Energy efficient hidden node detection for improving Quality of Service in wireless multimedia sensor networks

Full text

Energy Efficient Hidden Node Detection for
Improving Quality of Service in Wireless Multimedia
Sensor Networks
Adwan Alanazi and Khaled Elleithy
Computer Science and Engineering Department
University of Bridgeport, CT- USA
aalanazi@my.bridgeport.edu; elleithy@bridgeport.edu
Abstract— In many wireless multimedia sensor networks
(WMSNs) the nodes are static. However, node connection is
subjected to change due to disruptions in wireless communication,
power changes in transmission, or loss of synchronization
between neighboring nodes. A sensor should constantly be aware
of its immediate neighbors, through a process called continuous
neighbor discovery. In this paper we introduce an energy
efficient hidden node detection (EEHND) algorithm for
continuous neighbor discovery process in the wireless multimedia
sensor networks (WMSNs). We focus on the continuous neighbor
discovery process and regard it as a combined task of all the
nodes in every connected segment. Each sensor is entered as a
coordinate in an effort in order to reduce the time to detect
hidden sensors. Based on the simulation results, we demonstrated
that the protocol detected the hidden nodes in the network.
Keywords: wireless multimedia sensor networks, neighbor discovery,
hidden nodes, and energy efficiency.
I. I
NTRODUCTION
The latest advances in the complementary metal-oxide-
semiconductor (CMOS) technology have led to the
development of wireless multimedia sensor networks as a
class of wireless sensor network. WMSNs are capable of
capturing multimedia content such as images, audio and video
about the surrounding environment. Then the WMSNs send
that to the sink or a base station. WMSNs have offered
numerous applications in surveillance, environmental
applications military applications, health applications, and
home appellations. However, multimedia applications have
limitations that will affect the successful media transmission in
the sensor networks. Node connection is subjected to change
due to disruptions in wireless communication, power changes
in transmission, or loss of synchronization between
neighboring nodes [1-4]. When a sensor is aware of its
immediate neighbors, it must continuously maintain
information on its surroundings. In this work, we differentiate
between neighbor discovery during sensor network
initialization and continuous neighbor discovery. We focus on
the latter and regard it as a combined task of all the nodes in
every connected segment. Each sensor initiates the neighbor
discovery process to reduce power consumption by not
increasing the time required to detect hidden sensors. Despite
the static nature of the sensors in many sensor networks,
connectivity is still subject to change even after the network
has been established. The sensors must continuously try to
identify new neighbors in order to fulfill these requirements: 1)
Local synchronization loss because of accumulated clock
drifts [5]; 2) Wireless connectivity disruption between
adjacent nodes by a temporary event, such as a passing vehicle,
moving animals, or storms. When these events are over, the
hidden nodes must be revived for rediscovery; 3) the ongoing
accumulation of new nodes, in some networks to compensate
for nodes which have malfunctioned; 4) and the increase in
transmission power of some nodes, in response to several
events, such as detection of emergency situations [6-8]. Based
on these circumstances, detecting new links and nodes in
sensor networks must be considered an ongoing process [7]. In
the following discussion, we distinguish between the detection
of new links and nodes during initialization. When the node is
in an active state that performs normal recovery process. The
normal recovery process consists of initial neighbor discovery
and then continuous discovery process for finding the location
of the node. The former is referred to as initial neighbor
discovery whereas the latter is referred to as continuous
neighbor discovery. While previous works address initial
neighbor discovery and continuous neighbor discovery as
similar tasks, to be performed by the same scheme, we claim
that different schemes are required, based the following
reasons:
Initial neighbor discovery is usually conducted when the
sensor has no evidence about the structure of its immediate
surroundings. In this situation, the sensor cannot communicate
with the gateway and is therefore very limited in performing
its tasks. The immediate surroundings should be detected as
soon as possible in order to set a path to the gateway and
contribute to the operation of the network [9]. Hence, in this
state, more extensive energy use is justified. In contrast,
continuous neighbor discovery is performed when the sensor
is already functioning [8]. This is a long-term process,
whereby optimization is vital for extending the network
lifetime. When the sensor initiates the continuous neighbor
discovery, it shows that a node is already responsive from
many of its immediate neighbors, and can, therefore,
accomplish the task with these neighbors to consume
minimum energy [10, 11]. In contrast, each sensor in a distinct
fashion must accomplish an initial neighbor discovery.
The remainder of the paper is organized as follows: Section II
discusses the problem identification and its significance.
Section III gives an overview of existing approaches. Section
IV presents the proposed approach. Section V discusses the

simulation and experimental results. Finally the entire paper is
concluded in Section VI.
II. PROBLEM
IDENTIFICATION
Initial neighbor discovery is usually performed when
the sensor has no awareness of its immediate surroundings. In
such a situation, the sensor node cannot communicate with its
base station thus, limiting the node performance. Based on this
situation, several applications experience the following
problems:
• The network experiences heavy traffic; as a result the
performance is decreased.
• Extended process for forwarding the data; it takes
longer process for forwarding the data if node is not
found then it causes the latency in the network.
• Additional energy consumption. When the neighbor
node information is not available then the source
node starts sending the data so that all the nodes
receive the data, consume additional energy.
III. RELATED
WORK
In this section, some of important features of related
work are discussed. In [5], the neighbor discovery process is
introduced and attempts to determine the new node from the
station. Since, energy consumption is not a concern for the
base station, discovering new nodes is simplified. The base
station initiates the node discovery process periodically
broadcasting the special HELLO message. When a node gets
the message, it initiates the registration process. The node can
switch channels to find the best HELLO message. The best
Hello message might depend on the distinctiveness of the
broadcasting base station, on security considerations, or on the
physical layer quality. Another related work attempted to
reduce neighbor discovery time process by improving the
broadcast rate of the HELLO messages [10, 11]. In both works,
the central node performs the neighbor discovery process.
Similarly, the work did not focus on the energy consumption.
In addition, the hidden nodes were assumed to be able to listen
to the ‘HELLO’ messages broadcasted by the central node.
Every node conducted the neighbor discovery in sensor
networks, and hidden nodes could not be able to listen to the
HELLO messages when sleeping. In [9], an algorithm for
energy efficient node discovery is introduced. The approach
uses the temporal patterns of happenstances and exploits the
patterns to determine the duty cycling. Duty cycling is
perceived as a sampling process and articulated as an
optimization problem. Authors also used fortification-learning
methods to perceive and dynamically change the times at
which a sensor node should be aware, as it is likely to meet
other nodes. In [6], a simple Aloha-like algorithm is proposed
that supports synchronous node transmissions and the number
of known nodes. The approach shows that the time for all the
nodes to determine their corresponding neighbors is Θ (ln n) in
a perfect network, which led to an arbitrary number of nodes
to communicate simultaneously. The fundamental problem of
secure neighbor discovery is introduced in [12] and attempted
to secure the network from different forms of attacks. The
approach consists of a scalable key-distribution protocol that is
secure in the absence of conspiring malicious nodes. The
objectives were to present the secure neighbor discovery to
protect the attacks of hidden nodes. The static network was
proposed to secure the one-hop neighbor discovery process.
However, the work did not focus on energy efficiency. All
existing approaches tried to detect the hidden node, but did not
factor in energy efficiency. Our proposed protocol particularly
focuses on energy efficiency while detecting the hidden node
in the neighborhood.
IV.
DETECTING
HIDDEN
NODES
We focus on the continuous neighbor discovery process. In
our approach, each sensor node applies a simple approach in a
coordinate effort to minimize energy consumption without
increasing the required time to detect hidden nodes. In addition,
we focus on avoiding the message collision.
Our approach consists of the following components:
A. Hidden link participate inside a subdivision
Our approach is to determine the hidden links when a
new node joins the network. Subdivision nodes should detect
the joining of new node in the network in order to continue the
communication process. When new node joins that node
releases a particular synchronization (SYN) message to all
subdivision members, waking them up and periodically
broadcasting a group of HELLO messages. This SYN message
is dispersed over the already identified links of the subdivision.
Thus, it is assured to be received in every subdivision node by
having all the nodes wake up practically at the same time. For
a short period, we can guarantee that every wireless link
between the segment's members is detected.
B. Hidden link participate outside a subdivision
A random wake-up methodology is applied to reduce the
option of iterating collisions between the HELLO messages of
nodes in the same subdivision. In our approach, each node
coordinates with its other neighborhood nodes during the
wake-up periods to avert collisions and accelerate the
discovery process of hidden nodes. Since the wake up time
period is very small, and the time of forwarding the HELLO
transmission time is even smaller. In this case, there is
possibility that two nodes can be active at same time and
initiate neighbor node recovery process. In our approach, we
applied scheduling process so that there is marginal chance
that two nodes should be active same time and initiate
neighbor discovery process. During the scheduling process,
the nodes need to be synchronized with each other and follow
the schedule for the rest of communication time. Each
receiving node chooses the some timeslots and receives the
data during those timeslots. The time slot process is done
without conflicting the schedule of other node. Therefore, we
split the neighbor nodes into different subdivision groups,

where node of each subdivision chooses its slot assigned to
that group only.
C. Neighbor Discovery Model
The sensor node decides randomly, when to initiate the
transmission of a HELLO message. If its message does not
strike with another HELLO message, the node is referred to as
a discovered node. We also, are able to determine the load and
residual energy of each node after the node discovery process.
The load factor ′

′ depends on the buffer state of the node
that is estimated by using the equation (1).


= 



(1)
Where 

: The number of the Hello packets in the buffer and


: Maximum size of the buffer to accept the Hello
message during the node discovery process. When the node
discovery process is completed, we can determine the residual
energy of the node. Thus residual energy ′

′ of each node can
be calculated as


=

−(

+

)(2)
Where 

: Initial energy of node, 

: Receiver’s energy
consumed for hidden node discovery process and 

:
transmitter’s energy consumed for hidden node discovery
process.
The objective is to determine the HELLO transmission
frequency, and the duration of the neighbor discovery process.
Algorithm 1: An Energy Efficient Hidden Node Detection
Process
1) Initialization of sensor nodes in a network. (Ns :
sensor nodes)
2) Beginning of the simulation time (To).
3) Consider a hidden node in a network i.e Nh
4) If (Nh ∈ Ns), then
5) Nh broadcasts RREQ (root request) message
forwarding to neighbor nodes (Nn)
6) Else if ( dest = = true), then
7) detects neighbor nodes and connection found
8) end if
9) detection is not found and unreachable
10) end if
11) consider state of nodes ( Sn: normal state; Si initial
state) and probability “P1”
12) if (Sn<<Si && Ns ∈ Si), then
13) transmission of “ HELLO” message takes place
14) else if (1-(1-P1)D>=P), then
15) the function of DT(Discovery time) i.e TN(v )is
calculated
16) end if
17) neighbor node (Nn) responds to “HELLO” message
through invoking
18) finalize the setup of joint wireless link
19) end if
In algorithm, line 1 shows the initialization process. Line-2 sets
the simulation time. Lines3-4 shows the number of nodes that
belongs to the sensor network. From Lines5-10, neighbor
nodes broadcast the message at 1-hop neighbor nodes. When
the message is delivered and hidden nodes receive response,
the availability of hidden nodes is presented. From lines 11-18
the energy consumption is illustrated based on the probability
of hidden nodes and the discovery process. Finally, hidden
nodes are allowed to join the networks. We take a simple
example for the proposed algorithm. Let us assume that nodes
in the initial state should remain an active until they enter into
normal state. Let us consider a subdivision node in the normal
state, where continuous neighbor discovery node is
accomplished that determines the degree of its hidden
neighbors. In our approach, we assume that distance of hidden
node is 10 meters and number of hidden nodes are 2. Thus, the
hidden node wakes up after specified time and attempts to
broadcast the message at the 1-hop neighborhood. This kind of
process could help determine the availability of another hidden
nodes in the network.
V. SIMULATION
SETUP
AND
PERFORMANCE
ANALYSIS
In order to evaluate the performance of our proposed energy
efficient approach for the hidden node in the wireless
multimedia sensor network, our approach was implemented
using network simulator NS2. The network is designed to
cover 800 X 800 square meters. We distributed 90 nodes in
the network with homogenous capabilities. Each node has an
initial 6 joules of energy. The simulation’s objective is to
determine the consumed energy for hidden nodes to improve
the QoS. Furthermore, we also compared our approach with
two other known existing approaches: Hidden-Node
Avoidance Mechanism (H-NAMe) [13], and Hidden node
avoidance for IEEE 802.15.4 (HNA-IEE) [14]. We
demonstrated two scenarios, with and without hidden nodes.
The simulation scenarios consist of 15 end nodes, which create
flat topology. We set medium access layer to operate with the
Non-Beacon enabled mode that applies un-slotted carrier
sense multiple accesses with collision avoidance (CSMA/CA).
Radio range of each sensor nodes under free-space
propagation model is fixed to 45 meters. In first scenario, all
end nodes that contribute in the network can listen to each
other. The distance between each sensor node is set at 35
meters. In the second scenario, each sensor node can hear ten
of the fifteen other end sensor nodes. This gives 50 %
probability to determine the hidden nodes without wasting
energy. However, remaining 50% nodes need to detect in the
network. This scenario shows the near worst-case because and
each collision could lead to a whole packet damage. End
sensor node generates and transmits the acknowledged frame
whose destination is head node (HN). Acknowledgement

frame uses a constant application payload of 248 bits that
involve 88 bits of the packet overhead, 48 network layer
overhead, 64 bits medium access control overhead and 48
physical layer overhead. We generated the packets using inter-
arrival time based on Rayleigh probability density function. In
the first simulation experiment, the amount of received data
traffic is referred as goodput that is observed as a utility of the
generated data traffic for both scenarios (with and without
hidden nodes). The network traffic shows an amount of data
per time unit for non-acknowledgement frame that is either
received or generated by the neighbor node. The load can be
determined as


=

(3)
Where 

: Network load, : Number of nodes that generate
the data, : Number of generated packets in unit time and 

:
length of the frame that is equivalent to 248 bits.
The rest of parameters are explained in Table 1.
Table 1: Simulation parameters and its corresponding values
PARAMTERS VALUE
Size of network 800 × 800 square meters
Number of nodes 90
End nodes 15
Queue-Capacity 45 Packets
Number of frames 340frames
Maximum number of
retransmissions allowed
03
Initial energy of node 6 joules
free-space propagation 45 meters
Size of Packets 256 bytes
Data Rate 260 kilobytes/second
Sensing Range of node 35 meters
Simulation time 18 minutes
Average Simulation Run 08
Frame rate 38 fps
Reliability [0.78, 0.92]
Reporting rate 3 packet/s
Base station location (0,700)
Transmitter Power 12.7 mW
Receiver Power 13.6 mW
Speed 2-18 m/sec
Based on simulation, we are interested in the following
metrics.
 Successful delivered Packets
 End-to End delay
 Throughput with stationary nodes
A. Successful delivered packets
One of the penalties of the hidden node in WSNs is to drop the
packets because of the collisions that destroy the contents’
quality. Hence, the performance is highly degraded. We
determine the ratio of successful packets as follows:


=



 100(4)
Where 

: Ratio of successful packets, 

: delivered packets
and 

: Number of generated packets.
Figure 1 shows the results for the ratio of the successful
packet delivery with and without hidden nodes. Without
hidden nodes, the successful delivery is almost 96.9% with
180 kb/Sec generated application traffic. Without hidden
nodes, the network reaches its channel capacity; the numbers
of generated packets get greater than the number of packets
that can be transmitted. The packets that cannot be transmitted
are lost. In the existence of the hidden nodes, packets are lost
due to hidden node crashes and as traffic increases that lead to
number of collisions. As a result, great number packets are lost.
Figure 1: Successful delivery packets VS generated application traffic
B. End-to End delay
Packet delivery time shows the time that elapsed form its
construction to the time when the packet is positively received
by the destination node. Thus, this time is called End-to-End
delay. Throughput performance of network depends on the
end-to-end delay. If end-to-end delay is longer than usual time
then, throughput performance is degraded. In Figure 2, we
show the end-to- end delay of the network with and without
hidden nodes. We observed that when the size of the
application increases then end-to-end delay of the network is
highly degraded in presence of the hidden nodes. This leads to
the additional energy consumption and degradation in QoS.
Furthermore, we show in Figure 3, the performance of
network in presence of our proposed approach including the
hidden and non-hidden nodes. The results demonstrate that our
approach has substantially improved the performance of the
network. The statistical data shows that our approach has
minimum end-to-end delay up to 64 kb/sec. Once, the size of
generated application increases then our approach has slightly
higher end-to-end delay up to 139 kb/sec.

Figure 2: End-to-end delay VS generated application traffic
Furthermore, the trend of the results show that our approach
has relatively similar end-to-end delay as obtained without
hidden nodes. Based on the results, we conclude that our
approach substantially decreased the latency and increased the
energy efficiency that could lead improving the QoS in the
network.
Figure 3: End-to-end delay of our proposed EEHND, with hidden nodes and
without hidden nodes
C. Throughput with stationary nodes
The performance of EEHND is evaluated using 100% duty
cycles with constant frame size of 256 data frame ( including
payload and data frame format). We conducated several runs
to determine the network performance at the different loads.
Figures 4 and 5 show the throughput and success probability
rate based on the experiments. The result demonstrates the
performance of our proposed approach and other known
approaches: H-NAMe) and (HNA-IEE).
The average value for throughput and probability success rate
are calculated using 95% confidence interval for the sample
size of 5000 packets at each given load. Based on the results,
we observed that our proposed algorithm outperforms to other
competing algorithms even at lower loads. For example, at a
given offered load of 40%, the success probability rate of our
algorithm is greater than 36.8% than other competing
algorithms. During the experiment scenario, two packets were
retransmitted because of collision. If we deliberate one
retransmission for each lost packet, the increase in the number
of retransmission could be substantial in the case of the
network without our proposed algorithm. This situation could
lead to high-energy consumption.
Figure 4. Throughput with different offered load
Considering higher loads, it is clear that our proposed EEHND
increases the throughput up to 87.5% at 84% load. On the
other hand, HNA-IEE increases the throughput up to 49.2% at
the 54% load and H-NAMe increases up to 72.3 at the 70%
load. We observed our algorithm provides 68% probability
success at 108% load. While HNA-IEE and H-NAMe have
55.5% and 56.2% probability success respectively with same
load.
Based on the results, we conclude that our proposed algorithm
produced substantial progress in the network performance in
terms of throughput and probability success rate.

Figure 5. Probability success rate with different offered load
VI. CONCLUSION
In this paper, an energy efficient hidden node detection
algorithm is proposed for improving the Quality of Service of
the wireless multimedia sensor networks. Our algorithm is
based on continuous neighbor discovery process. Furthermore,
we also included formulation to determine the load and
residual energy of the nodes when completing the discovery
process. We argue that continuous neighbor discovery is
essential even if the sensor nodes are mobile. If the nodes in
an associated subdivision work together on this task, hidden
nodes are ensured to be discovered within a certain probability
and a certain time period. Our simulation results demonstrate
that our proposed EEHND algorithm improved the throughput
performance of the network and reduced the end-to-end delay.
Decrease in end-to-end delay can save the additional energy
consumption. In addition, we compared the performance of
our EEHND algorithm with HAN-IEE and NAMe approaches.
The statistical data shows that our algorithm provides higher
successful probability detection rate and throughput as
compared with other approaches. The simulation results
confirm that our proposed algorithm is capable of determining
hidden nodes for WMSNs application. In future, we will
measure the performance of our proposed algorithm in high
scale wireless multimedia sensor networks and will apply
hardware-testing process.
R
EFERENCES
[1] R. Madan and S. Lall, "An energy-optimal algorithm for neighbor
discovery in wireless sensor networks," Mobile Networks and
Applications, vol. 11, pp. 317-326, 2006.
[2] A. Alanazi and K. Elleithy, "Real-Time QoS Routing Protocols in
Wireless Multimedia Sensor Networks: Study and Analysis,"
Sensors, vol. 15, pp. 22209-22233, 2015.
[3] Z. Serhan and W. B. Diab, "Energy-efficient QoS routing in
wireless multimedia sensor networks," in Advanced Information
Networking and Applications (AINA), 2015 IEEE 29th
International Conference on, 2015, pp. 223-230.
[4] K. Malarvizhi, M. Brindha, and M. Kumar, "Evaluation of energy
efficient routing in wireless multimedia sensor networks," in
Electronics and Communication Systems (ICECS), 2015 2nd
International Conference on, 2015, pp. 1387-1391.
[5] D. J. Baker and A. Ephremides, "The architectural organization of
a mobile radio network via a distributed algorithm,"
Communications, IEEE Transactions on, vol. 29, pp. 1694-1701,
1981.
[6] A. Russell, S. Vasudevan, B. Wang, W. Zeng, X. Chen, and W.
Wei, "Neighbor Discovery in Wireless Networks with Multipacket
Reception," 2014.
[7] R. Jurdak, P. Baldi, and C. V. Lopes, "Adaptive low power
listening for wireless sensor networks," Mobile Computing, IEEE
Transactions on, vol. 6, pp. 988-1004, 2007.
[8] P. Dutta and D. Culler, "Practical asynchronous neighbor discovery
and rendezvous for mobile sensing applications," in Proceedings of
the 6th ACM conference on Embedded network sensor systems,
2008, pp. 71-84.
[9] V. Dyo and C. Mascolo, "Efficient node discovery in mobile
wireless sensor networks," in Distributed Computing in Sensor
Systems, ed: Springer, 2008, pp. 478-485.
[10] A. Keshavarzian, E. Uysal-Biyikoglu, F. Herrmann, and A.
Manjeshwar, "Energy-efficient link assessment in wireless sensor
networks," in INFOCOM 2004. Twenty-third AnnualJoint
Conference of the IEEE Computer and Communications Societies,
2004, pp. 1751-1761.
[11] S. A. Borbash, "Design considerations in wireless sensor
networks," University of Maryland, 2004.
[12] S. Bagchi, S. Hariharan, and N. Shroff, "Secure neighbor discovery
in wireless sensor networks," 2007.
[13] A. Koubâa, R. Severino, M. Alves, and E. Tovar, "H-NAMe: A
hidden-node avoidance mechanism for wireless sensor networks,"
in 8th IFAC International Conference on Fieldbuses and Networks
in Industrial and Embedded Systems, 2009, pp. 10-19.
[14] U. Pešović, J. Mohorko, S. Ranđić, and Ž. Čučej, "Hidden node
avoidance mechanism for IEEE 802.15. 4 wireless sensor
networks," Electronic Components and Materials, vol. 43, pp. 14-
21, 2013.
Adwan Alanazi is pursuing his Doctorate in Computer
Science and Engineering at the University of Bridgeport in
Bridgeport, Connecticut, USA. His interests are wireless
sensor networks, network routing, and mobile commutations.
Dr. Khaled Elleithy is the Associate Dean for Graduate
Studies in the School of Engineering at the University of
Bridgeport. He has research interests are in the areas of
network security, mobile communications, and formal
approaches for design and verification. He has published more
than two hundred and fifty research papers in international
journals and conferences in his areas of expertise.
Dr. Elleithy is the co-chair of the International Joint
Conferences on Computer, Information, and Systems Sciences,
and Engineering (CISSE). CISSE is the first
Engineering/Computing and Systems Research E-Conference
in the world to be completely conducted online in real-time via
the Internet and was successfully running for six years. Dr.
Elleithy is the editor or co-editor of 12 books published by
Springer for advances on Innovations and Advanced
Techniques in Systems, Computing Sciences and Software.