Computer Networks 193 (2021) 108116

Contents lists available at ScienceDirect

Computer Networks
journal homepage: www.elsevier.com/locate/comnet

An energy-efficient routing protocol for the Internet of Things networks

based on geographical location and link quality
Ramin Yarinezhad a, Sadoon Azizi b, *
a
Department of Computer Engineering and Information Technology, Amirkabir University of Technology, Tehran, Iran
b
Department of Computer Engineering and IT, University of Kurdistan, Sanandaj, Iran

A R T I C L E I N F O A B S T R A C T

Keywords: In the design of Internet of Things (IoT), sensor nodes are used in collecting data from an environment and
Internet of Things (IoT) sending it to a base station or sink for future processing. These sensor nodes have some resource constraints such
Energy optimization as power supply. Hence, in order to prolong the network lifetime, we need to manage the energy consumption of
Routing protocol
these sensor devices. Thus, in this paper, a tree-based routing protocol is proposed that is efficient in power
Sensor networks
consumption and reduces the end-to-end delay in energy-efficient green-IoT networks with a mobile sink. Two
new different mechanisms are introduced in the proposed protocol for managing the routing in the network. The
first mechanism is an improved version of the geographic routing algorithm, which is more reliable and energy-
balanced. The second mechanism is a tree-based structure that can be created with the minimum control packets
and updated with an efficient procedure. Simulation results indicate that the proposed routing algorithm is
superior to the existing ones in terms of energy consumption, network lifetime, delay, and throughput.

1. Introduction is their limited power supply, which is generally an irreplaceable battery

[12–14]. Since it is always desired to achieve the maximum network
In the recent decade, Internet of Things (IoT) plays an important role lifetime, saving the energy of the sensor nodes is the most important
in a wide variety of applications such as smart homes, smart grids, smart challenge in such networks [11,15,16].
cities, smart factories, etc. [1,2]. The basic idea behind IoT is that In designing a routing protocol for energy-efficient green-IoT net­
different things or objects around us such as mobile devices, sensors, works, the energy consumption of the nodes must not only be efficient
video cameras, mobile phones, actuators, and Radio-Frequency IDenti­ but also balanced among all nodes. If it is not so, the hot-spot or energy
fication (RFID) tags become able to interact with each other and share hole problem [17,18] occurs. The hot-spot problem refers to the nodes
data through the internet to reach common goals [3–5]. nearer to the sink being required to handle a higher traffic load
In some applications of IoT, there is a demand for extensive data compared to the ones farther from the sink due to the multi-hop for­
collection for controlling and monitoring a specific area such as smart warding fashion in these networks, which causes their shorter lifetime.
homes and smart cities [5,6]. On the other hand, Wireless Sensor Net­ An effective way to address this problem is using mobile sinks [19],
works (WSNs) that are a framework for better managing and formu­ which was introduced in [20].
lating the sensor nodes can be very useful for controlling and monitoring Although using mobile sinks has its advantages, it brings new chal­
a specific area [7–9]. Therefore, in these applications, sensor nodes are lenges for routing in these networks. One of the main challenges is the
considered as the fundamental components of the IoT model. propagation of the location of the sink through the nodes in the network
A WSN consists of a set of sensor nodes deployed in a specific area and updating the next-hop relay nodes of each node, due to the
that work together to monitor and collect the environment’s data and continuous change in the sink’s location. So far, different methods have
send it to a base station or sink, which is connected to the internet been proposed to solve this problem. For example, virtual based in­
through the gateways [10,11]. In IoT networks based on WSNs, there are frastructures have been proposed to help maintain the connection be­
several resource constraints on IoT devices such as limitations in radio tween the sensor nodes and the sink [21–25] or the clustering method
range and processing power. However, a key constraint on sensor nodes has been suggested that helps reduce the number of the nodes that need

* Corresponding author.
E-mail addresses: [email protected] (R. Yarinezhad), [email protected] (S. Azizi).

https://doi.org/10.1016/j.comnet.2021.108116
Received 6 July 2020; Received in revised form 16 March 2021; Accepted 11 April 2021
Available online 15 April 2021
1389-1286/© 2021 Elsevier B.V. All rights reserved.
R. Yarinezhad and S. Azizi Computer Networks 193 (2021) 108116

to send their data to the sink [26–29]. However, the clustering method the underlying routing solution.
cannot support the sink’s mobility quite well, and therefore, when there
is a mobile sink in the network, the clustering method is usually com­ 2.1. Non-geographic-basic routing protocols
bined with other methods such as the proposed protocol in [30].
Furthermore, when a virtual infrastructure is used for the entire Some protocols assume that the sensor nodes can send their data
network, it increases the likelihood of the occurrence of the hot-spot directly to the sink and hence, do not use a multi-hop forwarding
problem near the virtual infrastructure. Moreover, in most of the exist­ fashion. Such protocols are usually based on the clustering method
ing routing protocols, the geographic routing protocol is used as the because in this method only a subset of the nodes called Cluster Heads
underlying routing solution such as [21,31]. Unfortunately, the (CHs) needs to send their data to the sink. For example, in [32], the
geographic routing algorithm has some disadvantages. One disadvan­ clustering method is combined with an improved version of the Genetic
tage of this algorithm is that the energy consumption is not balanced Algorithm (GA) to design a routing protocol in which the CHs send their
among all the nodes in the network and as a result, the network lifetime data directly to the sink. In [33], in order to cluster the sensor nodes, an
is not maximized. Another is the low reliability of the routing algorithm improved version of Differential Evolution (DE) is combined with the
since some data packets may not be transferred to the sink. In Section 4, clustering method. In [34], a 1.5-approximation algorithm is proposed
these weaknesses are discussed in more detail. in order to perform a load-balanced clustering in the network. In [35], a
In this paper, we propose a new energy-efficient, energy-balanced, layer structure is suggested to design a clustering protocol for routing
and reliable routing protocol for power saving in green-IoT networks. among the sensor nodes towards the sink. In [36], a neuro-fuzzy system
The routing protocol uses two different mechanisms to overcome the is applied in selecting the optimal CHs to form the clusters in IoT net­
problems mentioned above. To reduce the chance of the occurrence of works. In [37], an fpt-approximation algorithm is designed to form the
the hot-spot problem and increase the network lifetime, the network clusters. In [38], a new improved version of the GA is given to select the
area is divided into two sections. In the smaller section, a virtual tree- CHs and form the clusters. In [39], in order to solve the optimal clus­
based structure is used to update the next-hop relay nodes of each tering problem, a multi-objective particle swarm optimization is used.
node towards the mobile sink with minimum changes and energy con­ The two main disadvantages of these protocols are the high-energy
sumption. In the larger section, an improved version of the geographic consumption of the sensor nodes due to the use of a single-hop
routing algorithm is used that addresses the weaknesses of the standard communication method and impracticality in the real world due to the
geographic routing algorithm. The proposed protocol is called RTG, limited radio range of the sensor nodes.
which stands for Routing based on the Tree and Geographic methods. Other protocols, instead of sending their data directly to the sink, use
The simulation results show that RTG provides a better network per­ a multi-hop communication fashion. Now, we review a number of these
formance compared to other similar protocols. Some key features and protocols in which the next-hop relay nodes towards the sink are
contributions of this paper are highlighted as follows. determined without using the geographic routing algorithm. In [40], a
novel Energy-Aware and layering-based Clustering and Routing Proto­
• This paper proposes a routing protocol for power saving in green-IoT col (EA-CRP) has been proposed that divides the network area into
networks that effectively manages the network area. It divides the several layers. In order to determine the next-hop of each node for the
area into two parts and provides a suitable routing mechanism for multi-hop communication, EA-CRP considers the two factors of the
each part based on its characteristics. Thus, it can also be applied to distance between the nodes and the remaining energy of each node.
large scale IoT networks. However, this method is practical only for a static sink. In [43] and [44],
• The "ImprovedGeographic" procedure is the first proposed mecha­ two other cluster-based routing protocols have been proposed that use
nism that is an improved version of the geographic routing algo­ the virtual layer structure. However, these two protocols have the same
rithm. It addresses the weaknesses of the traditional geographic weaknesses as EA-CRP.
routing algorithm and increases its efficiency. In [43], a new heuristic algorithm for forming the clusters in the
• The second procedure called "Tree-Based" is a new mechanism for network is proposed in which each cluster forms a tree whose root is the
routing data packets near the mobile sink that creates and updates CH. Then the mobile sink travels near the CHs and collects their data.
the paths with the least control packets. Furthermore, it prevents the However, in this method, each CH must wait for the sink to be in its
hot-spot occurrence in the network. vicinity in order to be able to send its data, which causes a high delay in
• In general, the proposed routing protocol selects the shortest paths the network.
for each node towards the sink, updates these paths efficiently, and The Sink oriented Tree-based Data Dissemination (STDD) [24] is
balances the energy consumption in the network. Thus, it can reduce another tree-based routing protocol. STDD forms a tree whose root is the
the end-to-end delay in the network and prolong the network lifetime sink and when the sink changes its location, STDD updates the tree based
as well as provide a reliable routing algorithm. on a number of rules to keep a correct routing in the network. In this
protocol, some nodes may be forced to transfer their data to the sink on a
The remainder of the paper is organized as follows: We review some long path leading to a high end-to-end delay. In [44], the Delay Aware
related works in the next section. An overview of the standard Energy-Efficient Reliable Routing (DA-EERR) protocol has been pro­
geographic routing algorithm is introduced in Section 3. Section 4 ex­ posed, which is a routing protocol based on a virtual ring structure. In
plains the proposed protocol. A detailed comparative evaluation of the DA-EERR, a new routing algorithm has been proposed to select the
proposed protocol against other works is reported in Section 5 and next-hop relay node of each node based on the remaining energy and the
finally, Section 6 concludes the paper. location of the nodes. In [45], a tree structure is created to form the
routing among the nodes towards the sink. However, in some scenarios,
2. Related Works this protocol may form long paths between some nodes and the sink,
which leads to a high delay.
In the last decade, different routing mechanisms have been proposed In [46], a layer structure and a clustering method are used to select
to use energy efficiently in battery-powered IoT networks. These the best possible next-hop for each node. In [47], a path selection
mechanisms use different methods to reach their goals. In this section, mechanism is suggested based on the shortest paths and number of hops.
some of the recent outstanding works are examined in two categories In [48], a cluster-based routing mechanism has been proposed that uses
based on their methods. The first category consists of those works that the genetic algorithm to select the CHs. In [49], a routing protocol based
do not use geographic routing in their protocol. The second category on the k-d tree algorithm has been introduced. However, these protocols
consists of the works that use geographic routing whether directly or as are only practical for networks with static sinks.

2
R. Yarinezhad and S. Azizi Computer Networks 193 (2021) 108116

In [50], an Ant Colony Optimization-based mobile sink path deter­ [56], a structure similar to a plus sign is defined in the network and the
mination protocol (ACO) is proposed that determines the movement next-hop of each node is determined based on this structure. In [57], a
path of the mobile sink by an ant colony optimization algorithm. In this virtual wheel-based structure is introduced to design an energy-efficient
method, an ant colony optimization algorithm determines the optimal routing protocol. In [27], a virtual square is formed in the network and
set of the RPs and a re-selection procedure balances the energy con­ the square is rotated to update the paths near the sink. In this work, a
sumption among the nodes by reselecting the RPs. However, this method new fpt-approximation algorithm produces an energy-balanced clus­
can have a high delay since the RPs must wait for the sink to arrive at one tering of the nodes. In the routing phase of this protocol, the geographic
of their adjacent sojourn locations in order to be able to send their data. routing algorithm is used along with the virtual square structure.
Rendezvous points (RPs) is another method that can be used to
2.2. Geographic-based routing protocols collect data from the sensor nodes in networks with mobile sinks. In this
method, the sink travels near a number of pre-determined nodes called
In geographic routing, each node selects its nearest neighboring node RPs and collects their data, which is received from the environment as
to the destination as its next-hop relay node. This method can be used well as the other nodes. The locations that the sink must travel in order
alone or along with other methods [41,42]. In [21], a virtual ring-based to collect data from the RPs are called sojourn locations. In [58], a new
infrastructure is introduced in which the nodes use geographic routing Virtual Grid-based Rendezvous point and Sojourn location Selection
to send their data and control packets towards the sink or the ring nodes. (VGRSS) method is suggested, which uses a grid structure to determine
The ring nodes store the sink’s location, which is the destination of the the RPs and sojourn locations. For each cell, an RP is selected and the
data packets. A node obtains the sink’s location by sending a packet to intersection of every four cells is assumed as a sojourn location. How­
the nearest ring node. Then, the node sends its data to the destination by ever, each RP node must wait for the sink to arrive at one of its adjacent
geographic routing. This method, which is called Ring Routing, is sojourn locations to be able to send its data, which can increase the
effective in energy conservation. However, due to the high traffic near network delay.
the ring structure, the nodes near the ring die sooner, which reduces the Two other geographic routing based protocols have been proposed in
network lifetime. To address this problem, Nested Routing [22] has been [59] and [60]. In these two works, two fpt-approximation algorithms
proposed. In this protocol, several nested rings are defined in the have been proposed for clustering. In the routing phases of these pro­
network. In this way, not only the hot-spot problem near the rings is tocols, two geographic routing based algorithms are suggested. In [59],
addressed but also the delay in the network is reduced. the next-hop of each node is determined based on the distance and en­
In [23], a Virtual Grid-based Dynamic Routes Adjustment scheme ergy. In [60], the next-hops are determined with the help of a grid
(VGDRA) has been proposed that uses a mobile sink around the network structure. However, these protocols are only practical for networks with
area. Each cell has a cell header node that is connected to the sink static sinks.
through its neighboring cell headers. In this method, geographic routing There are other approaches that have been proposed for green IoT
is used differently. In fact, the idea behind connecting the cell headers to networks [61–64]. Also, in [65], for industrial IoT, an adopted
the sink and the cell members to each other can be seen as a new version energy-efficient architecture has been proposed. It includes three layers,
of geographic routing. However, in this method, the sink moves outside which are the sensing layer, the gateway layer, and the control layer.
the network area leading to a high end-to-end delay. This model increases the lifetime of the system by balancing the traffic
The grid structure is an effective structure that has been used load. In [66], a new solution for increasing the battery lifetime in Ma­
recently by many researchers. In [31], a Virtual Grid-Based structure chine to Machine (M2M) communications has been proposed. However,
(VGB) routing protocol has been introduced that uses a grid structure to this method is related to the physical layer while the proposed algorithm
select several specific nodes that save the location of the sink and share it of this paper is related to the network layer. It means that both algo­
with the ordinary nodes in the network. In [51], a grid structure is rithms can be used in one network simultaneously to increase the
defined that can support multiple mobile sinks. In [52], a data dissem­ network lifetime.
ination model is proposed that uses a new grid structure. This protocol is Table 1 presents a summary and comparison of the various protocols
called Routing Based on Grid structure and Mobile sink (RBGM) in discussed in this section. According to the previous works, the clustering
which the sink’s path is pre-determined and the sink’s neighboring cells method cannot support having a mobile sink well enough and the virtual
are changed due to its movement, the routing protocol updates the infrastructure based protocols increase the likelihood of the occurrence
next-hop of the nodes based on their relative distance and grid structure. of the hot-spot problem near the virtual infrastructure. Moreover,
Routing based on Cellular structure and Clustering (RCC) [30] is another methods based on RPs have a high network delay and the standard
routing protocol based on the grid structure and the clustering method. geographic routing algorithm has some disadvantages such as unbal­
In this protocol, each cluster head in a cell is connected to the sink anced energy consumption and low reliability. Therefore, in this paper,
through its neighboring cluster heads. When the sink changes its loca­ we propose a new energy-efficient, energy-balanced, and reliable rout­
tion, RCC uses several energy-efficient rules to update the routing with ing protocol for power saving in green-IoT networks. The proposed
minimal changes. However, the protocol cannot support multiple mo­ protocol uses two different mechanisms to overcome the mentioned
bile sinks. problems. To prevent the hot-spot problem and increase the network
In [25], the Railroad protocol has been proposed that creates a vir­ lifetime as well as better manage the nodes in the network, the network
tual rail-based structure in which the nodes that form the rail share the area is divided into two sections. In the area nearer to the sink, a tree
sink’s location with the other nodes. The Hexagonal cell-based Data structure is formed that is updated with an efficient updating procedure.
Dissemination (HexDD) [53] uses a set of hexagons to form its virtual For the nodes located in the area farther from the sink, an improved
structure. The Line-Based Data Dissemination (LBDD) [54] protocol version of the geographic routing algorithm has been proposed that can
defines a virtual line-based structure in the middle of the network area. address the problem of the standard geographic routing algorithm. The
The sensor nodes send their data to the nodes located on this line so that simulation results show that the proposed protocol provides a better
the sink can collect it. The Geographical Cellular-like Architecture network performance than the other similar protocols.
(GCA) [55] protocol creates a cellular-like structure in the network.
Then, the cell headers deliver the data to the sink. The main disadvan­ 3. The standard geographic routing algorithm
tage of these virtual structure based protocols is that the nodes forming
the structure and the ones located near it die sooner than the other The geographic routing algorithm is a routing solution with position-
nodes, which causes the network lifetime not to be maximized. aware sensors [67,68]. It is an efficient and scalable routing method in
There are more protocols based on virtual structures. For example, In which each node needs only have the local information instead of the

3
R. Yarinezhad and S. Azizi Computer Networks 193 (2021) 108116

Table 1
Summary and Comparison of previous works.
Protocol Mobile Method Multiple Using geographic Advantage Disadvantage
sink sinks routing

Genetic Algorithm No GA and clustering No No Low delay Single-hop

[32]
Differential Evolution No DE and clustering No No Low delay Single-hop
[33]
Neuro-fuzzy [36] No A neuro-fuzzy system and clustering No No Low delay Single-hop
EA-CRP [40] No Layer structure No No Low energy consumption Short network lifetime
STDD [24] Yes Tree structure Yes No Supporting multiple sinks High delay
DA-EERR [44] Yes Virtual ring structure No No Low delay High energy
consumption
Genetic Algorithm No GA and clustering Yes No Low delay Short network lifetime
[48]
ACO [50] Yes Ant colony optimization No No Energy-balanced High delay
Ring Routing [21] Yes Ring structure No Yes Energy-efficient Short network lifetime
Nested Routing [22] Yes Nested rings structure Yes Yes Long network lifetime High energy
consumption
VGDRA [23] Yes Grid structure No Yes Energy-efficient High delay
RBGM [52] Yes Grid structure No Yes Long network lifetime Predefined sink
movement
RCC [30] Yes Grid structure No Yes Long network lifetime No multiple sinks
HexDD [53] Yes Hexagonal structure No Yes Energy-efficient Short network lifetime
GCA [55] Yes Cellular structure No Yes Energy-efficient Short network lifetime
VGRSS [58] Yes Rendezvous points No Yes Long network lifetime High delay
RTG (this work) Yes Improved geographic routing and tree- No Yes Long network lifetime and No multiple sinks
based routing Low delay

global topology to determine its next-hop relay node towards the sink. protocol works periodically. In other words, the data collection phase is
This method can be applied to the networks with static sinks. However, divided into several rounds in which all the sensor nodes sense the area
in this paper similar to some of the recent works that were mentioned and forward their data to their next-hop.
above, the geographic routing algorithm is used for networks with In this study, the energy model is similar to [26] in which the energy
mobile sinks. required by the radio to transmit a k-bit message over a distance d is
In the geographic routing algorithm, each node needs to be aware of calculated by equation (1).
the location of its neighboring nodes and the destination of the data
kEelec + kεfs d2 d < d0
packets. Then, each node selects its neighboring node closest to the ETx (k.d)= { , (1)
kEelec + kεmp d4 d ≥ d0
destination as its next-hop. When the destination (i.e. the sink) is static,
this method is energy-efficient due to selecting the shortest path for data
where Eelec is the energy required by the electronics circuit, εfs and εmp
transmission.
are the energy required by the amplifier, and threshold d0 is calculated
The geographic routing does not apply to scenarios with mobile sinks √̅̅̅̅̅̅̅̅̅̅̅̅̅̅
as d0 = εfs /εmp . Furthermore, the energy required by the radio to
since the nodes are not aware of the destination location. One way to
address this problem is for the protocol to continuously propagate the receive a k-bit message is calculated by equation (2).
latest destination location throughout the network. However, this 0ERx (k) = kEelec (2)
method is not energy-efficient. Thus, the geographic routing algorithm
needs to be adapted to be practical and energy-efficient for networks In this paper, all the nodes have a uniform data generation rate and
with mobile sinks. Furthermore, geographic routing has several disad­ the energy spent by sensing has been balanced among all the nodes in
vantages that are discussed in the next section. the network. Thus, we do not consider the energy consumption for data
sensing. Furthermore, compared with data communication, the energy
dissipated by data aggregation is much smaller and is not taken into
4. Proposed Protocol
account.
In this section, the proposed protocol is presented by first explaining
a number of assumptions taken in the network and describing the 4.2. Proposed scheme
network model.
The main goal of the proposed power-saving scheme is to collect
4.1. Assumptions and network model sensor data and aggregate them in the sink. In this paper, a tree-based
routing protocol is presented, which uses a modified version of the
We assume that a set of N stationary sensor nodes are deployed geographic routing algorithm, and thus, we call this protocol Routing
randomly in an area of size X × X for the purpose of sensing data from based on Tree and Geographic (RTG) protocol. RTG is designed in such a
the area and transmitting it to the mobile sink. In the proposed protocol, way to reduce and balance the energy consumption in the network and
the sink moves randomly in a certain part of the network area, which is as a result, prolong the network lifetime.
explained in the next section. The sink is initially positioned in the In the network, nodes sense data from surrounding and send it to
center of the area. Moreover, the sensors nodes possess an internal GPS, their next-hop relay node to be transmitted to the sink. Before starting
which makes them position-aware. They also know the size of the region this process, RTG determines the next-hop node towards the sink for
and the initial position of the sink, which can be stored in their memory each node. The network operations in a network that uses RTG as its
before deployment. This assumption can be considered in this paper routing protocol are as follows. First, the nodes are deployed in the
because the proposed protocol has been designed for IoT networks based network area. Second, every node runs RTG to determine its next-hop
on WSNs. In this paper, we consider a “round” as each time the sensor relay node towards the sink. Then, the nodes start sensing and sending
nodes sense the area and send their data to the sink. It means that the data while the maintaining procedure is running to keep a correct

4
R. Yarinezhad and S. Azizi Computer Networks 193 (2021) 108116

routing in the network as well as prolonging the network lifetime. RTG

works in a distributed way and it does not require a flooding broadcast
in the whole network.
In general, due to the hot-spot problem, the section closer to the sink
must be considered as highly significant in prolonging the network
lifetime. In other words, if the energy of the nodes in this important
section is not properly managed, the network lifetime will be decreased.
Accordingly, RTG proposes a specific routing mechanism for this section
to better manage the energy of the nodes near the sink. In this mecha­
nism, the network area is divided into two separate virtual sections. The
outer section, which is larger than the inner section is called the Out­
Section, and the inner one is called the InSection, which is the area in
which the sink is permitted to move. The sink is at first deployed in the
center of the area and starts moving inside InSection randomly. These
sections are shown in Fig. 1.
In RTG, first, each node determines whether it is located in InSection
or OutSection. The destination of all data packets generated by the nodes
in OutSection is considered the center of the area, which is wrapped by
InSection. Once a data packet enters InSection, the nodes in InSection
deliver the packet to the sink. The routing for the OutSection nodes is
done by the "ImprovedGeographic" procedure, which is an improved
version of the geographic routing method and works with nodes that are
aware of the destination of the data. However, since the proposed pro­ Fig. 2. The RTG protocol flowchart.
tocol is based on a mobile sink, the "ImprovedGeographic" procedure is
not applicable for routing in InSection as the destination of the data 4.3. Network partitioning
packets is a mobile sink and changes over time. Therefore, RTG in­
troduces the "Tree-Based" procedure for routing in InSection. The As mentioned previously, the area near the sink is more important
"ImprovedGeographic" procedure and the "Tree-Based" procedure are than the rest of the network area in the sense that its heavy traffic can
explained in detail in Section 4.4 and Section 4.5, respectively. cause an earlier death of the sink’s neighboring nodes and as a result
In InSection, the nodes transfer the data packets generated by both reduce the network lifetime. Hence, to avoid this problem, in this pro­
OutSection and InSection nodes to the sink. The sink moves inside tocol, the sink can move in the network in order to change its neigh­
InSection to change its neighbor nodes in order to solve the hot-spot boring nodes and better balance the network traffic. However, one
problem. As mentioned above, the routing in InSection is done by the downside of having a mobile sink is that routing in the network becomes
"Tree-Based" procedure, which creates an initial routing and maintains it more difficult. To address these problems and better manage the area
while the sink is moving in InSection. All these aforementioned steps are near the mobile sink, RTG divides the area into two separate sections,
shown in the form of a flow chart given in Fig. 2. one near the sink and another farther from it.
One point about this flowchart is that the "ImprovedGeographic" and InSection, which is of size X3 × X3, is the section in which the sink is
"Tree-Based" procedures do not run completely in each run of the loop. permitted to move. This means that the sink only moves in one-ninth of
The "ImprovedGeographic" procedure has its own condition for running. the area. Fig. 3 shows the division of the area in more detail. The size of
Furthermore, each time the loop is executed, only a specific part of the InSection is determined based on the functionalities of the "Impro­
"Tree-Based" procedure runs. These steps are described below. vedGeographic" and "Tree-Based" procedures and is tested by simula­
tion. The results prove that this value is appropriate.
We have two types of nodes: InSection nodes and OutSection nodes.
Nodes that are located in InSection are InSection nodes and nodes that

Fig. 1. The network area virtual division in RTG. The arrows indicate the di­
rection of data flow. Fig. 3. Division of the network area in RTG.

5
R. Yarinezhad and S. Azizi Computer Networks 193 (2021) 108116

are located in OutSection are OutSection nodes. The first step of RTG is balanced in the network and as a result, the network lifetime is not
that each node determines its type. Since each node is aware of its po­ maximized. Second, the reliability of the routing algorithm is low
sition and the size of the area as well as the size of InSection, it can because some of the data packets may not be transferred to the sink due
determine whether it is located in InSection or OutSection by comparing to the infinite loop problem. The "ImprovedGeographic" procedure,
its position with the border of the InSection. RTG proposes two different which is introduced in the next subsection, is the RTG’s suggested
routing mechanisms for these sub-areas. The OutSection nodes use the routing mechanism for OutSection. This procedure solves the problems
following mechanism to determine their next-hop relay node towards in geographic routing.
the sink.
4.4.1. The "ImprovedGeographic" procedure
In order to address the weaknesses of the geographic routing algo­
4.4. Routing in OutSection rithm and use an energy-efficient and energy-balanced routing algo­
rithm in OutSection, the proposed protocol introduces a modified
In this section, the proposed mechanism for routing in the Out­ version of this algorithm. In geographic routing, the only factor
Section is described. OutSection nodes are located at a greater distance considered by each node in choosing its next-hop is the potential next-
from the sink compared to InSection nodes as the sink only moves inside hop’s distance from the sink. However, in this improved version, both
InSection. OutSection nodes can send their data packets to the sink by distance and the node’s energy are taken into account. In fact, in the
forwarding them towards the center of the area. In fact, the center of the process of next-hop selection, a limit is defined on the remaining energy
area is the destination of the data packets generated by OutSection of the nodes that helps balance the energy consumption in the network.
nodes. In this way, we will make sure that these data packets finally In order to address the infinite loop problem, the improved version of
enter the InSection, which wraps the center of the area. Then, the geographic routing categorizes the nodes into two separate groups: the
InSection nodes transfer these data packets to the sink. ones that can cause the infinite loop problem and those that create an
In OutSection, for forwarding data packets, the geographic routing efficient path towards the destination. The improved version of
method can be a good choice because each node is aware of its own geographic routing prohibits the use of the infinite loop causing nodes as
position, its destination, and its neighbor nodes. In the geographic next-hops. Moreover, the nodes that are efficient in creating paths to­
routing method, each node sends its packet to its neighboring node wards the destination have a higher priority for being selected as next-
closest to the sink. Once a data packet is received by a node, the receiver hop nodes.
node acts in the same way. The “ImprovedGeographic" procedure, which is shown in Fig. 5, is
Using geographic routing in OutSection can cause a decrease in both the improved version of geographic routing and it is used for OutSection
delay and network lifetime. The delay is decreased since it selects the routing. This procedure runs independently within each node. In the
shortest path between each node and the destination. However, this also network, each node has a NeighborTable that consists of its one-hop
causes a decrease in network lifetime since the energy of the nodes on neighboring nodes and their characteristics. This table is filled and
these paths is depleted faster and it is not balanced. updated by the beacon packets received by each node from its neigh­
Another disadvantage of geographic routing is that it might cause an bors. Table 2 shows the list of important packets used in RTG. Each
infinite loop for data packets. This happens when there is a node that has beacon packet created by a node includes its remaining energy and
less distance to the sink compared to all its neighbor nodes. For example, position. Upon receiving this packet, the information inside it is stored in
in Fig. 4, the two nodes B and C are two neighbors of node A. The dis­ the NeighborTable of the receiver node.
tances from B and C to the sink are both individually greater than the The nodes are divided into two categories. The first category is called
distance between A and the sink. In the geographic routing algorithm, hopeless nodes. A node is a hopeless node if its distance from the sink is
node C is chosen as the next-hop for node A because it is closer to the less than each of its neighboring nodes’ distance from the sink. For
sink compared to node B. In the same sense, node A is chosen as the next example, in Fig. 4, node A is a hopeless node. The second category is
hop of node C, causing an infinite loop. Thus, the data packets that are called hopeful nodes. A node is a hopeful node if it has at least one
generated by A and C never reach the sink because there is an infinite neighboring node whose distance to the sink is less than the distance
loop consists of nodes A and C for those data packets. In the following, between the node itself and the sink. For example, in Fig. 4, nodes B and
we refer to this weakness of geographic routing as the infinite loop C are hopeful nodes. Each OutSection node is aware of its position by
problem. GPS and the position of its neighbors by its NeighborTable as well as the
Accordingly, if we use the standard geographic routing method, we position of the destination, which is predetermined as the center of the
will face two problems. First, the energy consumption will not be area. Thus, each node can determine to which category it belongs. The
“ImprovedGeographic" procedure selects the next-hop nodes among
hopeful nodes. If there is not any hopeful neighbor, then a hopeless node
will be chosen as the next-hop node. The category of each node is
transferred to its neighbors through beacon packets and is stored in
NeighborListTable.
The "ImprovedGeographic" procedure consumes the energy of the
nodes uniformly. At first, it only allows nodes with remaining energy of
at least half their initial value to be selected as next-hop nodes. If there is
not a node with this characteristic, this limit will drop to a quarter of the
initial energy, after that one eighth, and so on and so forth. In Fig. 5, a
variable α is defined in order to implement this process. In each attempt
to select a proper next-hop, α is multiplied by 12. In order to make sure
this loop ends, we define a threshold th1 , which is equal to the energy
needed to send a data packet. Accordingly, if α < th1 , then the node is no
longer able to send a data packet, and its α factor will no longer be
multiplied by 12 as the node is dead.
Fig. 4. An example scenario in which node A does not have a proper neigh­ In this way, the improved version of geographic routing balances the
boring node to be selected as its next-hop relay node towards the sink in energy consumption among all the nodes in OutSection and increases
geographic routing.

6
R. Yarinezhad and S. Azizi Computer Networks 193 (2021) 108116

Fig. 5. The "ImprovedGeographic" procedure pseudocode, which runs inside each node.

node. Each node has the position of its neighbors in its NeighborTable
Table 2
that has been created by the beacon packets. An example of the initial
List of the important packets used in RTG.
routing in InSection is given in Fig. 6. In this example, each node has
Message Description selected a neighbor that has the shortest distance to the sink.
data_pkt The sensor data is transferred to the sink by this type of packet. As can be seen from Fig. 6, the InSection nodes and the sink form a
beacon This type of packet is used to create and update NeighborListTable. spanning tree with the sink as its root. The second layer in this tree
change_pkt The sink uses this type of packet to determine the active nodes.
consists of the sink’s one-hop neighbors, which are called active nodes.
IamActive_pkt Each active node informs its neighbors of its status with this type of
packet An active node is one that can send its data packet directly to the sink
IamDying_pkt With this type of packet, each node informs its neighbors that it is without any intermediate node. When the sink moves and the active
dying nodes change, the "Tree-Based" procedure, whose steps are given in
Fig. 7, updates the spanning tree to maintain a correct routing in
InSection.
network reliability. Upon the decrease of the energy of the nodes, in
Let R be the radio range of the sensor nodes. In this procedure, when
order to prolong the network lifetime, the "ImprovedGeographic" pro­
the sink moves 2R 3 meter, it broadcasts a "change_pkt" packet, which
cedure needs to be run to change the paths and balance the energy
consumption among all nodes. However, every time this procedure is includes its location. In other words, if the displacement of the sink is
run, each node broadcasts its beacon packet, which is energy consuming. equal to 2R
3 , it broadcasts a "change_pkt" packet. This value is specified by
Thus if this procedure runs too many times, it ends up wasting the en­
ergy of the nodes instead of preserving it. In this regard, we consider a
constant value called numR in each node. When the number of rounds
exceeds this constant, the nodes rerunning the "ImprovedGeographic"
procedure. An optimal value for numR is determined in Section 5.

4.5. The "Tree-Based" procedure

The "ImprovedGeographic" procedure cannot be applied to InSection

because the destination’s position of the data packets is not constant as
the sink is mobile. The reason behind the sink’s mobility is to prevent the
hot-spot problem by constantly changing the sink’s neighboring nodes.
However, the weakness of this method is that the nodes are not aware of
the sink’s position. One way to inform all InSection nodes about the
sink’s position is to use the flooding method. This method is not energy-
efficient as it needs to run continuously due to the continuous movement
of the sink. Thus, RTG proposes a new routing mechanism for InSection,
which is called the "Tree-Based" procedure.
The initial position of the sink is the center of the area and all the
nodes are aware of that. Thus, at first, each InSection node selects a
neighbor node that has the minimum distance to the sink as its next-hop Fig. 6. An example of the initial routing in InSection.

7
R. Yarinezhad and S. Azizi Computer Networks 193 (2021) 108116

Fig. 7. The steps of the "Tree-Based" procedure.

the simulation done in Section 5. The purpose of the "change_pkt" When the sink goes out of the radio range of an active node, the
packets is to update the list of the active nodes and make them aware of active node must change its status to ordinary node. Thus, in each node,
the sink’s location. In this paper, by broadcasting a packet, we mean that we define a timer with duration t seconds. An active node that does not
the source node broadcasts the packet in its radio range and thus only its receive a "change_pkt" packet for longer than t seconds is no longer an
one-hop neighbor nodes can receive the packet. active node. Thus, it sets its ActiveNode to false. This node is aware of the
Let S = {s1 , s2 , …, sN } be the set of sensor nodes. Each node si ∈ S has previous location of the sink and since the nodes close to this location
an internal variable called "ActiveNode", which is equal to true if si is a are more likely to be the currently active nodes, it chooses its neighbor
neighboring node of the sink, and false otherwise. The initial value of node that is closest to the previous location of the sink as its next-hop.
each node’s ActiveNode variable is false. A sensor node, upon receiving a Even though this choice of next-hop may not be optimal, it does not
"change_pkt" packet, sets this variable to true and broadcasts an pose a problem because an optimal next-hop will be chosen for this node
"IamActive_pkt" packet. This packet includes the sink’s location and in the next step of the procedure. In Section 5, an optimal value for t is
informs the normal nodes of their neighboring active nodes. The pseu­ determined.
docode for the updating part of the "Tree-Based" procedure that runs The "IamActive_pkt" packets that are broadcasted by the active nodes
inside each node is shown in Fig. 8. update the next-hop of the sink’s two-hop neighbor nodes. When a non-
According to this, "change_pkt" determines the active nodes, which active node receives an "IamActive_pkt" packet from a node that is closer
are the sink’s one-hop neighbor nodes, and "IamActive_pkt" transfers the to the sink than its current next-hop, it sets the sender of the packet as its
location of the sink to sink’s two-hop neighbor nodes. Other InSection next-hop node. These steps are repeated until the end of the network
nodes can connect to the sink through these nodes. In Fig. 9, an example lifetime.
of these nodes is shown. When the network is working, it is natural for some nodes to die

Fig. 8. The pseudocode for the "Tree-Based" procedure, which runs inside each sensor node.

8
R. Yarinezhad and S. Azizi Computer Networks 193 (2021) 108116

"IamDying_pkt" packet, it removes the sender node from its

NeighborListTable.
So far, the next-hop of each node has been determined, which means
the sensor nodes can begin collecting data and transmitting it. In each
round, sensor nodes collect data from the area and send that data in
addition to the data packets they received from other nodes to their next-
hop nodes.
By the routing mechanisms described above, RTG effectively man­
ages the network area. It divides the area into two parts and provides a
suitable routing mechanism for each part based on the characteristics of
that part. The first mechanism, which is called "ImprovedGeographic",
addresses the weaknesses of the traditional geographic routing method
and increases its efficiency. The second mechanism called "Tree-Based"
is a new way of routing near the mobile sink that updates the paths with
the least control packets and prevents the hot-spot problem. In general,
RTG selects the shortest paths and balances the energy consumption in
the network. Thus, it can reduce the delay in the network, prolong the
network lifetime, and provide a reliable routing algorithm. In Section 5,
the performance of the proposed protocol is evaluated by simulation.
Fig. 9. A sample InSection: The hollow Circle displays the sink, the sink’s radio Before that, an example of the "Tree-Based" procedure is given in the
range is marked by a dotted circle. The squares show active nodes. The two-hop next section.
neighbor nodes of the sink are represented with pentagons, and R denotes the
radio range of the nodes.
4.6. Example
sooner than others. Thus, the proposed procedure needs to have a
mechanism to update the NeighborTable in the nodes. One way to ach­ In InSection, when the sink moves, the "Tree-Based" procedure up­
ieve this goal is for the nodes to broadcast the beacon packets periodi­ dates the next-hop of the InSection nodes to maintain a correct routing
cally, which is very energy wasteful. Thus, the "Tree-Based" procedure among the nodes in this section towards the sink. Updating the next-hop
uses a different mechanism. In the "Tree-Based" procedure, each node of the nodes is done by several rules mentioned above. In this section, an
whose remaining energy is less than threshold th1 , which was defined in example of the updating process done by "Tree-Based" is given. Consider
the previous section, broadcasts an "IamDying_pkt" packet to inform its a network scenario that is shown in Fig. 10 (a) in which the sink is
neighbors that its energy is being depleted. Whenever a node receives an located at the center of the area. In Fig. 10 (b), the sink has moved from
point A to point B and has updated the set of active nodes. The old active

Fig. 10. An example of the "Tree-Based" procedure in a sample InSection. The hollow Circle displays the sink. The sink’s radio range is marked by a dotted circle. The
squares show active nodes. Capital letters show the sink’s locations.

9
R. Yarinezhad and S. Azizi Computer Networks 193 (2021) 108116

nodes have also been updated and other nodes have changed their next- path are set according to the corresponding reference. The value of the
hop based on the new set of active nodes. Then, in Fig. 10 (c), the sink other parameters and the parameters of the energy model are deter­
has moved from point B to point C. Fig. 10 (d) shows the changes in the mined similar to [26], which are given in Table 3.
next-hop of the nodes when the sink has moved from point C to point D. In these simulations, the values of the parameters numR and t that
As can be seen from this example, in "Tree-Based", the rules that are were defined in Section 4 are equal to 75 and 5, respectively. These
proposed for updating the routing do not create long paths. In fact, most values have been obtained by rerunning RTG’s simulation for different
of the paths are the shortest paths possible or at least near to the shortest values of these parameters and selecting the ones that lead to the
paths possible. Furthermore, if a long path is generated, it will be optimal results.
changed to the shortest path shortly after the sink moves. When a green-IoT network is designed, the main goal is to create a
network with the maximum network lifetime and the minimum delay in
5. Performance Evaluation delivering data messages. It is also desired to transfer as many data
messages to the destination as possible. Thus, the most important met­
The performance of the RTG protocol is evaluated using simulation. rics in such networks are the network lifetime, end-to-end delay, and the
This section presents the simulation scenarios, performance metrics, and number of data messages that are received by the sink. In order to
the results. The simulation results of RTG are compared with the con­ prolong the network lifetime, the nodal energy consumption must be
ducted simulation results of the ACO[50], RCC [30], VGRSS [58], and reduced, and thus, another important metric in such networks is energy
STDD [24] protocols. The reasons for comparing RTG with the consumption. These metrics are defined as follows.
mentioned protocols are mentioned briefly in the following. ACO is a
routing protocol based on a mobile sink and an ant colony optimization • Network lifetime is defined as the number of the rounds until 75
algorithm, which is a meta-heuristic algorithm. RTG is compared with percent of the nodes die. 75 percent is selected because all the pro­
ACO to show the effect of using the proposed protocol instead of using tocols need to have a sufficient number of alive nodes in the network
meta-heuristic algorithms. The RCC protocol uses a virtual infrastruc­ in order to be able to implement their procedures.
ture, which is similar to a grid structure. Thus, comparing RTG with RCC • Delay is defined as the average time between the moment a source
illustrates the advantage of using the proposed virtual tree-based transmits a packet to the moment the sink receives the packet.
structure instead of other virtual structures such as grids. VGRSS uses • Throughput at the sink is defined as the number of data packets
a mechanism that is a combination of a virtual grid structure and successfully received at the sink.
rendezvous points. Hence, comparing RTG with VGRSS shows that the • Energy consumption is defined as the total energy that is consumed
proposed mechanism is a lot more efficient than the combination of the by the communication of the nodes.
two mentioned methods. Finally, STDD is a routing protocol based on a
tree structure. Therefore, evaluating RTG with STDD shows that 5.2. Simulation results and analysis
combining the tree structure with the improved geographic routing al­
gorithm makes it more effective. In this section, the mentioned metrics are examined by varying the
number of nodes in each scenario. First, the network performance of
5.1. Simulation scenarios and performance metrics RTG is evaluated in terms of network lifetime. We ran the protocols by
varying the number of sensor nodes from 300 to 900 in both NET#1 and
We use the OMNeT++ network simulator to implement RTG and all NET#2. Fig. 11 shows the results. In NET#2, the network area is bigger
the other protocols. OMNeT++ is a modular and component-based C++ than that of NET#1. Therefore, the energy consumption in NET#2 is
simulation library and framework. For the sake of simulation, two higher than NET#1, and thus its network lifetime is shorter. Further­
network scenarios NET#1 and NET#2 are defined. In NET#1, the more, increasing the number of nodes will increase the number of
network area is of size 350m × 350m and R = 60m, and in NET#2, the packets, which will reduce the network lifetime.
network area is of size 800m × 800m and R = 80m. As mentioned The simulation results show that the RTG protocol has the longest
before, the sensor nodes and the sink all have the same radio range of R. network lifetime among all the mentioned protocols. This is because the
The initial energy of each node is equal to 4.0 J. In each scenario, the improved geographic routing algorithm balances the energy consump­
number of the nodes varies from 300 to 900. The results that are shown tion in OutSection and the tree structure as well as the sink’s mobility
in the next section are obtained from the average of 25 runs of the balance the energy consumption in InSection. These two procedures
simulations. The simulation results of all of the protocols are conducted together by balancing the energy consumption among all the nodes in
with the same parameter values shown in Table 3. In each network the network, increase the lifetime of each individual node which leads to
scenario, there is one mobile sink that moves at a constant speed of 10 prolonging the overall network lifetime. The network lifetimes of ACO
m/s. For each protocol, the other parameters related to the sink and its and RCC are quite close as they both balance the energy in the network
and reduce the overall energy consumption. VGRSS and ACO both use
Table 3 RPs; however, VGRSS has a shorter network lifetime than ACO because
Simulation Parameters. ACO uses a more accurate method to select its RPs. Moreover, STDD
does not aim to select the shortest paths between the source nodes and
Parameter Value
the sink and as a result, these paths may become too long, which can be
Number of nodes 300 - 900
very energy consuming in the long run. Therefore, STDD has the shortest
Sensor field in NET#1 350m × 350m
network lifetime among all the mentioned protocols.
Sensor field in NET#2 800m × 800m
Next, the protocols are compared in terms of throughput at the sink.
R in NET#1 60m
R in NET#2 80m
Fig. 12 shows the results in both NET#1 and NET#2. The throughput at
Data packet size 40 bytes the sink in RTG is higher than those of the other protocols because RTG
Control packet size 200 bits has a longer network lifetime.
The initial energy of sensor nodes 4.0 J Next, the protocols are compared in terms of delay. Fig. 13 illustrates
The speed of sink 10 m/s
the results of both NET#1 and NET#2. When the nodes in the network
Number of simulation iterations 25
Eelec 50 nJ/bit are increased, the number of packets increases as well, which increases
εfs 10 pJ/bit m2 the end-to-end delay. RTG has the shortest delay among the mentioned
εmp 0.0013 pJ/(bit m4 )
protocols because of three reasons. First, the improved version of the
geographic routing algorithm selects the shortest paths in OutSection.

10
R. Yarinezhad and S. Azizi Computer Networks 193 (2021) 108116

Fig. 11. Comparison in terms of network lifetime in (a) NET#1 and (b) NET#2.

Fig. 12. Comparison in terms of throughput at the sink in (a) NET#1 and (b) NET#2.

Fig. 13. Comparison in terms of delay in (a) NET#1 and (b) NET#2.

Second, the tree-based procedure connects the nodes to the sink with the RCC. On the basis of delay, the difference between RCC and RTG is that
shortest paths in InSection. Third, all these paths always lead to the sink; RCC selects the shortest paths between the cell headers and the sink but
meaning that the nodes can always send their data to the sink without RTG selects the shortest paths between each individual sensor and the
any delay. In ACO and VGRSS, the RPs need to wait for the sink’s sink. This causes RTG to have a shorter delay than RCC. However, RCC’s
presence in their radio range in order to be able to send their data. Thus, delay is still less than the other three protocols since it selects relatively
these protocols are higher in delay compared to all the other mentioned short paths that are always connected to the sink.
protocols. As mentioned before, STDD may select long paths between Finally, we ran the protocols for 600 nodes and measured the total
the nodes and the sink and in this way increase the end-to-end delay. In energy consumption in both NET#1 and NET#2. Fig. 14 shows the re­
fact, the simulation results show that the delay in STDD is higher than sults. Based on these results, RTG consumes less energy in each round

11
R. Yarinezhad and S. Azizi Computer Networks 193 (2021) 108116

Fig. 14. Comparison in terms of total energy consumption in (a) NET#1 and (b) NET#2.

which causes the network to work for more rounds and thus live longer Declaration of Competing Interest
compared to the other protocols. This can be mainly due to two reasons.
First, in this protocol, the shortest paths between each node and the sink All authors declare that they have no conflict of interest. Further­
are selected, which is less energy consuming. Second, the tree-based more, the authors declare that they have no known competing financial
updating procedure updates the paths upon the sink movement with interests or personal relationships that could have appeared to influence
the minimum energy consumption possible. The simulation results also the work reported in this paper.
suggest that the sum of the energy consumption of all the rounds in RTG
is higher than those of the other protocols, which means that RTG uses Reference
the energy of the nodes more efficiently.
[1] M. Stoyanova, Y. Nikoloudakis, S. Panagiotakis, E. Pallis, and E. K. Markakis, “A
survey on the internet of things (IoT) forensics: challenges, approaches, and open
7. Conclusion and future works issues,” IEEE Commun. Surv. Tutorials, vol. 22, no. 2, pp. 1191–1221, 2020.
[2] R.P. Meenaakshi Sundhari, K. Jaikumar, IoT assisted Hierarchical Computation
Strategic Making (HCSM) and Dynamic Stochastic Optimization Technique (DSOT)
In this paper, a tree-based routing mechanism for power saving and for energy optimization in wireless sensor networks for smart city monitoring,
reducing the end-to-end delay in energy-efficient green-IoT networks Comput. Commun. 150 (2020) 226–234.
with a mobile sink has been proposed. This routing algorithm creates [3] E. Borgia, The internet of things vision: Key features, applications and open issues,
Comput. Commun. 54 (2014) 1–31.
and maintains a virtual tree structure consisting of the sensor nodes and [4] S. Li, L.Da Xu, S. Zhao, The internet of things: a survey, Inf. Syst. Front. 17 (2)
the mobile sink in an efficient and low-cost matter. This tree structure (2015) 243–259.
provides a routing tree in the network that connects the sensor nodes to [5] A. Al-Fuqaha, M. Guizani, M. Mohammadi, M. Aledhari, M. Ayyash, Internet of
Things: A Survey on Enabling Technologies, Protocols, and Applications, IEEE
the mobile sink on the shortest paths with proper intermediate nodes for
Commun. Surv. Tutorials 17 (4) (2015) 2347–2376.
data transmission. Furthermore, the proposed routing protocol has [6] M. binti Mohamad Noor, W.H. Hassan, Current research on Internet of Things (IoT)
introduced a new version of the geographic routing algorithm. The security: A survey, Comput. Networks 148 (2019) 283–294.
traditional geographic routing algorithm is unreliable and does not have [7] M. Kassab, J. DeFranco, P. Laplante, A systematic literature review on Internet of
things in education: Benefits and challenges, J. Comput. Assist. Learn. 36 (2)
an overall balanced energy consumption. In the proposed version of the (2020) 115–127.
geographic routing algorithm, these weaknesses have been addressed. [8] L. Chettri, R. Bera, A comprehensive survey on Internet of Things (IoT) toward 5G
The simulation results show that the proposed routing protocol provides wireless systems, IEEE Internet Things J 7 (1) (2019) 16–32.
[9] Z. Abbas, W. Yoon, A survey on energy conserving mechanisms for the internet of
a better network performance compared with the state-of-the-art ap­ things: Wireless networking aspects, Sensors (Switzerland) 15 (10) (2015)
proaches. Since in some applications of green-IoT networks there are 24818–24847.
multiple mobile sinks, improving the proposed algorithm in such a way [10] S. Abbasian Dehkordi, K. Farajzadeh, J. Rezazadeh, R. Farahbakhsh,
K. Sandrasegaran, M. Abbasian Dehkordi, A survey on data aggregation techniques
for it to support this feature increases the applicability of this protocol. in IoT sensor networks, Wirel. Networks 26 (2) (2020) 1243–1263.
Hence, we consider this as a future topic of interest to improve the [11] F. Fanian, M. Kuchaki Rafsanjani, Cluster-based routing protocols in wireless
proposed protocol. sensor networks: a survey based on methodology, J. Netw. Comput. Appl. 142
(2019) 111–142.
[12] C. Science, S. Engineering, Energy efficiency in wireless sensor network : a review,
Statement Comput. Networks 4 (3) (2014) 539–544.
[13] R.M. Curry, J.C. Smith, A survey of optimization algorithms for wireless sensor
network lifetime maximization, Comput. Ind. Eng. 101 (2016) 145–166.
The work presented in the submission „A Green Multipath TCP [14] R. Yarinezhad, S.N. Hashemi, Distributed faulty node detection and recovery
Framework for Industrial Internet of Things Applications” does not scheme for wireless sensor networks using cellular learning automata, Wirel.
Networks 25 (5) (May 2019) 2901–2917.
involve the use of human subjects.
[15] H. Yetgin, K.T.K. Cheung, M. El-Hajjar, L. Hanzo, A survey of network lifetime
maximization techniques in wireless sensor networks, IEEE Commun. Surv.
CRediT authorship contribution statement Tutorials 19 (2) (2017) 828–854.
[16] Z. Al Aghbari, A.M. Khedr, W. Osamy, I. Arif, D.P. Agrawal, Routing in wireless
sensor networks using optimization techniques: a survey, Wirel. Pers. Commun.
Ramin Yarinezhad: Conceptualization, Methodology, Software, (2019) 1–28.
Validation, Formal analysis, Investigation, Resources, Data curtion, [17] Y. Gu, F. Ren, Y. Ji, J. Li, The evolution of sink mobility management in wireless
sensor networks: a survey, IEEE Commun. Surv. Tutorials 18 (1) (2016) 507–524.
Writing – original draft, Writing – review & editing, Visualization. [18] S. Olariu, I. Stojmenović, Design guidelines for maximizing lifetime and avoiding
Sadoon Azizi: Conceptualization, Methodology, Validation, Formal energy holes in sensor networks with uniform distribution and uniform reporting,
analysis, Investigation, Resources, Data curtion, Writing – original draft, in: Proceedings - IEEE INFOCOM, 2006, pp. 1–12.
Writing – review & editing, Visualization, Supervision, Project admin­
istration, Funding acquisition.

12
R. Yarinezhad and S. Azizi Computer Networks 193 (2021) 108116

[19] M.I. Khan, W.N. Gansterer, G. Haring, Static vs. mobile sink: The influence of basic [50] P.K. D., T. Amgoth, C.S.R. Annavarapu, ACO-based mobile sink path determination
parameters on energy efficiency in wireless sensor networks, Comput. Commun. 36 for wireless sensor networks under non-uniform data constraints, Appl. Soft
(9) (2013) 965–978. Comput. J. 69 (2018) 528–540.
[20] R.C. Shah, S. Roy, S. Jain, W. Brunette, Data MULEs: Modeling a three-tier [51] H. Luo, F. Ye, J. Cheng, S. Lu, L. Zhang, TTDD: Two-tier data dissemination in
architecture for sparse sensor networks, Proc. 1st IEEE Int. Work. Sens. Netw. large-scale wireless sensor networks, Wirel. networks 11 (1–2) (2005) 161–175.
Protoc. Appl. SNPA 2003 1 (2–3) (2003) 30–41. [52] R. Yarinezhad, S.N. Hashemi, An efficient data dissemination model for wireless
[21] C. Tunca, S. Isik, M.Y. Donmez, C. Ersoy, Ring routing: an energy-efficient routing sensor networks, Wireless Networks 25 (6) (2019) 3419–3439, 11-Oct-.
protocol for wireless sensor networks with a mobile sink, IEEE Trans. Mob. [53] A.T. Erman, A. Dilo, P. Havinga, A virtual infrastructure based on honeycomb
Comput. 14 (9) (2015) 1947–1960. tessellation for data dissemination in multi-sink mobile wireless sensor networks,
[22] R. Yarinezhad, Reducing delay and prolonging the lifetime of wireless sensor Eurasip J. Wirel. Commun. Netw. 2012 (1) (2012) 17.
network using efficient routing protocol based on mobile sink and virtual [54] E. Ben Hamida, G. Chelius, A line-based data dissemination protocol for wireless
infrastructure, Ad Hoc Networks 84 (2019) 42–55. sensor networks with mobile sink, in: IEEE International Conference on
[23] A.W. Khan, A.H. Abdullah, M.A. Razzaque, J.I. Bangash, VGDRA: a virtual grid- Communications, 2008, pp. 2201–2205.
based dynamic routes adjustment scheme for mobile sink-based wireless sensor [55] X. Chen, M. Xu, A geographical cellular-like architecture for wireless sensor
networks, IEEE Sens. J. 15 (1) (2015) 526–534. networks, in: Lecture Notes in Computer Science (including subseries Lecture Notes
[24] A. Hawbani, et al., Sink-oriented tree based data dissemination protocol for mobile in Artificial Intelligence and Lecture Notes in Bioinformatics), 3794, LNCS, 2005,
sinks wireless sensor networks, Wirel. Networks 24 (7) (2018) 2723–2734. pp. 249–258.
[25] J.H. Shin, J. Kim, K. Park, D. Park, Railroad: Virtual infrastructure for data [56] R.K. Lenka, A.K. Rath, S. Sharma, Building Reliable Routing Infrastructure for
dissemination in wireless sensor networks, in: PE-WASUN’05 - Proceedings of the Green IoT Network, IEEE Access 7 (2019) 129892–129909.
Second ACM International Workshop on Performance Evaluation of Wireless Ad [57] S. Jain, K.K. Pattanaik, A. Shukla, QWRP: Query-driven virtual wheel based routing
Hoc, Sensor, and Ubiquitous Networks, 2005, pp. 168–174. protocol for wireless sensor networks with mobile sink, J. Netw. Comput. Appl. 147
[26] W.B. Heinzelman, A.P. Chandrakasan, H. Balakrishnan, An application-specific (2019), 102430.
protocol architecture for wireless microsensor networks, IEEE Trans. Wirel. [58] A. Mehto, S. Tapaswi, K.K. Pattanaik, Virtual grid-based rendezvous point and
Commun. 1 (4) (2002) 660–670. sojourn location selection for energy and delay efficient data acquisition in wireless
[27] R. Yarinezhad, S.N. Hashemi, Increasing the lifetime of sensor networks by a data sensor networks with mobile sink, Wirel. Netw. 26 (5) (2020) 3763–3779.
dissemination model based on a new approximation algorithm, Ad Hoc Networks [59] R. Yarinezhad, S.N. Hashemi, A routing algorithm for wireless sensor networks
100 (Jan. 2020), 102084. based on clustering and an fpt-approximation algorithm, J. Syst. Softw. 155 (2019)
[28] R.S.Y. Elhabyan, M.C.E. Yagoub, Two-tier particle swarm optimization protocol for 145–161.
clustering and routing in wireless sensor network, J. Netw. Comput. Appl. 52 [60] R. Yarinezhad, S.N. Hashemi, Solving the load balanced clustering and routing
(2015) 116–128. problems in WSNs with an fpt-approximation algorithm and a grid structure,
[29] M. Azharuddin, P.K. Jana, PSO-based approach for energy-efficient and energy- Pervasive Mob. Comput. 58 (2019).
balanced routing and clustering in wireless sensor networks, Soft Comput 21 (22) [61] C. Estevez, J. Wu, Recent advances in Green Internet of Things, in: 2015 7th IEEE
(2017) 6825–6839. Latin-American Conference on Communications, LATINCOM 2015, 2016, pp. 1–5.
[30] R. Yarinezhad, S.N. Hashemi, A cellular data dissemination model for wireless [62] R. Atat, L. Liu, J. Wu, G. Li, C. Ye, and Y. Yi, “Big data meet cyber-physical systems:
sensor networks, Pervasive Mob. Comput. 48 (2018) 118–136. a panoramic survey,” arXiv, vol. 6, pp. 73603–73636, 2018.
[31] R. Yarinezhad, A. Sarabi, Reducing delay and energy consumption in wireless [63] J. Wu, S. Guo, J. Li, D. Zeng, Big Data Meet Green Challenges: Greening Big Data,
sensor networks by making virtual grid infrastructure and using mobile sink, AEU - IEEE Syst. J. 10 (3) (2016) 873–887.
Int. J. Electron. Commun. 84 (2018) 144–152. [64] R. Atat, L. Liu, J. Wu, J. Ashdown, Y. Yi, Green massive traffic offloading for cyber-
[32] P. Kuila, S.K. Gupta, P.K. Jana, A novel evolutionary approach for load balanced physical systems over heterogeneous cellular networks, Mob. Networks Appl. 24
clustering problem for wireless sensor networks, Swarm Evol. Comput. 12 (2013) (4) (2019) 1364–1372.
48–56. [65] K. Wang, Y. Wang, Y. Sun, S. Guo, J. Wu, Green industrial internet of things
[33] P. Kuila, P.K. Jana, A novel differential evolution based clustering algorithm for architecture: an energy-efficient perspective, IEEE Commun. Mag. 54 (11) (2016)
wireless sensor networks, Appl. Soft Comput. J. 25 (2014) 414–425. 48–54.
[34] D. Wei, S. Kaplan, H.A. Chan, Energy efficient clustering algorithms for wireless [66] H. Chao, Y. Chen, J. Wu, Power saving for Machine to Machine communications in
sensor networks, IEEE Int. Conf. Commun. 31 (4) (2008) 236–240. cellular networks. 2011 IEEE GLOBECOM Workshops, GC Wkshps 2011, 2011,
[35] K.A. Darabkh, M.Z. El-Yabroudi, A.H. El-Mousa, BPA-CRP: a balanced power- pp. 389–393.
aware clustering and routing protocol for wireless sensor networks, Ad Hoc [67] H. Mostafaei, Energy-efficient algorithm for reliable routing of wireless sensor
Networks 82 (2019) 155–171. networks, IEEE Trans. Ind. Electron. 66 (7) (2018) 5567–5575.
[36] K. Thangaramya, K. Kulothungan, R. Logambigai, M. Selvi, S. Ganapathy, [68] M. Sajwan, D. Gosain, A.K. Sharma, Hybrid energy-efficient multi-path routing for
A. Kannan, Energy aware cluster and neuro-fuzzy based routing algorithm for wireless sensor networks, Comput. Electr. Eng. 67 (2018) 96–113.
wireless sensor networks in IoT, Comput. Networks 151 (2019) 211–223.
[37] R. Yarinezhad, S.N. Hashemi, Exact and approximate algorithms for clustering
problem in wireless sensor networks, IET Commun 14 (4) (Oct. 2020) 580–587.
[38] M.O. Oladimeji, M. Turkey, S. Dudley, HACH: Heuristic Algorithm for Clustering Ramin Yarinezhad received the BS degree in computer sci­
ence from Yazd University, Iran, and MS degree in computer
Hierarchy protocol in wireless sensor networks, Appl. Soft Comput. J. 55 (2017)
452–461. science from Amirkabir University of Technology, Iran. He is a
[39] S. Randhawa, S. Jain, MLBC: multi-objective load balancing clustering technique in Research Assistant in Computer Engineering at Amirkabir
wireless sensor networks, Appl. Soft Comput. J. 74 (2019) 66–89. University of Technology.
[40] K.A. Darabkh, N.J. Al-Maaitah, I.F. Jafar, A.F. Khalifeh, EA-CRP: a novel energy-
aware clustering and routing protocol in wireless sensor networks, Comput. Electr.
Eng. 72 (2018) 702–718.
[41] S. Halder, A. Ghosal, M. Conti, LiMCA: an optimal clustering algorithm for lifetime
maximization of internet of things, Wirel. Networks 25 (8) (2019) 4459–4477.
[42] R. Yarinezhad, A. Sarabi, MLCA: a multi-level clustering algorithm for routing in
wireless sensor networks, J. Commun. Eng. 8 (2) (2019) 249–265.
[43] S. Chowdhury, C. Giri, EETC: energy efficient tree-clustering in delay constrained
wireless sensor network, Wirel. Pers. Commun. 109 (1) (2019) 189–210.
[44] S. Maurya, V.K. Jain, D.R. Chowdhury, Delay aware energy efficient reliable
routing for data transmission in heterogeneous mobile sink wireless sensor Sadoon Azizi received the M.Sc. degree in computer science,
network, J. Netw. Comput. Appl. 144 (2019) 118–137. with focus on high-performance computing, and the Ph.D.
[45] H. Zhao, S. Guo, X. Wang, F. Wang, Energy-efficient topology control algorithm for degree in computer science, with focus on cloud data centers,
maximizing network lifetime in wireless sensor networks with mobile sink, Appl. from Amirkabir University of Technology, Tehran, Iran, in
Soft Comput. J. 34 (2015) 539–550. 2012 and 2016, respectively. He is currently an Assistant Pro­
[46] J. Huang, D. Ruan, W. Meng, An annulus sector grid aided energy-efficient multi- fessor with the Department of Computer Engineering and IT,
hop routing protocol for wireless sensor networks, Comput. Netw. 147 (2018) University of Kurdistan, Sanandaj, Iran. He is also the director
38–48. of the Internet of Things and Computing Systems Research
[47] L. Farhan, R. Kharel, O. Kaiwartya, M. Hammoudeh, B. Adebisi, Towards green Laboratory (IoTComSys Lab) and Head of the High-
computing for Internet of things: energy oriented path and message scheduling Performance Computing (HPC) center, University of Kurdi­
approach, Sustain. Cities Soc. 38 (2018) 195–204. stan. His current research interests include Cloud Computing,
[48] S. Verma, N. Sood, A.K. Sharma, Genetic Algorithm-based Optimized Cluster Head Fog/Edge Computing, Internet of Things, Green Computing,
selection for single and multiple data sinks in Heterogeneous Wireless Sensor and Data Center Networks. For additional information: https://research.uok.ac.ir/sazizi/
Network, Appl. Soft Comput. J. 85 (2019), 105788. en/.
[49] J. Anzola, J. Pascual, G. Tarazona, R. González, A clustering WSN routing protocol
based on k-d tree algorithm, Sensors (Switzerland) 18 (9) (2018) 2899.

13