×
1000 m2 . 2000 m2 .

lower communication overhead and overall shows a better
performance than existing protocols.
Conflict of Interests
The authors declare that there is no conflict of interests
regarding the publication of this paper.
Acknowledgments
This research was supported by the Ministry of Science, ICT
and Future Planning (MSIP) (2014H1C1A1066391), Republic
of Korea, and partially supported by the Education and
Research Promotion Program of KUT.
References
[1] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci, “A
survey on sensor networks,” IEEE Communications Magazine,
vol. 40, no. 8, pp. 102–105, 2002.
[2] G. Pottie, “Wireless sensor networks,” in Proceedings of the
Information Theory Workshop (ITW ’98), San Diego, Calif, USA,
1998.
[3] T.-H. Kim, D. Tipper, and P. Krishnamurthy, “Localized algo-
rithm to improve connectivity and topological resilience of
multi-hop wireless networks,” Journal of Information and Com-
munication Convergence Engineering, vol. 11, no. 2, pp. 69–81,
2013.
[4] P. Mitra and C. Poellabauer, “Efficient group communications in
location aware mobile ad-hoc networks,” Pervasive and Mobile
Computing, vol. 8, no. 2, pp. 229–248, 2012.
International Journal of Distributed Sensor Networks 15

[5] N. Fareena, A. S. P. Mala, and K. Ramar, “Mobility based energy International Conference on System Sciences (HICSS ’00), IEEE,
efficient multicast protocol for MANET,” in International Con- Maui, Hawaii, USA, January 2000.
ference on Modelling Optimization and Computing, pp. 2473– [22] K.-T. Feng, C.-H. Hsu, and T.-E. Lu, “Velocity-assisted predic-
2483, April 2012. tive mobility and location-aware routing protocols for mobile
[6] M. Ni, Z. Zhong, and D. Zhao, “MPBC: a mobility prediction- Ad Hoc networks,” IEEE Transactions on Vehicular Technology,
based clustering scheme for ad hoc networks,” IEEE Transac- vol. 57, no. 1, pp. 448–464, 2008.
tions on Vehicular Technology, vol. 60, no. 9, pp. 4549–4559, 2011. [23] S. Ganesh and R. Amutha, “Efficient and secure routing
[7] S. Soro and W. B. Heinzelman, “Cluster head election tech- protocol for wireless sensor networks through SNR based
niques for coverage preservation in wireless sensor networks,” dynamic clustering mechanisms,” Journal of Communications
Ad Hoc Networks, vol. 7, no. 5, pp. 955–972, 2009. and Networks, vol. 15, no. 4, pp. 422–429, 2013.
[8] A. S. M. Hosen, S.-H. Kim, and G.-H. Cho, “An energy
efficient cluster formation and maintenance scheme for wireless
sensor networks,” Journal of Information and Communication
Convergence Engineering, vol. 10, no. 3, pp. 276–283, 2012.
[9] Y. Xia and C. K. Yeo, “Measuring group mobility: a topology
based approach,” IEEE Wireless Communications Letters, vol. 2,
no. 1, pp. 54–57, 2013.
[10] W. C.-W. Tan, S. K. Bose, and T.-H. Cheng, “Power and mobility
aware routing in wireless ad hoc networks,” IET Commu-
nications, vol. 6, no. 11, pp. 1425–1437, 2012.
[11] Y. Chen and K. Lee, “A study on the context-awareness architec-
ture for energy conserving routing in MANET,” in Proceedings
of the International Conference of Korea Institute of Maritime and
Communication Science (ICKIMICS ’07), vol. 1, Ulaanbaatar,
Mongolia, July 2007.
[12] Y. Zhang, J. M. Ng, and C. P. Low, “A distributed group mobility
adaptive clustering algorithm for mobile ad hoc networks,”
Computer Communications, vol. 32, no. 1, pp. 189–202, 2009.
[13] H.-C. Jang and H.-T. Huang, “Moving direction based greedy
routing algorithm for VANET,” in Proceedings of the Interna-
tional Computer Symposium (ICS ’10), pp. 535–540, Tainan,
Taiwan, December 2010.
[14] L. Hu, Z. Ding, and H. Shi, “An improved GPSR routing strategy
in VANET,” in Proceedings of the 8th International Conference on
Wireless Communications, Networking and Mobile Computing
(WiCOM ’12), pp. 1–4, IEEE, Shanghai, China, September 2012.
[15] C.-C. Chiang, H.-K. Wu, W. Liu, and M. Gerla, “Routing in
clustered multihop, mobile wireless networks with fading chan-
nel,” in Proceedings of the Singapore International Conference on
Networks (SICON ’97), Singapore, April 1997.
[16] P. Basu, N. Khan, and T. D. C. Little, “A mobility based metric
for clustering in mobile ad hoc networks,” in Proceedings of
the International Workshop on Wireless Networks and Mobile
Computing (WNMC ’01), Phoenix, Ariz, USA, April 2001.
[17] P. Santi, Topology Control in Wireless Ad Hoc And Sensor
Network, John Wiley & Sons, London, UK, 2005.
[18] M. Baysan, K. Sarac, R. Chandrasekaran, and S. Bereg, “A
polynomial time solution to minimum forwarding set problem
in wireless networks under unit disk coverage model,” IEEE
Transactions on Parallel and Distributed Systems, vol. 20, no. 7,
pp. 913–924, 2009.
[19] T. Wang and W. N. N. Hung, “Reliable node clustering for
mobile Ad hoc networks,” Journal of Applied Mathematics, vol.
2013, Article ID 285967, 8 pages, 2013.
[20] W. B. Heinzelman, A. P. Chandrakasan, and H. Balakrish-
nan, “An application-specific protocol architecture for wireless
microsensor networks,” IEEE Transactions on Wireless Commu-
nications, vol. 1, no. 4, pp. 660–670, 2002.
[21] W. R. Heinzelman, A. Chandrakasan, and H. Balakrishnan,
“Energy-efficient communication protocol for wireless micro-
sensor networks,” in Proceedings of the 33rd Annual Hawaii