Professional Documents
Culture Documents
ScienceDirect
Karbala International Journal of Modern Science xx (2017) 1e17
http://www.journals.elsevier.com/karbala-international-journal-of-modern-science/
Abstract
It is extremely important to safeguard our harbours from intruders and smugglers who aim to benefit from unlawful activities
and cause harm. Theft, terror attacks in commercial boats and cargo ships docked on harbours needs to be prevented. Camera
surveillance, radar, satellites images have not been very reliable so far as they fail to work in drastic weather conditions and can be
manipulated as well. Underwater Wireless Sensor Network (UWSN) could be installed in both shallow and deep water of the
harbour for detecting various types of harbour activities. Ship movements namely heave, sway, surge, yaw, pitch and roll, could be
detected and classified using pressure, position and underwater sensors. Such information can help in tracking ship motions,
movements, loading and unloading activities. Any unplanned unloading activity can thus be detected and necessary alarms can be
raised for ship owners and harbour official's attention. However designing such a network needs one to ensure that the severe energy
constraints of the UWSN are well addressed. Bayesian Network based approach is explored in this paper for scheduling the sleep
and active cycle of network nodes. Our proposed technique of ship motion monitoring system reduces energy consumption of the
network nodes and enhances the network lifetime by balancing network load intelligently. The tracking mechanism proposed here
explores the fundamental behaviour of ships motions in waves with reference to translating coordinate system.
© 2017 The Authors. Production and hosting by Elsevier B.V. on behalf of University of Kerbala. This is an open access article
under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Keywords: Underwater Wireless Sensor Network (UWSN); Ship motion monitoring; Bayesian network
https://doi.org/10.1016/j.kijoms.2017.11.001
2405-609X/© 2017 The Authors. Production and hosting by Elsevier B.V. on behalf of University of Kerbala. This is an open access article under
the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Please cite this article in press as: M. Rao, N.K. Kamila, Bayesian network based energy efficient ship motion monitoring, Karbala International
Journal of Modern Science (2017), https://doi.org/10.1016/j.kijoms.2017.11.001
+ MODEL
2 M. Rao, N.K. Kamila / Karbala International Journal of Modern Science xx (2017) 1e17
consumption of nodes in transmission is the major unloading activity of the ship (theft) and raises alarm.
concern of such a delay tolerant network, so that the It is also designed to raise an alarm when a ship docked
network life time is prolonged. Yuanming [2] has in the harbour is manoeuvred away from the dock
proposed energy balanced routing algorithm for wire- (robbery) without prior information. It ensures that the
less sensor network that avoids routing loop and routing information collected by the sensor nodes is
redundancy in path selection in order to save energy. forwarded to the base station in an energy efficient
But this approach cannot be applicable in the case of manner while avoiding load imbalance and achieving
underwater wireless sensor network, due to dynamic an enhanced network lifetime.
topology due and water current. Li et al. [3] have
proposed a compression algorithm for wireless sensor 2. Problem statement
network deployed on sea surface for sea route moni-
toring. Their algorithm is based on computation of Fig. 1 depicts the satellite impression of ship traffic
distortion based on L1-norm. Their results prove that along harbours on oceans as per the information
data transmission and network survivability is available at www.maritime.com [13]. Red arrows
improved. However, their system is not helpful in indicate tanker ship positions while green arrows
detecting intrusions and thefts. The concern of indicate the cargo ship positions as received via sat-
detecting ship intrusions has been dealt by Luo et al. ellite. However, this information is not enough for
[4]. Their technique helps in detecting ships trespass- detecting and evading intruders. It will certainly help
ing harbour areas by classifying ship generated waves identifying various ships and their movements. When
and ocean waves. Underwater wireless sensor networks assisted with radar technology, more accurate infor-
have some inherent constraints due to the acoustic mation of objects surrounding a ship on ocean can be
medium of data transmission. Kohli and Bhattacharya known. However, it cannot help in intrusion detection.
[5] have pointed out the implication of depth, tem- Chainas [14] attempts to find alternative ships for
perature and salinity of ocean water in acoustic mode serving ports in situation when a desired ship fails to
of data propagation. Li et al. [6] have proposed a delay reach its course, probably due to malfunction. Ports are
tolerant MAC layer protocol inspired by coupon col- informed of the arrival and departure of ships, but
lector problem to address the issues of long propaga- monitoring such huge entities needs more than camera
tion delay and swarm mobility of underwater sensors. and human surveillance. Underwater Wireless Sensor
Similarly there are other issues related to protocol and Network can be installed in harbours for detecting
network design of underwater sensors. Coles et al. [7] various motions of ships. Currently there is no reliable
argue that instead of deploying a huge number of static technology to detect when a ship is being robbed and
nodes for surveillance, a set of mobile nodes could be therefore cannot be prevented. Neither is there a
deployed guided by decentralized discrete Bayesian mechanism to know when the cargo of a ship is being
network implementing an optimized grazing strategy. stolen. These crimes are reported only later by human
Many researchers have aimed to reduce energy con- surveillance or by the help of satellites images. By
sumption of nodes (Huang et al. [8]; Latif et al. [9]; using sensors we can enable the harbour officials with
Yang et al. [10]; Shazzad et al. [11]). Techniques like (at the moment) information for efficient decision
depth based routing, sleep cycles; efficient MAC pro- making in combating with such crimes.
tocols with collision avoidance have been explored. Here, we have attempted to develop a system that
Mandal [12] has attempted to study TDMA variants detects various motions of ships to help ship owners
that could be suitable for underwater sensor network. and harbour officials to track ship activities in order to
But none of the authors talk about energy efficient ship avoid theft and robbery. Robbery can be differentiated
motion detection. from theft as a situation where an intruder exercises
The work here, significantly contributes towards coercion. Ship robbery can be very dangerous than ship
building an autonomous ship motion detection system theft. However, both these situations need to be cur-
that is equipped with strategies for identifying different tailed. Here we consider taking away of the ship
types of ship motions, namely heave, sway, surge, yaw, permanently from the harbour by an intruder as rob-
pitch and roll that vary due to different angles of force bery. Our goal is prevent such situations with the help
and gravity. Using underwater wireless sensor network of underwater wireless sensor networks in a robust and
the proposed technique predicts a possible unplanned scalable way.
Please cite this article in press as: M. Rao, N.K. Kamila, Bayesian network based energy efficient ship motion monitoring, Karbala International
Journal of Modern Science (2017), https://doi.org/10.1016/j.kijoms.2017.11.001
+ MODEL
M. Rao, N.K. Kamila / Karbala International Journal of Modern Science xx (2017) 1e17 3
Underwater wireless sensor network have unique 3.2. Behaviour of ships in waves
challenges and constraints, unlike seen in other
distributed mobile network. Gkikopouli et al. [1], A ship motion can be said to have six degrees of
Zenia et al. [15], Shazzad et al. [11], Diamant and freedom as depicted in Fig. 2. Its motion can be
Lampe [16], Jafri et al. [17], Han and Fei [18] and
Mandal [12] have all tried to address these special
concerns of underwater wireless sensor networks. They
are listed below for better understanding:
Please cite this article in press as: M. Rao, N.K. Kamila, Bayesian network based energy efficient ship motion monitoring, Karbala International
Journal of Modern Science (2017), https://doi.org/10.1016/j.kijoms.2017.11.001
+ MODEL
4 M. Rao, N.K. Kamila / Karbala International Journal of Modern Science xx (2017) 1e17
classified into three mutually perpendicular trans- considered for an average position of ship's centre of
lations and three rotations around its centre of Gravity gravity. This threshold would vary on various climatic
(G). These six degrees of freedom of ship motion conditions. Eq. (7) expresses how the threshold value is
namely: heave, sway, surge, yaw, pitch and roll can be considered.
mathematically expressed in steadily translating coor- z ¼ za cosðue tÞ ð7Þ
dinate system O (x, y, z) [20,21]. Let us, imagine a
person on board of a ship. What he sees on the front is These various motions of a ship can be measured
known as bow side (front end) and stern side is on his using position sensors. The sensor data can help in
back, to his right is the starboard side and on his left is classifying the various types of motions. When a theft
the port side. All ship motions are described and define of intrusion happens it is more likely that only the three
as follows [20]. motions related to rotation will occur. When a ship is
The motion surge can be defined as a translation loaded or unloaded, one would also find the three
motion along the longitudinal x-direction as expressed rotation type of motions to occur. Here, we consider
in (1). the problem of detecting theft or intrusion in a ship by
Here x is the coordinate along which the motion is relating to the ship rotation based motions. The next
explained, u is the circular wave frequency and z is the section explains the proposed system.
harmonic elevation of the water surface. xa is the wave
amplitude along the x coordinate and time t. 4. Proposed system
x ¼ xa cosðue t þ εxz Þ ð1Þ
A ship docked at a harbour needs to be monitored
Similarly, swaying is also a translation type of continuously. Camera surveillance may not be enough.
motion in the lateral direction of the y axis and with We propose to design a classifier to identify various
respect to the port side. It is expressed in Eq. (2), where ship motions. Using the data collected by the position
ya is the wave amplitude along y axis, u is the circular and pressure sensors, we can extract features such as
wave frequency and z is the harmonic elevation of the weight of the ship, angle of displacement and further
water surface. perform classifications to learn if a ship is being rob-
bed, or being stolen (theft). Necessary alarms can then
y ¼ ya cos ue t þ εyz ð2Þ
be raised to combat the intrusion. Fig. 3 illustrates the
Heaving on the contrary is a translation type of proposed system. In our earlier work (Rao et al. [22]),
motion in the vertical z-direction and upwards. It is we have demonstrated with the help of simulations,
defined mathematically as in Eq. (3) how a network of nodes can be deployed in the harbour
for detecting various objects based on spatial and
z ¼ za cosðue t þ εzz Þ ð3Þ
temporal correlation with the help of an Aerial Mobile
Roll, pitch and yaw are rotation motions. Roll is Sink Node (AMS). In this work, we extend the
motion in right side of the positive x axis explained in detection technique further and enable it to learn from
Eq. (4). Pitch similarly is the positive right turning of its experiences. This can be achieved by modelling
the boat in the y axis and expressed in (5). Yaw is the with Bayesian belief network because of its efficiency
positive right turning as seen in the z axis, explained in and simplicity. It not only handles issues due to
Eq. (6). missing data, but also achieves incremental learning.
The proposed system comprises of two sections,
namely it has a detection system and a communication
4 ¼ 4a cosðue t þ ε4z Þ ð4Þ
system. The detection system comprises of pressure
q ¼ qa cosðue t þ εqz Þ ð5Þ nodes and position sensor nodes as shown in Fig. 3.
The communication system includes the nodes that
J ¼ Ja cosðue t þ εJz Þ ð6Þ transmit the data to the aerial mobile sink with the
algorithm [22]. It comprises of underwater sensors,
These motions are of and around the centre of
which report to their cluster heads selected after every t
gravity (G) of a ship and therefore any motion in any
time duration based on residual energy. Here we pro-
point of the ship can be detected by the help of su-
pose an algorithm that aims to run on the aerial mobile
perposition. A threshold of wave elevation could be
Please cite this article in press as: M. Rao, N.K. Kamila, Bayesian network based energy efficient ship motion monitoring, Karbala International
Journal of Modern Science (2017), https://doi.org/10.1016/j.kijoms.2017.11.001
+ MODEL
M. Rao, N.K. Kamila / Karbala International Journal of Modern Science xx (2017) 1e17 5
Fig. 3. Architecture of cognitive underwater wireless sensor network for ship motion detection.
sink node to route information to the base station. This routing, path reinforcement learning, path reliability
enables the base station to learn on any unplanned and clustering have been explored, but efficient
unloading activities. It perceives it as a theft or robbery tracking techniques in Underwater sensor network is
and raises alarms It also helps the base station to still to be discovered.
predict nodes in the network that are about to drain
their energy using spatial and temporal correlation. 4.1. Proposed algorithm
This shall enable nodes to look for alternate routing
paths. Singh and Kaki [23] have proposed an adaptive We propose a ship motion recognition algorithm
MAC protocol that schedules the sleepeactive cycle of here. The algorithm enables the base station to track
nodes by using reinforcement learning techniques. any unplanned activities. The ship motion recognition
However this will require each node to compute more algorithm continuously tracks the changes in weight,
than required. Coutinho et al. [24] have proposed an pressure and position of the ship. When any change
opportunistic routing technique that could extend the beyond a threshold is reached, necessary alarms are
network lifetime by controlling the sleep interval of the raised. However, designing such a system has to
nodes. Geographical routing techniques have been address the challenge imposed by the limited power of
proposed for energy conserving of nodes in Refs. the sensor nodes. The network is likely to experience
[25,26], while [27,28] suggest duty cycles of nodes for load imbalance leading to network disconnectivity and
conserving energy. resulting in system fault. Hence, an energy efficient
Researchers so far have only addressed the need for routing technique is needed for such a system. The ship
optimised routing techniques; but none have addressed motion recognition and classification algorithm pro-
the need for optimized tracking techniques. Many posed here takes advantage of Bayesian belief network
routing techniques [27e35] based on agent based for dealing with the problem of routing load imbalance
Please cite this article in press as: M. Rao, N.K. Kamila, Bayesian network based energy efficient ship motion monitoring, Karbala International
Journal of Modern Science (2017), https://doi.org/10.1016/j.kijoms.2017.11.001
+ MODEL
6 M. Rao, N.K. Kamila / Karbala International Journal of Modern Science xx (2017) 1e17
P (A) P (B)
P(a1) P(a2) P(b1) P(b2) X
0.5 0.5 0.5 0.5
P(x|a,b)
P(x1| P(x2| P(x3| P(x4| P(x5| P(x6|
ai bj) ai bj) ai bj) ai bj) ai b j ) ai bj)
a1,b1 0.2 0.1 0.1 0.1 0.2 0.3 C D
a1,b2 0.3 0.2 0.1 0.1 0.1 0.2
a2,b1 0.1 0.2 0.3 0.3 0.1 0
a2,b2 0.1 0.2 0.3 0.3 0.1 0
P(c|x) P(d|x) E
P(c1|xi) P(c2|xi) P(c3|xi) P(d1|xi) P(d2|xi) P(d3|xi)
x1 0.6 0.2 0.2 x1 0.5 0.4 0.1
x2 0.2 0.4 0.4 x2 0.2 0.7 0.1 P(e|c,d)
x3 0.4 0.3 0.3 x3 0.3 0.4 0.3 c1d1 c1d2 c1d3 c2d1 c2d2 c2d3 c3d1 c3d2 c3d3
x4 0.5 0.3 0.2 x4 0.3 0.5 0.2 e1 0.7 0.8 0.5 0.6 0.9 0.4 0.2 0.3 0.1
x5 0.4 0.5 0.1 x5 0.1 0.4 0.5 e2 0.3 0.2 0.5 0.4 0.1 0.6 0.8 0.7 0.9
x6 0.2 0.5 0.3 x6 0.3 0.6 0.1
while tracking any possibilities of ship being robbed B. Node A represents the time of the day and therefore
(unplanned unloading: alarm 1) or taken out of port can have two values, such A ¼ {a1, a2}, where, a1 as
(alarm 2). Bayesian belief network represents casual forenoon and a2 as afternoon. Node A is significant in
dependencies of the structural information when representing the impact of density, frequency and time
reasoning about the functioning of a system. Bielza of ocean as it translates in the behaviour of ocean
et al. [36] express how multidimensional classification waves; this directly influences the type of ship motion.
can be achieved. We shall represent the ship motion Node B here represents the type of ship design. We
classification as a problem of multidimensional clas- consider this as very important feature for classifying
sification with the help of discrete variables in belief ship motions. Journee and Pinkster [21] have pointed
nets. out, how a ship's length to its beam, affects when it
Fig. 4 represents the belief network for ship motion needs to turn. Other features such as location of lon-
classification. The nodes labelled in upper case in gitudinal centre of buoyancy, block and prismatic co-
Fig. 4 represent the discrete states associated in this efficient also affect the pattern of ship motion. In this
classification problem. Nodes A and B have a direct paper, we consider only two types of ships design,
influence on the output of ship motion classification namely: cargo and tanker ship. A cargo ship is essen-
represented by node X. The child nodes of node X, i.e, tially designed to carry goods such as coal, grains, iron
nodes C and D have indirect influence of nodes A and ores and similar commodities; where as a tanker ship
Please cite this article in press as: M. Rao, N.K. Kamila, Bayesian network based energy efficient ship motion monitoring, Karbala International
Journal of Modern Science (2017), https://doi.org/10.1016/j.kijoms.2017.11.001
+ MODEL
M. Rao, N.K. Kamila / Karbala International Journal of Modern Science xx (2017) 1e17 7
mostly carries oil. Node B is expressed as B ¼ {b1, topology as illustrated in Fig. 4. We can evaluate this
b2}, Were, b1 ¼ cargo; b2 ¼ tanker. Node X, which by summing up the joint distribution, P (a, b, c, d, e),
represents the type of ship motion can have 6 values, as over all the variables other than e, as e needs to be
explained in previous section and expressed in Eq. (8). estimated. It is given by Eq. (11).
X
X ¼ fx1 ; x2 ; x3 ; x4 ; x5 ; x6 g ð8Þ PðeÞ ¼ Pða; b; c; d; eÞ ð11Þ
a;b;c;d
here, x1 ¼ heave, x2 ¼ sway, x3 ¼ surge, x4 ¼ yaw,
x5 ¼ pitch, x6 ¼ roll. At time instant t, nodes which have a weight larger
Node C and node D, represent the data collected by than the average weight wavg of the network nodes are
various sensors namely pressure sensors and position selected to be in transmit state at t þ 1 instant of time
sensors. We represent node C as follows and consider and the other nodes are allowed to be in a sleep state.
generalized terms for ease of understanding. wi of a node is computed using Eq. (12) as explained in
Ref. [37].
C ¼ fc1 ; c2 ; c3 g ð9Þ
Gi
where, c1 ¼ light, c2 ¼ average, c3 ¼ heavy. wi ¼ w1 Er þ w2 ð12Þ
Di
Node D is expressed as in Eq. (10) and it has an
indirect influence of node A and B. It is the real time here wi is the weight of a node i and it decides whether a
data collected by position sensors about the angle of node would be sleep or transmit mode. w1 and w2 are
ship bending from its centre of gravity (G). These weight factors that can be estimated by Eq. (11). Er is
values are likely to be recorded in three axes, x, y the residual energy of the ith node, while Gi represents
and z. detection range and Di represents the signal strength.
The residual energy Er of an ith node at time instant t, is
D ¼ fd1 ; d2 ; d3 g ð10Þ equal to the difference of the initial energy of the node
and the energy consumed G [34] by the node so far.
where, d1 ¼ change of angle in x axes; d2 ¼ change of Energy consumption of an UWSN node is computed
angle in y axes; d3 ¼ change of angle in z axes. using Eq. (13) [34].
Node E represents the energy state of the nodes. An
Dlupp
underwater wireless sensor node is likely to be in t0 þ Ts
transmit or sleep state, which can be represented as X X e
E ¼ {e1, e2}, where e1 ¼ transmit and e2 ¼ sleep state. G¼ EðiÞ; where EðiÞ ¼ P0 U RCðiÞðtÞ
vi ~t¼t0
A node forwarding a packet towards sink is in e1 state
otherwise it is in sleep state, which is denoted by e2. If ð13Þ
the probability of the state of node is 1, then we can
infer e2 as e2 ¼ 1 e1. Using Bayes theorem one can In Eq. (13), E(i) is the energy consumed by the ith node
calculate the posterior probability. Fig. 4 is an illus- in a Ts time duration, Dlupp is the upper bound of for-
tration of a simple belief net explaining the working warding delay, P0 is the minimum power required for
e
principle of the proposed algorithm with an instant of guaranteed reception of a packet and UðRCðiÞðtÞ Þ is the
possible data samples. The data base is synthesized for component that represents the underwater signal
the proposed model. When the model is functioning, attenuation [38]. Our proposed algorithm using
the values will be collected and updated by the sensor Bayesian inference can prolong the network lifetime of
nodes in every t instant of time. The time of the day has nodes as it can avoid nodes from draining their energy
a direct impact on the behaviour of waves and there- inefficiently using Eq. (12). The proposed algorithm has
fore directly affects the type of ship motion. The type following objectives:
of motion indeed affects the nodes to report changes in
pressure and position. Using the property of Bayesian 1. Identify type of ship motion and classify type of
network, we can determine the probability distribution threat to raise alarms
over the variables e1 and e2 at node E in the network 2. Node scheduling to extend network lifetime
Please cite this article in press as: M. Rao, N.K. Kamila, Bayesian network based energy efficient ship motion monitoring, Karbala International
Journal of Modern Science (2017), https://doi.org/10.1016/j.kijoms.2017.11.001
+ MODEL
8 M. Rao, N.K. Kamila / Karbala International Journal of Modern Science xx (2017) 1e17
Please cite this article in press as: M. Rao, N.K. Kamila, Bayesian network based energy efficient ship motion monitoring, Karbala International
Journal of Modern Science (2017), https://doi.org/10.1016/j.kijoms.2017.11.001
+ MODEL
M. Rao, N.K. Kamila / Karbala International Journal of Modern Science xx (2017) 1e17 9
Please cite this article in press as: M. Rao, N.K. Kamila, Bayesian network based energy efficient ship motion monitoring, Karbala International
Journal of Modern Science (2017), https://doi.org/10.1016/j.kijoms.2017.11.001
+ MODEL
10 M. Rao, N.K. Kamila / Karbala International Journal of Modern Science xx (2017) 1e17
Please cite this article in press as: M. Rao, N.K. Kamila, Bayesian network based energy efficient ship motion monitoring, Karbala International
Journal of Modern Science (2017), https://doi.org/10.1016/j.kijoms.2017.11.001
+ MODEL
M. Rao, N.K. Kamila / Karbala International Journal of Modern Science xx (2017) 1e17 11
Fig. 5. Comparison of energy consumption of nodes with DSDV protocol v/s DSDV with proposed algorithm.
Fig. 6. Energy consumption of nodes with traditional tracking technique using AODV protocol.
Please cite this article in press as: M. Rao, N.K. Kamila, Bayesian network based energy efficient ship motion monitoring, Karbala International
Journal of Modern Science (2017), https://doi.org/10.1016/j.kijoms.2017.11.001
+ MODEL
12 M. Rao, N.K. Kamila / Karbala International Journal of Modern Science xx (2017) 1e17
Fig. 7. Energy Consumption of nodes with traditional AODV & DSDV protocol v/s with proposed algorithm.
outlines the parameters considered for carrying out the algorithm which schedules the transmit and sleep cy-
simulation. cles of nodes in a manner that avoids load imbalance.
Fig. 5, plots the comparison of the energy drainage of
5.2. Performance evaluation nodes as simulation time increases, with traditional
DSDV protocol v/s DSDV protocol with our proposed
The proposed tracking algorithm was tested with tracking algorithm. Energy of nodes in represented
two types of routing protocols namely, Ad Hoc on along the y axis, while the simulation time is repre-
Demand Distance Vector (AODV) and Destination sented in x axis. The results obtained show that DSDV
Sequence Distance Vector (DSDV) routing protocol. In protocol with our tracking algorithm is more efficient
order to evaluate its reliability and stability, following in saving energy of the network with an aerial mobile
performance metrics were considered: a) Energy con- sink node. DSDV is a proactive protocol and therefore
sumption of nodes b) Delay c) Jitter d) Throughput d) some amount of energy is spent initially in periodic
Packet Delivery Fraction (PDF). Delay is defined as advertisements and hop by hop route learning mecha-
the time taken by a packet to reach from source to nism. Our approach extends the network lifetime by
destination. Jitter on the other hand is defined as the 30%, in case of DSDV routing protocol. Fig. 6 depicts
variation in delay of the arrival of packets at the the comparison of energy consumption of network
receiver node. Throughput can be defined as the nodes using
number of packets that successfully reach the desired AODV is essentially a reactive protocol and route
destination. Packet Delivery Fraction (PDF) is the ratio for routing is discovered on demand. It may be noted
of number of packets lost by the total number of that compared to DSDV, AODV protocol fails in
packets generated. conserving energy of the network to an extent that
DSDV does. This is mainly due to the routing overhead
5.2.1. Impact of energy consumption of nodes that AODV protocol creates for route requests. How-
It is very important for WSN applications to ever, when AODV protocol is employed with our
conserve energy of the nodes for prolonging the proposed tracking algorithm, it yields a slightly better
network lifetime. To achieve this we have proposed an approach in conserving energy of the network nodes,
Please cite this article in press as: M. Rao, N.K. Kamila, Bayesian network based energy efficient ship motion monitoring, Karbala International
Journal of Modern Science (2017), https://doi.org/10.1016/j.kijoms.2017.11.001
+ MODEL
M. Rao, N.K. Kamila / Karbala International Journal of Modern Science xx (2017) 1e17 13
Fig. 8. Comparison of DSDV protocol with and without proposed algorithm in terms of delay, jitter and throughput.
Please cite this article in press as: M. Rao, N.K. Kamila, Bayesian network based energy efficient ship motion monitoring, Karbala International
Journal of Modern Science (2017), https://doi.org/10.1016/j.kijoms.2017.11.001
+ MODEL
14 M. Rao, N.K. Kamila / Karbala International Journal of Modern Science xx (2017) 1e17
Fig. 9. Comparison of AODV protocol with and without proposed algorithm in terms of delay, jitter and throughput.
Please cite this article in press as: M. Rao, N.K. Kamila, Bayesian network based energy efficient ship motion monitoring, Karbala International
Journal of Modern Science (2017), https://doi.org/10.1016/j.kijoms.2017.11.001
+ MODEL
M. Rao, N.K. Kamila / Karbala International Journal of Modern Science xx (2017) 1e17 15
Table 2
Summary of simulation results.
Quality of Service parameters DSDV DSDV with AODV AODV with proposed
proposed Algorithm Algorithm
Packets generated 227 456 235 412
Packets dropped 20 2 26 5
Packet Delivery Fraction 8.810573% 0.438596% 11.063830% 1.213592%
Minimum delay (seconds) 0.014485542 0.007724276 0.013845924 0.013826
Average delay (seconds) 0.24690241 0.43063643 0.24456037 0.22594407
Maximum delay (seconds) 0.79444606 23.112526 1.3555875 2.4200597
Minimum packet jitter (seconds) 1.9823E-005 0 1.5964E-005 2.97E-007
Average packet jitter (seconds) 0.015232387 0.29697927 0.098027964 0.063073965
Maximum packet jitter (seconds) 0.77996052 23.0977735 1.118258 2.3810584
Minimum throughput of transferred packets (KB/s) 14 0 0 29
Average throughput of transferred packets (KB/s) 62 28 23 75
Maximum throughput of transferred packets (KB/s) 93 96 69 95
Minimum throughput of generated packets (KB/s) 40 77 0 0
Average throughput of generated packets (KB/s) 78 95 25 38
Maximum throughput of generated packets (KB/s) 98 113 87 99
First packet transferred (seconds) 1.794461 1.0077243 1.0653573 1.098604
Last packet transferred (seconds) 5.1358639 24.120403 13.76673 8.3033117
Time in between (seconds) 3.341417 23.120403 12.76673 15.544089
First packet generated (seconds) 1 1 1 1
Last packet generated (seconds) 4.9636642 7.9907896 13.648371 16.544089
Lifetime (seconds) 3.9636642 6.9907896 12.648271 15.544089
than otherwise and indeed keeps the network con- the number of packets generated in our approach is
nected to some larger extent, as can be seen from about 750 and more, where as DSDV alone was
Fig. 6. Fig. 7, presents a comparison of energy con- generating only 400 packets. More packets are gener-
sumption of network nodes with DSDV and AODV ated due to the overhead caused by maintaining
protocols with DSDV and AODV protocols when scheduling of clock cycles in our approach.
assisted with our proposed algorithm. We find that our The green lines in graph 8(e) and 8(f) indicate
algorithm has lesser energy consumption in the case of acknowledgement packets, while red lines indicate
DSDV protocol. messaging packets and yellow line indicates data
This is essentially due to the fact that DSDV is a packet. It can be noted from 8(f) that the messages are
proactive protocol, while AODV is a reactive protocol exchanges periodically for updating clock cycles.
that needs energy to maintain active links. Our tech- Fig. 9 above, summarizes a comparative study of the
nique however yields better results for both ADOV and AODV protocol with our proposed algorithm. Here
DSDV. 9(a) represents the delay per packet, 9(c) represents
jitter per packet and 9(e) represents throughput of the
5.2.2. The aspect of network performance parameters network with AODV protocol. When combined with
UWSN has limitations imposed by the acoustic our tracking algorithm, we find that though the average
communication where signals are propagated 2,00,000 delay per packet as depicted in 9(b), and average jitter
times slower than WSN signals. depicted in 9(d) have increased but throughput as
Traditional AODV and DSDV need to be modified depicted in 9(f) has improved. As AODV is a reactive
to accommodate acoustic communication. protocol, it is likely to have more delay due to more
In Fig. 8, we represent the comparative study of the number of packets being generated for route requests.
proposed algorithm with DSDV protocol in terms of Table 2 below, summarizes the benefit of the proposed
delay, jitter and throughput. Fig. 8(a) is the represents algorithm. From the table it can be noted that our
the delay per packet by just using DSDV protocol for proposed tracking algorithm has a better Packet De-
tracking, however the delay per packet has reduced livery Fraction (PDF) in both routing protocols. Our
with our algorithm as depicted in 8(b). Jitter per algorithm also has a longer lifetime and better
packets has reduced from what can be seen in 8(c) to throughput. The packet delay and jitter have increased
8(d) with our tracking algorithm. 8(e) and 8(f) repre- in our approach due to the overhead of the updating of
sent the throughput of data packets. It can be seen that sleep cycles. However our algorithm also ensures that
Please cite this article in press as: M. Rao, N.K. Kamila, Bayesian network based energy efficient ship motion monitoring, Karbala International
Journal of Modern Science (2017), https://doi.org/10.1016/j.kijoms.2017.11.001
+ MODEL
16 M. Rao, N.K. Kamila / Karbala International Journal of Modern Science xx (2017) 1e17
Please cite this article in press as: M. Rao, N.K. Kamila, Bayesian network based energy efficient ship motion monitoring, Karbala International
Journal of Modern Science (2017), https://doi.org/10.1016/j.kijoms.2017.11.001
+ MODEL
M. Rao, N.K. Kamila / Karbala International Journal of Modern Science xx (2017) 1e17 17
[19] C. Petrioli, R. Petroccia, J.R. Potter, D. Spaccini, The SUNSET [29] L. Jin, D.D. Huang, A slotted CSMA based reinforcement
framework for simulation, emulation and at-sea testing of un- learning approach for extending the lifetime of underwater
derwater wireless sensor networks, Ad Hoc Netw. 34 (2015) acoustic wireless sensor networks, Comput. Commun. 36
224e238. (2016) 1094e1099.
[20] https://commons.wikimedia.org/wiki/File%3ABrosen_ [30] G. Han, L. Liu, N. Bao, J. Jiang, W. Zhang, J.J.P.C. Rodrigues,
shipsmovemensonthewave.svg. AREP: an asymmetric link-based reverse routing protocol for
[21] J.M.J. Journee, J. Pinkster, Introduction in Ship Hydrome- underwater acoustic sensor networks, J. Netw. Comput. Appl.
chanics, Lecture MT519, Deft University of Technology, Draft (2017), https://doi.org/10.1016/j.jnca.2017.01.009.
Edition, 2002. [31] N. Nowsheen, G. Karmakar, J. Kamruzzaman, PRADD: a path
[22] M. Rao, N.K. Kamila, K.V. Kumar, Underwater wireless sensor reliability-aware data delivery protocol for underwater acoustic
network for tracking ships approaching harbour, in: Paper sensor networks, J. Netw. Comput. Appl. 75 (2016) 385e397.
Presented in IEEE International Conference on Signal Pro- [32] M. Ahmed, M. Salleh, M.I. Channa, Routing protocols based
cessing, Communication, Power & Embedded System, Paral- on node mobility for Underwater Wireless Sensor Network
ekhemundi, Odisha, India, Oct, 2016. (UWSN): a survey, J. Netw. Comput. Appl. 78 (2017)
[23] A. Singh, K.K. Kaki, Relative reward based scheduling (RRS): 242e252.
an adaptive duty cycle based WSN MAC protocol, in: Paper in [33] N. Javaid, S. Hussain, A. Ahmad, M. Imran, A. Khan, Region
Proceedings of International Conference on Advanced based cooperative routing in underwater wireless sensor net-
Communication Control and Computing Technologies works, J. Netw. Comput. Appl. (2017), https://doi.org/10.1016/
(ICACCCT), 2016, pp. 779e782. j.jnca.2017.01.013. Article in Press.
[24] R.W.L. Coutinho, A. Boukerche, L.F.M. Vieira, A.A.F. Loureiro, [34] L. Liu, P. Wang, R. Wang, Propagation control of data for-
Modeling the sleep interval effects in duty-cycled underwater warding in opportunistic underwater sensor networks, J. Com-
sensor networks, in: Paper in Proceedings of IEEE International put. Netw. 114 (2017) 80e94.
Conference on Communications (ICC), 2016, pp. 1e6. [35] L. Liu, N. Zhang, Y. Liu, Topology control models and solu-
[25] C. Zhu, L.T. Yang, L. Shu, V.C.M. Leung, J.J.P.C. Rodrigues, tions for signal irregularity in mobile underwater wireless
L. Wang, sleep scheduling for geographic routing in duty- sensor networks, J. Netw. Comput. Appl. 51 (2015) 68e90.
cycled mobile sensor networks, IEEE Trans. Ind. Electron. 61 [36] C. Bielza, G. Li, P. Larra~naga, .Multi-dimensional classification
(11) (Nov 2014) 6346e6355. with Bayesian networks, Int. J. Approx. Reason. 52 (6) (Sep
[26] J. Jiang, G. Han, H. Guo, L. Shu, J.J.P.C. Rodrigues, 2011) 705e727.
Geographic multipath routing based on geospatial division in [37] B. Wu, Y.-P. Feng, H.-Y. Zheng, X. Chen, Dynamic cluster
duty-cycled underwater wireless sensor networks, J. Netw. member scheduling for target tracking in sensor networks,
Comput. Appl. 59 (2016) 4e13. IEEE Sens. J. 16 (19) (2016) 7242e7249.
[27] M.R. Bharamagoudra, S.S. Kumar, B. Gonen, Event driven [38] L. Berkhovskikh, Y. Lysanov, Fundamentals of Ocean Acous-
energy depth and channel aware routing for underwater acoustic tics, third ed., Springer-Verlag, 2003.
sensor networks: agent oriented clustering based approach, J. [39] A. Papanikolaou, Selection of main dimensions and calculation
Comput. Electr. Eng. 58 (2017) 1e19. of basic ship design values (Chapter 2), in: Springer Book on Ship
[28] X. Ding, X. Sun, C. Huang, X. Wu, Cluster-level based link Design, 2014, https://doi.org/10.1007/978-94-017-8751-2_2.
redundancy with network coding in duty cycled relay wireless [40] C.C. Mei, Waves in Water, Wave Propagation (Chapter 4), MIT,
sensor networks, J. Comput. Netw. 99 (2016) 15e36. 2004, pp. 1e40.
Please cite this article in press as: M. Rao, N.K. Kamila, Bayesian network based energy efficient ship motion monitoring, Karbala International
Journal of Modern Science (2017), https://doi.org/10.1016/j.kijoms.2017.11.001