You are on page 1of 9

Available online at www.sciencedirect.

com
Available online at www.sciencedirect.com
ScienceDirect
ScienceDirect
Procedia
Available Computer
online Science 00 (2018) 000–000
at www.sciencedirect.com
Procedia Computer Science 00 (2018) 000–000 www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia
ScienceDirect
Procedia Computer Science 155 (2019) 351–359

The 14th International Conference on Future Networks and Communications (FNC)


The 14th International Conference on 2019,
August 19-21, FutureHalifax,
Networks and Communications (FNC)
Canada
August 19-21, 2019, Halifax, Canada
An Analysis of the Directional Preference ETX Measure for the
An Analysis of the Directional Preference ETX Measure for the
Collection Tree Protocol in Mobile Sensor Networks
Collection Tree Protocol in Mobile Sensor Networks
David Krynickiaa, Ramiro Liscanobb*
David Krynicki , Ramiro Liscano *
a
Ontario Power Generation, 889 Brock Rd. Pickering ON, Canada
b
Ontario
a Tech
Ontario University,
Power 2000889
Generation, Simcoe St.Rd.
Brock North, Oshawa,
Pickering ON,ON, Canada
Canada
b
Ontario Tech University, 2000 Simcoe St. North, Oshawa, ON, Canada

Abstract
Abstract
There has been a growing interest in Wireless Sensor Networks (WSN) that utilizes mobile nodes for various purposes. These
There
mobilehas wireless
been asensor
growing
networks
interesttendin to
Wireless
suffer from
Sensor
constant
Networks
link breakages
(WSN) that mainly
utilizes
caused
mobile
by connected
nodes fornodes various
moving
purposes.
apart,These
often
mobile
movingwireless sensorThese
very quickly. networks tend to sufferrequire
lost connections from constant
WSNs tolink breakages
constantly mainly
repair caused by
the network connected nodes
connections; movingmaintenance
this constant apart, often
moving
in turn very
causesquickly.
powerThese lost connections
and packet losses andrequire
very WSNs to constantly
noisy network repair the
conditions. networkaconnections;
However directional this constantmetric
preference maintenance
can be
in turn causesinpower
implemented andreduce
order to packet thelosses and very
frequency noisy network
and occurrence conditions.
of lost However
links between a directional
a parent node and its preference
child for metric
mobilecan be
sensor
implemented
networks. As insuchorder to reduce the
a directional ETX frequency
measureandwasoccurrence
developedoffor lostthelinks betweenTree
Collection a parent node(CTP)
Protocol and itsinchild
orderfortomobile
create sensor
longer
networks. As This
lasting links. such paper
a directional
describes ETXandmeasure
measureswasthedeveloped
performance forofthethis
Collection Tree
Directional Protocol CTP
Preference (CTP)(DP-CTP)
in order routing
to createprotocol
longer
lasting
utilizinglinks. Thismetrics
various paper describes and measures
such as Packet the Ratio
Reception performance
(PRR), of this Directional
average number ofPreference CTP (DP-CTP)
beacon transmissions per routing
node and protocol
parent
utilizing
changes. various metricsbased
A comparison such onas PRR
Packet Reception
to other Ratio
popular WSN (PRR), average
algorithms suchnumber
as Leachof beacon transmissions
and Geographic Greedy perForwarding
node and (GGF)
parent
changes. A comparison
is presented. Based on thebased
PRR onthePRR to otheralgorithm
DP-CTP popular WSN algorithms
improves such as Leach
the performance and such
of CTP Geographic
that it isGreedy
capableForwarding (GGF)
of outperforming
is presented.
Leach and GGFBased on the PRR
in various the DP-CTP algorithm improves the performance of CTP such that it is capable of outperforming
scenarios
Leach and GGF in various scenarios
© 2019 The Authors. Published by Elsevier B.V.
© 2019 The Authors. Published by Elsevier B.V.
This
© 2019is an
Theopen accessPublished
Authors. article under the CCB.V.
by Elsevier BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
This
Peer-review
is an open
Peer-review under
access
under responsibility
article under
responsibility ofofthe
the
theConference
Conference
CC BY-NC-ND Program
Program license
Chairs.
Chairs.(http://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review under responsibility of the Conference Program Chairs.
Keywords: Collection Tree Protocol; CTP; Mobile Wireless Sensor Networks
Keywords: Collection Tree Protocol; CTP; Mobile Wireless Sensor Networks

* Corresponding author. Tel.: +1-905-721-8668;


E-mail address:author.
* Corresponding rliscano@ieee.org
Tel.: +1-905-721-8668;
E-mail address: rliscano@ieee.org
1877-0509 © 2019 The Authors. Published by Elsevier B.V.
This is an open
1877-0509 access
© 2019 Thearticle under
Authors. the CC by
Published BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Elsevier B.V.
Peer-review
This under
is an open responsibility
access of the CC
article under Conference
BY-NC-NDProgram Chairs.
license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review under responsibility of the Conference Program Chairs.

1877-0509 © 2019 The Authors. Published by Elsevier B.V.


This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review under responsibility of the Conference Program Chairs.
10.1016/j.procs.2019.08.050
352 David Krynicki et al. / Procedia Computer Science 155 (2019) 351–359
2 Author name / Procedia Computer Science 00 (2018) 000–000

1. Introduction

Mobility introduces an increase to overhead to routing protocols within WSNs due to the fact that nodes that are in
communication may leave each other’s radio range causing the protocol to have to find a new parent. This causes the
network’s topology to be fluid which in turn forces the routing protocol to rebuild the routing tree. In order for a
routing protocol to cope with such fluid changes in topology it must quickly react to lost links, repair only the portions
that are broken, be able to identify loops and choose a parent that would be unlikely to be lost.
There are three common types of network topologies that are utilized in wireless sensor networks, these being Tree,
Clustering, Chain, and Flooding topologies. In this paper the focus is on a tree topology.
Tree topology WSNs utilize protocols that are tree based which are designed to create connections that branch from
the sink node towards the rest of the sensor nodes which create in essence a network map that would resemble a tree.
One of the more popular tree based protocols for WSNs is the Collection Tree Protocol (CTP). CTP is an address free
routing protocol where sensor nodes are not responsible for selecting a full path to the sink, but only towards a parent
which in turn selects its own parent; this is repeated until the sink is selected as a parent [1]. In other words the sink is
chosen implicitly by choosing a neighbour to send data to [2]. Parents are chosen based on beacons that are broadcast
which are used to assist in generating paths between nodes in order to create a path to a sink in network [3].
In an effort to improve the performance of CTP in mobile situations modifications to CTP are necessary. Since
mobile nodes tend to lose connections to their parent and neighbour nodes mostly due to leaving radio range and less
so to causes commonly found in static WSN, it is evident that an improvement to the tree construction or parent
selection will provide better results.
In this paper we present a Directional Preference ETX measure for the Collection Tree Protocol (DP-CTP)
algorithm in order to better support mobile sensing environments. Details of DP-CTP were previously presented [4]
that included a comparison of DP-CTP to conventional CTP. In this paper a more comprehensive analysis of DP-CTP
is presented along with some comparative analysis to the LEACH [5] clustering algorithm and a Greedy Geographic
Forwarding (GGF) [6] algorithm.

2. Related Work

In this section we review some work related to collection tree protocols for mobile sensing as well as works related
to directional measures for routing in wireless sensor networks.
For mobile wireless sensor networks a significant amount of work has occurred in hierarchical routing algorithms
[7] such as in cluster-based network topologies but not as much in the flat structures such as tree-based network
topologies and in particular CTP.
For tree-based routing protocols in mobile sensor networks, most existing work is on the planning of moving
trajectory for nodes or sinks [8, 9] or prediction of mobile nodes to improve routing efficiency [10, 11] (Kusy et al.
2009; Lee et al. 2010). Our work is closer to the works that predict the direction of the mobile nodes except that we
integrate this measure into the calculation of the ETX node edge quality value.
The work by Ben Otman et al. [12] investigated a modification to CTP in mobile environments to attempt to
improve its Packet Reception Ratio (PRR) by incorporating support nodes that are static. In their work they discovered
that the trickle algorithm [13] used by CTP to increase the beacon interval time was not effective in reducing the
number of beacons due to the effect of the mobile nodes on the stability of the network and was disabled for the mobile
nodes. In our work we observed that our algorithm performed better when the trickle algorithm is enabled due to the
fact that network topology is more stable because of the addition of the directional measure into the ETX calculation.
Singh et al. [14] presented a tree-based routing algorithm based on spatial positioning of the nodes focusing on the
definition of levels based on geographic distance of the nodes to the sink. Our approach takes into account relative
position and direction of travel of the nodes and not just distance to the sink as theirs did.
The work by Kumar et al. [15] is closer to our work in that it leverages relative node mobility information in the
formation of clusters. It is similar to LEACH in that it selects cluster heads which service local nodes and aggregate
data before forwarding it towards the sink but the cluster selection is based on a mobility metric rather than signal
strength. This inspired us to compare our approach to an implementation of LEACH development for the Castalia
simulator [16] operating in a mobile node environment. It is worth noting that this implementation of LEACH was
David Krynicki et al. / Procedia Computer Science 155 (2019) 351–359 353
Author name / Procedia Computer Science 00 (2018) 000–000 3

not customized for mobile scenarios (i.e. it did not take into account the spatial relationship among the nodes to create
the clusters.)

3. Directional Preference Collection Tree Protocol (DP-CTP)

As previously stated in the Introduction section the details of the DP-CTP algorithm have been published in
reference [4] and in this section we present a brief overview of CTP and DP-CTP.
CTP creates a tree from sensor nodes to the sink based on cumulative ETX values such that eventually the path
with the minimum cumulative ETX value from the sensor nodes to the sink are chosen. ETX values that are used in
the creation of the neighbour table are calculated by a Link Estimator. A neighbour cannot be selected as a parent until
the 1-Hop ETX value is calculated. The calculation for the 1-Hop ETX value is based on the incoming and outgoing
link quality. The link quality is calculated based on the number of successfully transmitted data packets sent to a
node’s parent over a period of time.
The main purpose of DP-CTP is the selection of a parent node that is traveling in the same direction as the child is
to reduce the frequency these connections have to be changed. When a connection is stable the tree requires fewer
instances of rebuilding which means it can save processing time and power, which in turn would allow the network
to forward more data instead of focusing on rebuilding failed connections. If the nodes know their current direction,
they can selectively choose a different parent utilizing a simple modification to the ETX. Specifically this modification
is the value of ETX with the addition the calculation of the difference in the nodes directional vectors.
This can be accomplished by each node maintaining the knowledge of their current direction and broadcasting it
with their beaconing; to limit the amount of data required all nodes would assume the same domain, such as 0° being
due east, 90° being north, 180° West, 270° South. Utilizing a simple calculation a node can calculate its own theta
based on its current location and its intended location. The formula used is Eqn. 1.

∆𝑦𝑦
𝜃𝜃 = 𝑡𝑡𝑡𝑡𝑡𝑡−1 ( ) ∗ 180/𝜋𝜋 (1)
∆𝑥𝑥

Eqn. 1 provides a θ between 0° and 90° and the vectors are then used to extrapolate the correct θ within a Cartesian
plane. When the other node receives this value it will compute its own and will calculate the difference. If the
difference is greater than 180 the difference is subtracted from 360° to account for directions that are headed in a
similar direction such as 1° North of East and 359° which would actually be 1° South of East.
The idea is that if nodes are heading in the same direction the nodes will stay in range for longer periods of time
and thus will require to reselect a new parent less often. Once the difference of θ is calculated, it can be simply summed
with the current ETX with the θ, as shown in Eqn. 2, which provides enough adjustment to select a more appropriate
node.

𝐸𝐸𝐸𝐸𝐸𝐸𝐷𝐷𝐷𝐷−𝐶𝐶𝐶𝐶𝐶𝐶 = 𝐸𝐸𝐸𝐸𝐸𝐸 + 𝜃𝜃 (2)

4. Analysis of DP-CTP

In this section a comprehensive analysis of DP-CTP is presented based on a modified CTP implementation running
on the Castalia WSN simulator [17].

4.1. Simulation parameters

Three simulation field areas were defined that are primarily differentiated by the physical size of the simulation.
These are small (50m x 50m), medium (100m x 100m), and large (150m x 150m) size. The transmission power of the
nodes is -3 dBm which is approximately equivalent to a 37.5 m range based on the conventional radio model presented
in [12]. The data transmission rate for all the sensor nodes is 0.333 packets/s and the beaconing interval can vary from
10s to 95s in all 3 scenarios (this is the beaconing interval managed by the CTP trickle algorithm). The number of
nodes is a parameter that varies between the simulations and is the values are listed in Table 1. The density ratio of
354 David Krynicki et al. / Procedia Computer Science 155 (2019) 351–359
4 Author name / Procedia Computer Science 00 (2018) 000–000

nodes to area is consistent between the different areas and correspond to the density values of .002, .004, .006, .008,
and 0.010 nodes / m2. The location of the sink varies among areas only to maintain a location that is at the top center
of the experimental area.

Table 1. Simulation parameters that vary between the different areas.


Parameter Small Field Medium Field Large Field
Field size 50m x 50m 100m x 100m 150m x 150m
Variation in the number of nodes 5,10,15,20,25 20,40,60,80,100 45,90,135,180,225
Sink location x=25m y=50m x=50m y=100m x=75m y=150m

The mobility model chosen for the experiments is the random way point mobility model with fixed speeds of 1, 2,
3, and 4 m/s as it emulates the most challenging mobility that could be encountered in real life. This is justified by
Lagkas et al. [18] which concluded in their research paper that “In general, the more independent, dynamic, and less
coherent the node movement is, the less paths can be created, hence, less transmissions can be successfully
completed”.

4.2. Packet Reception Ratio Experiments

In all of these experiments the Packet Reception Ratio (PRR) was captured relative to the different node density
values for different node speeds. Fig. 1 show the results of the simulation for the small area. In a small area the PRR
values are relatively good (in 19 of 20 cases above 97%). The variation is PRR is about 1.5% among the different
speeds and density values.
As can be witnessed from the results in Fig. 1 and Fig. 2 the PRR values become worse as the experimental size
increases even though the node density values are consistent across the different sizes. We attribute this to the increase
in the number of nodes within the range of other nodes (calculated at 37.5 m) that increases the amount of beacons
and packets being sent which can account for the increase in collisions. This is reflected in Fig. 3 that shows the
increase in packet loss due to interference as the field size and density increases.

Fig. 1. PRR vs node density for different node speeds for the small field

An additional reason why the large area performed poorly in comparison to the smaller areas is that when a parent
closer to the sink has a deeper tree it is possible that when a node in the middle of the tree disconnects the connection
to the rest of the nodes below the disconnected node will have lost their ability to forward their packets as well, further
lowering the PRR.
There are other factors in packet loss such as nodes that cannot service incoming packets because they are too busy
servicing other requests. This trend has similar behavior to that of Fig. 3 with higher dropped packets when the number
David Krynicki et al. / Procedia Computer Science 155 (2019) 351–359 355
Author name / Procedia Computer Science 00 (2018) 000–000 5

of nodes increases and when the field size is larger. These dropped packets are influenced less by node mobility as
they are by an increase in the number of nodes.

(a) (b)
Fig. 2. PRR vs Node Density for different node speeds; (a) medium field and (b) large field size

Fig. 3. Packet failures due to interference for all field sizes and node speeds for different node density

4.3. Parent Selection and Beacon Broadcasting

A quick look at the Beacon broadcast results for the three field size simulations shown in Fig. 4 and Fig. show that
the nodes have been broadcasting their beacons excessively in the small size field as compared to the larger fields.
This can be attributed to an increase in nodes traveling in completely opposite directions after randomly choosing a
new direction.
The smaller area provides few trajectories to choose from especially since the nodes reach the edges of the field
more often, because of this the ETX values get updated more frequently which means that parents must update their
children with new ETX values. However since there were fewer nodes in the network and the radio range was large
enough to generally keep the parent within range there was in comparison very few parent changes, which can be seen
in Fig. 6.
Only at higher speeds were there additional changes in the parents however this can be attributed to more frequent
directional changes that cause changes to the ETX values sufficient enough to exceed the minimum required to select
a new parent. Furthermore since there are fewer neighbours in the small area network there is a high chance that a
neighbour with a poor ETX could have replaced the worst case neighbour in the neighbour table. This would mean
356 David Krynicki et al. / Procedia Computer Science 155 (2019) 351–359
6 Author name / Procedia Computer Science 00 (2018) 000–000

that the trickle algorithm would have been reset to the minimum interval beacon transmit time more often causing an
increase in beacons being broadcast

Fig. 4. Average number of beacon transmissions vs node density for different node speed in a small field

(a) (b)

Fig. 5. Average number of beacon transmissions vs node density for different node speed in (a) medium field and (b) large field size

Conversely in the medium and large fields, there are many more nodes that can be selected meaning that it would
be harder to replace a node in the neighbour table especially if more are selected traveling in the same direction. Since
nodes in the larger networks are not being removed from the neighbour table the trickle algorithm had the opportunity
to approach their maximum beacons transmit time, thus lowering the amount of beacons broadcast. However due to
the larger areas and limited radio range there were more opportunities for nodes to leave and enter the neighbour
ranges and possibly replace a nodes parent as seen in comparing Fig. 6 and Fig. 7.
Author name / Procedia Computer Science 00 (2018) 000–000 7
David Krynicki et al. / Procedia Computer Science 155 (2019) 351–359 357

Fig. 6. Number of parent changes vs node density for different node speeds in a small field area.

(a) (b)

Fig. 7. Number of parent changes vs node density for different node speeds in (a) medium field and (b) large field area

5. Comparison of DP-CTP to LEACH and GGF

Looking at a comparison between LEACH, DP-CTP and GGF in Fig. 8 for 2 different node speeds it can be seen
that DP-CTP outperforms the other algorithms for most scenarios for node quantities ranging from 40 to 90 in terms
of Packet Reception Ratio. Focusing on LEACH the PRR that it experiences is quite low, which is not usually the
case for LEACH. However LEACH is not a multi-hop protocol as such and is limited to the range of the cluster head
and sink, assuming that the cluster head is 30m away from the sink the furthest this network can reach is 60m from a
sensing node, this means that there are numerous nodes in an area that are unable to send their payloads. This affects
the PRR because it is based on the expected number of packets to be sent versus the total number of unique packets
received by the sink. The variation in the slightly higher speeds and PRR that is received for LEACH can be attributed
to nodes moving into range of a cluster head to deliver their payload.
GGF was run in two different styles, the first where all nodes were mobile except the sink and the second where
all the nodes were static while the sink was mobile. The first method where all nodes were mobile while the sink was
358 David Krynicki et al. / Procedia Computer Science 155 (2019) 351–359
8 Author name / Procedia Computer Science 00 (2018) 000–000

static had poor PRR, this could be attributed to the fact that each node had to constantly update the location of its
neighbours to determine to whom to forward, and because the sink was broadcasting the same location the further
away a node was from the sink the less exact it could determine its own location. When the sink was mobile but all
nodes were static the PRR performance was much improved, again this can be attributed to the fact that the sink could
more rapidly update the nodes in the network with its location and because the nodes are static they were able to better
pin point their own locations and not have to update their neighbor table compared to a volatile scenario.

(a) (b)
Fig. 8. PRR vs number of nodes for DP-CTP. LEACH, and GGF for node speeds of (a) 4 m/s and (b) 5 m/s

6. Conclusion

It was determined through simulation analysis that the performance of DP-CTP degrades as the volume of nodes
in the network increases. For each network field size there is a number of nodes that is ideal, too few nodes and there
is not enough coverage and too many and the interference caused by the additional noise can prevent successful
communication. It was also observed that dropped packets can occur due to buffer overflows in nodes that are closer
to the sin as such either the buffer should be increased, or an aggregation system should be implemented.
A comparison against LEACH was also conducted. LEACH utilizes scheduling to prevent or lower the effects of
collisions and interference, and it was observed that packets sent by the sensor nodes are successfully received by
their cluster heads, however because LEACH does not support multi-hop the application layer reports fewer data
packets received by the sink. There is a trade-off between Tree based algorithms and clustering algorithms such as
LEACH and DP-CTP. For clustering algorithms in order to have complete coverage of an area the radio range has to
be increased which requires additional power usage. Tree based algorithms have a limitation to the number of nodes
they can directly support due to the number of packets that need to be forwarded especially near the sink where nodes
can be overloaded, an increase in buffer can allow for fewer packet drops but would again require additional power.
If however DP-CTP aggregated data then the number of packets sent would be reduced thus possibly reducing the
congestion occurring near the sink, while not having a large impact on power consumption.

References

[1] Hakoura, Bassel Zuhair. (2011) “Comparison of Collection Tree Protocols with Gossip Algorithms for Data Aggregation in Wireless Sensor
Networks.” Doctoral dissertation McGill University Libraries.
[2] Fonseca, Rodrigo, Omprakash Gnawali, Kyle Jamieson, Sukun Kim, Philip Levis, and Alec Woo. (2006) “The collection tree protocol (CTP).”
TinyOS TEP 123(2).
[3] De Couto, Douglas Seraphim James. (2004) “High-throughput routing for multi-hop wireless networks.” Doctoral dissertation Massachusetts
Institute of Technology.
David Krynicki et al. / Procedia Computer Science 155 (2019) 351–359 359
Author name / Procedia Computer Science 00 (2018) 000–000 9

[4] Krynicki, David and Ramiro Liscano. (2017) “Directional Preference Collector Tree Protocol for Mobile Wireless Sensing.” In International
Conference on Network-Based Information Systems: 339-350.
[5] Heinzelman, W. R., A. Chandrakasan, and H. Balakrishnan. (2000) “Energy-efficient communication protocol for wireless microsensor
networks,” in Proceedings of the 33rd Annual Hawaii International Conference on System Sciences: 10-pp.
[6] Karp, Brad and H.T. Kung. (2000)" GPSR: Greedy Perimeter Stateless Routing for Wireless Networks." in Proc. of the 6th Annual ACM/IEEE
International Conference on Mobile Computing and Networking: 243-254.
[7] Sabor Nabil, Shigenobu Sasaki, Mohammed Abo-Zahhad, and Sabah M. Ahmed. (2017) “A Comprehensive Survey on Hierarchical-Based
Routing Protocols for Mobile Wireless Sensor Networks: Review, Taxonomy, and Future Directions”, Wireless Communications and Mobile
Computing, 2017, https://doi.org/10.1155/2017/2818542 (Accessed 10 May 2019).
[8] Tan, Rui, Guoliang Xing, Jianping Wang, and Hing Cheung So. (2010) “Exploiting Reactive Mobility for Collaborative Target Detection in
Wireless Sensor Networks”, IEEE Trans. Mobile Computing, 9 (3): 317-332.
[9] Xing, Guoliang, Tian Wang, Zhihui Xie, and Weijia Jia. (2008) “Rendezvous Planning in Wireless Sensor Networks with Mobile Elements,”
IEEE Trans. Mobile Computing, 7 (12): 430-1443.
[10] Kusy, Branislav, HyungJune Lee, Martin Wicke, Nikola Milosavljevic, and Leonidas Guibas. (2009) “Predictive QoS Routing to Mobile Sinks
in Wireless Sensor Networks,” Proc. ACM Int’l Conf. Information Processing in Sensor Networks (IPSN): 109-120.
[11] Lee, HyungJune, Martin Wicke, Branislav Kusy, Omprakash Gnawali, and Leonidas Guibas. (2010) “Data Stashing: Energy-Efficient
Information Delivery to Mobile Sinks through Trajectory Prediction,” Proc. ACM/IEEE Ninth Int’l Conf. Information Processing in Sensor
Networks (IPSN): 291-302.
[12] Ben Otman, Nadra. (2015) “Design Considerations for Fixed Node Assisted Multi-hop Mobile Sensor Network.” MASc thesis dissertation
University of Ontario Institute of Technology.
[13] Levis, Philip, Thomas Clausen, Jonathan Hui, Omprakash Gnawali, and J. Ko. (2011) “The trickle algorithm.” No. RFC 6206
[14] Singh, Mrityunjay, M. Sethi, N. Lal, and Saroj Poonia (2010) “A tree based routing protocol for mobile sensor networks,” IJCSE International
Journal on Computer Science and Engineering, 02 (01S): 55-60.
[15] Kumar, G. Santosh, Paul MV Vinu, and K. Poulose Jacob. (2008) “Mobility metric based leach-mobile protocol.” In 2008 16th International
conference on advanced computing and communications: 248-253.
[16] Pires, Adonia and Claudio Silva. (2011) “LEACH Clustering Protocol for Castalia Simulator 0.2.
[17] Boulis, Athanassios (2011).“Castalia A simulator for Wireless Sensor Networks and Body Area Networks.” NICTA: National ICT Australia
83
[18] Lagkas, Thomas D., Argyro.Lamproudi, Panagiotis Sarigiannidis, and Charalabos Skianis. (2015) “The impact of mobility patterns on the
efficiency of data forwarding in MANETs.” IEEE Int. Conf. Commun.,2015: 6037-6042.

You might also like