Computer Networks 227 (2023) 109691

Contents lists available at ScienceDirect

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

Anchor-based void detouring routing protocol in three dimensional IoT

networks✩
Naveen Kumar Gupta a ,∗, Rama Shankar Yadav b , Rajendra Kumar Nagaria c , Deepak Gupta d ,
Achyut Mani Tripathi e , Om Jee Pandey f
a
Department of Information Technology, Dr B R Ambedkar National Institute of Technology Jalandhar, 144011, Punjab, India
b Computer Science and Engineering Department, Motilal Nehru National Institute of Technology Allahabad, Prayagraj 211004, India
c Electronics and Communication Engineering Department, Motilal Nehru National Institute of Technology Allahabad, Prayagraj 211004, India
d Faculty of Engineering and Technology, University of Lucknow, Lucknow 226007, Uttar Pradesh, India
e
Department of Computer Science and Engineering, Indian Institute of Technology Guwahati, 781039, Assam, India
f
Department of Electronics Engineering, Indian Institute of Technology (BHU) Varanasi, Varanasi 221005, Uttar Pradesh, India

ARTICLE INFO ABSTRACT

Keywords: In recent years, several applications of Internet of Things (IoT) have been observed in various areas including
IoT networks environmental monitoring, healthcare systems, cognitive smart agriculture, industrial control, smart homes,
Geographical routing intelligent transportation systems, and traffic management. For such applications, wireless sensor networks
Void node problem
(WSNs) are generally deployed to gather the sensed data from the targeted application field. In order to
Anchor node
transfer the sensor node data to the gateway (sink node), novel routing protocols need to be developed,
Orthogonal distance
3-dimensional routing
leading to reduced data transmission delay, high data throughput, and improved energy efficiency across the
Routing stretch network. In this context, geographical routing protocol has been considered as a promising approach for the
path selection in WSNs. This approach is full of scalability and multi-hop routing is performed using local
decisions. However, geographical routing protocols suffer from the void node problem (VNP) i.e., a region
where active nodes are not available in the direction closer to the destination. Numerous protocols have been
designed to get recovery from VNP in 2D networks which cannot be directly applied to 3D networks. The
3D routing includes the networks deployed in the hilly area, high buildings, airborne region, underground,
underwater and so forth. On applying the 2D routing protocols on complex 3D topology, the network may face
additional problems like packet looping, routing failure, ambiguity, or increased data latency due to longer
path. Further, the majority of geographical routing protocols follow the boundary of void which leads to a
longer path. In order to address the aforementioned challenges, this paper presents a novel anchor-based void
detouring routing (AVDR) protocol where anchor node is treated as a sub-destination which provides the
direct smaller path between source and gateway nodes. The proposed method bypasses the void boundaries
and directly connects source to anchor, anchor to destination, or two successive anchors. Further, anchor
information is distributed to the desired region to reduce the periodic anchor advertisement process. The
effectiveness of the proposed method has been tested over both, real field data set and simulated testbed with
OMNET++ simulator. The results obtained over real field data set claim that the proposed method takes only
29.09 ms (ms) for transferring the data on an average. However, this value is 32.37 ms, 34.32 ms, 33.61 ms,
37.20 ms, and 38.73 ms, respectively, using A3DR, EDGR, GPSR-3D, BSMH, and RPL methods. Moreover, it
is also noted that the proposed method achieves an improvement of 8.2%, 7.54%, 7.66%, 8.49%, and 8.22%,
in routing stretch when compared to aforementioned methods, respectively. This improvement with respect to
network overhead is 30.25%, 57.45%, 51.05%, 75.56%, and 58.89% using the proposed method.

✩ This document is the results of the research project funded by the Digital India Corporation (formerly media lab asia) under Ministry of Electronics and
Information Technology, Govt. of India, New Delhi.
∗ Corresponding author.
E-mail addresses: [email protected] (N.K. Gupta), [email protected] (R.S. Yadav), [email protected] (R.K. Nagaria), [email protected] (D. Gupta),
[email protected] (A.M. Tripathi), [email protected] (O.J. Pandey).

https://doi.org/10.1016/j.comnet.2023.109691
Received 24 May 2022; Received in revised form 3 March 2023; Accepted 7 March 2023
Available online 16 March 2023
1389-1286/© 2023 Elsevier B.V. All rights reserved.
N.K. Gupta et al. Computer Networks 227 (2023) 109691

1. Introduction

With the advancement in industries, the transformation of the

business system has increased through the use of the Internet of Things
(IoT). The IoT solutions offer connectivity anytime and anywhere
through a huge network of IoT devices containing sensors or softwares.
Sensors are crucial part of IoT network where a large number of sensing
devices are responsible for monitoring the application field [1]. The
IoT network is a network of several connected IoT devices where
each device is composed of a mandatory sensing unit, processing unit,
communication unit and an optional positioning unit. It is responsible
for sensing, processing, collecting and delivering the sensed data to the
gateway (sink node) for further processing. The gateway provides con-
nectivity between local sensor network and global network (Internet).
In order to achieve the vision of interconnecting the devices, there is a
requirement for a suitable routing approach to enable communication
among sensing devices. The routing protocol is responsible for commu-
nication between gateway and sensing devices where gateway may not Fig. 1. Route with following boundary nodes of void region and backtracking in case
be available in the direct transmission range of the current sensor [2, no other local alternative.

3]. Location based protocols are shown as promising approaches for

efficiently handling such types of conditions. Geographical routing
protocols (GRPs) consider the geographic locations of wireless nodes considers 3D topology where source node 74 transfers data packet to
where location is supported by global positioning system (GPS) or sink node 47. Fig. 2(a) represents the problem caused by the void in the
location-aware protocols [4–6]. The GRPs rely on local geographic network. Here, packet is stuck at void node 19 and starts following the
information and perform multi-hop routing with the help of local void boundary nodes through node 18 or node 56. In both of the cases
decisions. These protocols are also known as position-based, location- packet has to travel the longer path. However, the path length may be
based or geometric routing protocols. GRPs start with greedy mode minimized through utilizing the concept of anchor nodes. The concept
and switch to recovery mode on the occurrence of void node problem of anchor node is presented in Fig. 2(b) where node 18 and node 56 are
(VNP) [7–9]. The VNP is a big issue of geographic routing schemes considered as anchor nodes. The source node directly approaches the
where a node cannot find a neighbor that is closer to the destination. anchor nodes which reduces the path length between source and sink
The face routing based approach provides a deterministic path toward pair. Thus, it is observed that direct approaching the anchor nodes will
the destination in connected 2D networks. This approach follows a be more effective for reducing the route length in case of void in IoT
longer path through boundary nodes to detour the network void network.
region [10,11]. However, most existing geographical route selection The current trend of IoT networks is not limited to the territory,
schemes are designed for two dimensional networks, but these schemes these networks are also required at high buildings, hilly areas, un-
cannot be directly employed in practical three-dimensional networks. derground or underwater [5,13,14]. In such networks, the location of
While going through the communication from source to the sink nodes is represented by a three-dimensional (3D) coordinate system.
that has a void in the intermediate path, it has been observed that Traditional 3D GRPs also have provision to get a recovered path from
each packet is required to travel a longer path, i.e., following the path VNP problem, but the packets have to follow the boundary nodes of
around the void. However, for communication from another source, it the void region. This type of recovery strategy unnecessarily increases
is again required to search the path even if another source is nearer the path length where an alternate smaller path might be possible. The
to the previous one. The communication with longer distance searched anchor-based methods focus on minimizing the path length in the larger
path incurs a noticeable amount of search time overhead, which may volume of IoT network with sufficiently increased void in the network.
suffer multiple backtracking. As a result, it consumes a lot of energy. Moreover, these methods are more suitable for multi-hop routing in
This observation can be seen in the example given in Fig. 1. Consider broad geographical region e.g., disaster prone hilly areas, step farming
the case where node 1 wants to send the data to the sink node. Here, fields, and multi-floor buildings, among others. In this work, we have
communication starts from node 1 and the packet is stuck at node 5. addressed the problem of the longer path during detouring the void and
There is no directly focused forwarding node available towards the proposed a novel anchor-based void detouring routing protocol (AVDR)
sink. As a result, it has to select node 12 and start following the path for three-dimensional IoT networks. This work is focused on locating
around boundary. Further, it can be seen that while communicating and advertising intermediate nodes that play vital roles in reducing the
from node 5 to node 7, it is moving toward the source. Node 7 is path length. These nodes act as intermediate source nodes as well as
closer to the source and can be reached from source node 1 via node local destinations, also termed anchor nodes. The major contributions
numbers 6, 8 requiring length 3 hop count. Thus, there is a need for of this paper are summarized as follows:
direct communication between a source node and node 7 for next time
without following the longer path. It will avoid additional searching (i) We have proposed a novel method for locating the anchor
time for node 7 via node 5. Further, for the source node 23, the packet node(s) in the large scale 3-dimensional networks. First, we
is required to backtrack then, forwarding the packet to node 16. Here, construct a plane by using source node, destination node and
node 16 can be reached from node 23 via node 29 with the length of void node’s 3D coordinates. Then, we create a perpendicular
2 hop count. Further, previously gathered information can be used to plane with respect to the said plane to divide the desired part
get the path without involvement of new searching. In this case, node of the network. Subsequently, we apply an orthogonal distance
number 7 and node 16 can be used to avoid searching for a particular based procedure in each sub-part of the network.
segment. These nodes may act as intermediate nodes for a group of (ii) We have developed a novel method to determine the scope of
source nodes and their information is available in each node of the anchor nodes to avoid repeated anchor discovery process. Here,
group. we adaptively identified a group of nodes which lies inside
Another case is shown in Fig. 2 where real field data is taken from of this scope. Further, we distribute the anchors’ location to the
testbed available at FIT-IoT Lab [12], Strasbourg, France. This network identified group of nodes. The nodes from this group may utilize

2
N.K. Gupta et al. Computer Networks 227 (2023) 109691

Fig. 2. Illustrates real field networks developed using FIT-IoT lab datasets [12], (a) consisting void node (b) consisting anchor nodes addressing the void node problem shown in
(a).

the location of already discovered anchor nodes for subsequent MDT protocol uses virtual links to handle the VNP problem and dy-
communication. Additionally, each node adaptively selects the namic topology changes. The MDT produces a good packet reception
suitable anchor from available anchor nodes. Therefore, this ratio and routing stretch, but the end-to-end delay is very high. The
method helps in reducing the path length, avoiding unnecessary WEAVE [10] protocol used the concept of checkpoints and waypoints
searching of the path, and minimizing the network overhead and to route the packets in large-scale 2D and 3D networks. Most of these
overall energy consumption. approaches use periodic beacons to ensure connectivity. In reality,
(iii) A novel energy measurement model is developed correspond- maintaining connectivity all the time is not a good idea in any ad hoc
ing to the energy consumption by investigating the practical or sensor network. A node requires the next hop only when it has data
scenarios for the data transfer over the developed IoT networks. packets to transmit. Further, the comprehensive survey on 3D routing
protocols is given in [4,5,30] where it has been observed that packets
The rest of the paper is organized as follows: Section 2 describes follow void boundary nodes to reach the sink node [19,24,25,31,32].
related research work done in the literature. The modeling and as- As a result, boundary nodes get overloaded and each packet has to
sumptions related to the proposed method are given in Section 3. go through a longer path, whereas some other alternative may be
Section 4 presents the detailed description of the proposed anchor available [33]. Considering this problem, researchers have proposed
based protocol. Further, experimental work, along with its analysis, is intermediate sub-destination based protocols [11,34–37]. The interme-
given in Section 5. The experimental work includes simulated testbed diate sub-destination is also termed as anchor node where packets may
and real field testbed. Finally, the paper’s conclusion is available in follow several anchors to avoid the void boundary and minimize the
Section 6. path length. The main issue with the anchor-based approach is to find
a suitable anchor node that could optimize the average path length.
2. Related work An effort has been made to find suitable anchor nodes in the
literature [4,11,34,37]. The idea of projected distance based approach
The geographical routing protocols rely on local decisions where has been proposed in EAGR [34], EMGR [37] and EDGR [11] routing
packets follow a greedy forwarding approach to reach the destination protocols. In case of network void, the anchor discovery packet tra-
node [15–18]. There is a possibility of unavailability of a suitable verses around boundary of the void and monitors the projected distance
forwarding node which satisfies the greedy approach criteria [19–21]. on source to sink connecting line. Initially, projected distance starts
This condition causes VNP problem [22]. The researchers have made an increasing and reaches at peak node before decreasing. This peak node
effort to fix the VNP problem with face routing and proposed Greedy- is considered as an anchor node. The protocols use a reactive approach
Face-Greedy (GFG) [23], Greedy Perimeter Stateless Routing (GPSR) to find the anchor list prior to starting the communication. The reactive
[19] and other variants [24,25]. These protocols rely on the planarity of approach unnecessarily utilizes the network resources every time. The
the network [5,26] which may not be possible in practical complex 3D EDGR is an anchor based dual path routing protocol where two anchor
topology [4,13,26]. These protocols follow the longer paths in practical nodes are to be discovered based on projected distance. In this protocol,
scenarios due to boundary based void recovery methods and lack of the source node starts anchor discovery process and then switches to
consideration of the third dimension in 3D networks. The path length data forwarding after identifying the anchor nodes. Each source node
followed by packet affects overall energy consumption. has to perform an anchor discovery process before data forwarding,
Many efforts have been made to get recovery from VNP in 3D which causes a high initial delay. Further, it relies on periodically bea-
networks [4,5]. Taking care of the third dimension, GDSTR-3D [27], con exchange, which creates overhead. As a result, this protocol suffers
Multihop Delaunay Triangulation (MDT) [28], WEAVE [10], 3D real- from initial delay as well as additional packet transmission overhead. It
time geographical routing protocol (3DRTGP) [7], OnionMap [13], is also mentioned that EDGR protocol considers 3D network topology.
GPSR-3D [29] and some other protocols came into the picture. Zhou However, it uses a 2D plane, ignoring one dimension during the anchor
et al. [27] proposed GDSTR-3D where each node keeps two-hop neigh- discovery process. Ignorance of one dimension may cause actual longer
bor information and sets up convex hull tree to route the packet around path in practical 3D networks. The protocol is restricted to the ideal
the void nodes. The maintenance of two hop information is quite conditions where left and right side anchors are available, but in real
memory consuming. Simon S. Lam and Chen Qian [28] have extended networks, void area may touch network boundary where such anchors
S4 algorithm in 3D network and introduced MDT routing protocol. may not be available.

3
N.K. Gupta et al. Computer Networks 227 (2023) 109691

Nguyen et al. proposed RVMH [38] protocol to deal with routing etc.) and communicates via wireless channel (e.g., Wi-Fi, ZigBee, UVB,
path length issues. Further, it is improved in the BSMH protocol [33] Bluetooth etc.).
and divided the entire approach into three parts. In the first part, apply
Location of sensor node: The location of sensor consists of their
tent rule to detect the void boundary nodes of entire network and create
parameters (X, Y, Z) where X, Y and Z are longitude, latitude and
convex polygons. In second part, propagate the polygon information to altitude of location where sensor node is deployed.
the entire network which reduces overall network overhead of repeated
polygon creation. After getting the void boundary locations, in the Sensor network: It consists of 𝑚 sensor nodes, i.e., 𝑛1 , 𝑛2 , 𝑛3 ,...., 𝑛𝑚 , sen-
third part, the packet forwarding has been started. However, approach sor nodes where a 𝑖th node 𝑛𝑖 have location (𝑥𝑖 , 𝑦𝑖 , 𝑧𝑖 ), communication
achieved better routing stretch and minimized network overhead, but it range 𝑅𝑖 , capable of performing some processing, gathering sensory
still has some functional limitations. Creating and maintaining the void information and communicating with other connected nodes.
convex polygons in practical 3D networks is very difficult. Further, each Volume of network (𝑉𝑛 ): The volume of a network is defined as X ×
node requires a lot of memory to store the polygons’ information. Apart Y × Z where X, Y and Z are the maximum value of longitude, latitude
from routing strategies, routing protocols rely on certain baseline model and altitude of all nodes deployed in the network.
based on current technology. In this context, Routing Protocol for Low-
Volume of 𝑖th node (𝑉𝑖 ): The volume of an 𝑖th sensor node is the
Power and Lossy Networks (RPL) [39,40] provides a baseline model for
spherical communication range covered by the node.
energy constrained IoT networks. RPL has been considered as de-facto
routing model for IoT networks which offers IPv6 addressing scheme Network density: It is defined as ratio of number of sensor nodes in
atop of IEEE 802.15.4 MAC for Low-Power and Lossy Networks. This network to the volume of network.
model can quickly create a primary route using a proactive approach, Solid angle (𝛼𝑖 ): For a node, solid angle is defined as angle of cone
share routing information and efficiently adapt the topology. where node represents vertex of cone and both node & sink are on the
The literature survey indicates that addressing the void node prob- vertical axis of the cone.
lem is the biggest challenge of any geographical routing protocol. In
case of 2D topology, this can be deterministically solved using face Volume of sector (𝑉𝑐 ): It is the volume of cone formed at any node for
routing approach by traversing the boundary nodes left-hand side or given solid angle.
right-hand side. In case of 2D plane, two anchor nodes are sufficient Neighbor node of 𝑖th node (𝑁𝑖 ): It is the set of nodes, which are at
to handle void node problem. One anchor node on left most side and one hop distance from the 𝑖th node and are lying in its communication
another anchor for right most side of void boundary can handle the range.
connectivity. On the other hand, there is no such deterministic left or
Diameter of cone: It is the base diameter of the cone.
right-side boundary present in 3D networks. Thus, the identification
of anchor node(s) is a challenging task in 3D IoT networks. Further, Forwarding node: A node lying in conical section and at one hop
it becomes more challenging while ensuring no packet looping, no distance of a node can be forwarding node.
routing failure, minimized network overhead, unambiguous routing Anchor node: This node is a special intermediate node acts as sub-
with minimized data latency. To fill these gaps, in this work, we shall destination as well as intermediate source node. It is responsible for
develop a novel anchor-based void detouring routing (AVDR) protocol receiving the data from source node or any previous sub-destination
to improve the performance of 3D IoT network. The applicability and transmitting it to the next sub-destination or sink node.
and effectiveness of the proposed method are illustrated over both
simulated and real field data sets. Plane 1: Plane containing 3D coordinates of three nodes source node
(S), void node (V) and sink node (D).

3. Network model Plane 2: Perpendicular plane to the plane 1 passing through the line
VD.
The AVDR routing protocol is used to route a packet from a source Orthogonal distance or projected distance: This distance is com-
to a gateway using the concept of local forwarding. The local forward- puted after packet passing through void node and it is distance between
ing is governed by constraints available at local node. These constraints node C𝑝 and VD line.
include direction of the next node, availability of forwarding node,
arep message: Anchor reply message is generated from recently dis-
selection of more effective forwarding node. Here, first of all, we
covered anchor node after completion of anchor discovery.
shall discuss the assumptions considered while modeling the proposed
protocol. These assumptions are based on the literature [41]. Top-left anchor: The anchor node discovered top of plane 1 and left
of plane 2 if projecting void node to the sink.
1. The network consists of three dimensional static randomly
Top-right anchor: The anchor node discovered top of plane 1 and right
deployed sensor nodes where node location after its deployment
of plane 2 if projecting void node to the sink.
in remains fixed.
2. There is a destination node, known as gateway/sink node, and Bottom-left anchor: The anchor node discovered bottom of plane 1
its location is known to each node in the network. Gateway and and left of plane 2 if projecting void node to the sink.
sink keywords are used interchangeably. Bottom-right anchor: The anchor node discovered bottom of plane 1
and right of plane 2 if projecting void node to the sink.
3.1. Terms used
3.2. Network initialization
In this section, we discuss the terms and symbols used in the paper.
First, we provided a list of terms along with their description and then, The network is initialized by deploying 𝑚 number of randomly
distributed sensor nodes in 3D space specified by latitude, longitude
we prepared a table for special symbols. All the special symbols are
and altitude. The volume of network space where sensor nodes are
given the Table 1 followed by their brief description.
deployed in X × Y × Z volume. X, Y and Z are the respective maximum
Sensor node: It is a node consists of a set of embedded sensors, along difference in latitude, longitude and altitude among the coordinates of
with limited resources (e.g., computing, memory, battery and GPS, the network space. A node on its deployment, knows its own position

4
N.K. Gupta et al. Computer Networks 227 (2023) 109691

Table 1 The energy needed to transmit the 𝑘-bit data with data-rate 𝑟
Symbol table.
bits/second, considering path loss in the channel, is 𝑘 × 𝑐1 × 𝑑𝑖𝑗𝛽 . Here,
Symbol Description 𝑐1 is the distance-relative constant of energy consumption and 𝑑𝑖𝑗 is
𝑅𝑖 Communication range of 𝑖th node the euclidean distance between node i and node j. The 𝛽 is path loss
𝑉𝑛 Volume of network deployment area constant, generally between 2 and 6, that depends on the transmission
𝑉𝑐𝑖 Volume of conical sector generated with vertex environment [43]. Further, transmitted data is received by all the
as 𝑖th sensor node listener nodes in the transmission range of node i, so total energy
𝑉𝑖 Volume of 𝑖th sensor node covered with consumption during data reception is proportional to the number of
communication range nodes in the transmission range of node i. Therefore, the proposed
𝛼𝑖 Solid angle of a core formed with vertex as 𝑖th protocol adjusts the maximum power according to selected next hop,
node which minimizes the occurrence of other nodes in transmission region
𝑅𝑒𝑠𝑒𝑛𝑑 – List (at 𝑖th node) of unexpired recently of the sender node. Each listener utilizes power 𝑃 during transmission
𝑄𝑢𝑒𝑢𝑒𝑖 transmitted packets waiting for forwarded to time and consumes 𝑃 × 𝑘𝑟 energy to receive the transmitted data. The
next hop
one-bit receiving energy consumption by the nodes in the neighborhood
𝑇 𝑟𝑎𝑛𝑠𝑚𝑖𝑡𝑡𝑒𝑑– List (at 𝑖th node) of unexpired packetIds of is 𝑐3 = 43 𝜋 × 𝐷𝑛 × 𝑃 × 1𝑟 , where 𝐷𝑛 is the average network density in
𝑄𝑢𝑒𝑢𝑒𝑖 transmitted packets, which are forwarded by
next hop
terms of number of nodes per unit volume and 43 𝜋 × 𝑑𝑖𝑗3 is the volume
of transmission region covered by sender node 𝑖.
𝑇𝑑 Waiting delay time for packets to be forwarded
to next hop
The data is processed to the intended next hop only and other nodes
simply discard it. So, data is processed to sender node 𝑖 and receiver
𝑚 Total number of sensor nodes
node 𝑗 only, which consumes constant energy, i.e., 𝑘 × 𝑐2 . Finally, total
𝑁𝑖 Set of nodes corresponding to 𝑖th sender
energy consumption 𝐸𝑖𝑗 in one packet transmission is given in Eq. (1).
𝑆 Source node where packet is generated
𝐷 Destination node/sink node
𝐸𝑖𝑗 = 𝑘 × 𝑐1 × 𝑑𝑖𝑗𝛽 + 𝑘 × 𝑐2 + 𝑘 × 𝑐3 × 𝑑𝑖𝑗3
𝐶𝑖 Current node other than source and sink where [ ] (1)
packet is currently available ⟹ 𝐸𝑖𝑗 = 𝑘 𝑐1 × 𝑑𝑖𝑗𝛽 + 𝑐2 + 𝑐3 × 𝑑𝑖𝑗3
node j Forwarding candidate, inside the conical sector
of node 𝐶𝑖 , which is currently considering A node i receives data from the previous sender, process it and
forwarded to node j then its energy consumption 𝐸𝑖 is equivalent to
𝑑𝑖 Euclidean distance between current sender and
sink the sum of all three components. It can be computed using formula
provided in Eq. (2).
𝑐𝑗 Euclidean distance between potential
forwarding node and sink 1
𝐸𝑖 = 𝑘 × 𝑐1 × 𝑑𝑖𝑗𝛽 + 𝑘 × 𝑐2 + 𝑘 × 𝑃 ×
𝜃𝑖,𝑗 Computed angle from potential forwarding 𝑟
[ ] (2)
node to sink concentrating current sender 𝑖 𝛽 1
⟹ 𝐸𝑖 = 𝑘 𝑐1 × 𝑑𝑖𝑗 + 𝑐2 + 𝑃 ×
node V Starting point of void, termed as void node, 𝑟
with 3D coordinates V𝑥 , V𝑦 , V𝑧
4. Proposed anchor based void detouring routing (AVDR) protocol
VD line Line joining void node and sink
node C𝑝 current node, computing projected distance Utilizing the sub-destination node concept, the objective here is
with VD line, with 3D coordinates C𝑝𝑥 , C𝑝𝑦 , C𝑝𝑧 to reduce the path length by removing the redundant path segments
𝑑(𝑉 𝐷 → 𝐶𝑝 ) distance between VD line and node C𝑝 and their availability for future course of communication to the sink
∣ 𝑉⃗𝐷 ∣ magnitude of vector VD node. Future communication may include data transfer between same
anchor𝑙𝑒𝑓 𝑡 (A𝑙 ) left side anchor with respect to line joining source to sink or another source node to the sink. The proposed
void node and sink with 2D coordinates 𝐴𝑙𝑥 , 𝐴𝑙𝑦 protocol used an orthogonal distance based approach while discovering
anchor𝑟𝑖𝑔ℎ𝑡 (A𝑟 ) right side anchor with respect to line joining its communication path to the sink node. In the beginning, no node
void node and sink with 2D coordinates 𝐴𝑟𝑥 , 𝐴𝑟𝑦 knows about the void in network. When a node starts communicating
with sink node, it may get stuck with void. At this point, the current
node starts anchor discovery process which requires finding a suitable
sub-destination node. The number of anchors may vary according to
and position of sink node. These positions are measured in terms of
network structure, position or size of the void. The anchors’ information
3D coordinates. Further, it computes the network density and network
needs to be advertised to other nodes, which are in the scope of an-
density is stored at each deployed node. The network density is the
chors, upon discovering the anchor node(s). The anchor advertisement
ratio of number of nodes deployed to the total volume of the de-
ployed network i.e., network density 𝐷𝑛 = 𝑚∕𝑉𝑛 . The network density process minimizes the overhead of repeated anchor discovery. There
measures probabilistic availability of next available node as next hop. is a provision of sharing anchors’ information with other nodes in the
More density is more chance for availability of many forwarding nodes, scope of anchors. So, these nodes could utilize the anchors’ information.
whereas only fewer number node availability is possible in core of The protocol is distributed in nature where the entire algorithm is
sparsely deployed network. The non availability of network forwarding available at all nodes. Depending on the current condition, any node
node is creating the void. Further, network model and base protocol may work as source, destination or intermediate. The current node
(A3DR) are taken from previous research work [41]. react on specific events and takes appropriate action on receiving
the packet. In next subsections, we first discuss the anchor discovery
3.3. Energy consumption model process in detail followed by anchor advertisement.

The energy needed to send 𝑘-bit data from node i to one-hop 4.1. Anchor discovery
neighbor node j can be calculated using the energy consumption model,
which is similar to the models given in [11,34,42]. The major por- In the beginning, a source node transmits the packet focusing on
tion of energy consumption needed during (a) data transmission at the sink node using A3DR protocol. The packet may be stuck at a node
sender node i, (b) data reception at nodes (𝑗1 , 𝑗2 , 𝑗3 , …) available in the where it has to increase its angle 𝛼 repeatedly but did not find any
transmission range of node i and, (c) data processing at i and j nodes. neighbor in the forwarding direction. The neighbor may be available

5
N.K. Gupta et al. Computer Networks 227 (2023) 109691

Fig. 3. Orthogonal distance demonstration.

Fig. 4. Projection on source–sink line.

at higher value of angle 𝛼 where next selected forwarding node has

( )2 (
more distance from sink than the previous node. A node is said to be where 𝑄 = (𝑉𝑦 − 𝐶𝑝𝑦 )(𝑉𝑧 − 𝐷𝑧 ) − (𝑉𝑦 − 𝐷𝑦 )(𝑉𝑧 − 𝐶𝑝𝑧 ) − (𝑉𝑥 − 𝐶𝑝𝑥 )(𝑉𝑧 −
)2 ( )2
void node from where distance to the sink node starts increasing. The 𝐷𝑧 ) − (𝑉𝑥 − 𝐷𝑥 )(𝑉𝑧 − 𝐶𝑝𝑧 ) + (𝑉𝑥 − 𝐶𝑝𝑥 )(𝑉𝑦 − 𝐷𝑦 ) − (𝑉𝑥 − 𝐷𝑥 )(𝑉𝑦 − 𝐶𝑝𝑦 )
void node is responsible for starting anchor discovery process, finalizing
the anchors and broadcasting the anchors’ information to the scope of
Example 4.1. The demonstration of orthogonal distance is represented
anchor nodes.
in Fig. 3 where nodes D, V, C1 and C2 with their coordinates are node
The anchor discovery process is based on computation of orthog-
15 (800, 400, 0), node 5 (350, 430, 0), node 12 (320, 480, 0) and
onal distance between current node and straight line joining void
node and sink node. Suppose the coordinates of current node (C𝑝 ), node 19 (250, 540, 0) respectively. The straight line between node
void node (V), source node (S) and sink node (D) are (𝐶𝑝𝑥 , 𝐶𝑝𝑦 , 𝐶𝑝𝑧 ), 5 and node 15 is VD line and dotted line represents its continuation.
(𝑉𝑥 , 𝑉𝑦 , 𝑉𝑧 ), (𝑆𝑥 , 𝑆𝑦 , 𝑆𝑧 ) and (𝐷𝑥 , 𝐷𝑦 , 𝐷𝑧 ) respectively. The mathematical The orthogonal distance from node 12 and node 19 (represented using
formulation for orthogonal distance is shown in Eq. (3). red lines) can be computed using Eq. (7). The computed orthogonal
distances are 154.32 m from node 12 and 209.53 m from node 19.
|𝑉 ⃗𝐶𝑝 × 𝑉⃗𝐷|
𝑑(𝑉 𝐷 → 𝐶𝑝 ) = (3)
|𝑉⃗𝐷| Anchor Discovery Process: In case of void, forwarding nodes are
where 𝑉 ⃗𝐶𝑝 = (𝑉𝑥 − 𝐶𝑝𝑥 , 𝑉𝑦 − 𝐶𝑝𝑦 , 𝑉𝑧 − 𝐶𝑝𝑧 ) and 𝑉⃗𝐷 = (𝑉𝑥 − 𝐷𝑥 , 𝑉𝑦 − selected around the boundary of void, where orthogonal distances of
𝐷𝑦 , 𝑉𝑧 − 𝐷𝑧 ) are 3D vectors. The cross product of two vectors (𝑉 ⃗𝐶𝑝 and each node keep increasing. It means that we are moving away from the
sink node. During this selection, a node with the highest orthogonal
𝑉⃗𝐷) is as below in Eq. (4)
distance from other nodes, selected around the boundary of void. The
| |
| 𝑖 𝑗 𝑘 | node with the highest orthogonal distance is termed as an anchor node.
| |
𝑉 ⃗𝐶𝑝 × 𝑉⃗𝐷 = |𝑉𝑥 − 𝐶𝑝𝑥 𝑉𝑦 − 𝐶𝑝𝑦 𝑉𝑧 − 𝐶𝑝𝑧 | This anchor node is also known as intermediate destination toward sink
| |
| 𝑉𝑥 − 𝐷 𝑥 𝑉𝑦 − 𝐷𝑦 𝑉𝑧 − 𝐷𝑧 ||
| from the source node. The orthogonal distance starts decreasing after
⇒ 𝑉 ⃗𝐶𝑝 × 𝑉⃗𝐷 = (4) reaching the anchor node while traversing the void boundary. This can
𝑖[(𝑉𝑦 − 𝐶𝑝𝑦 )(𝑉𝑧 − 𝐷𝑧 ) − (𝑉𝑦 − 𝐷𝑦 )(𝑉𝑧 − 𝐶𝑝𝑧 )] be seen in Fig. 4 where node 9 has the highest orthogonal distance
whereas node 5 has least orthogonal distance.
− 𝑗[(𝑉𝑥 − 𝐶𝑝𝑥 )(𝑉𝑧 − 𝐷𝑧 ) − (𝑉𝑥 − 𝐷𝑥 )(𝑉𝑧 − 𝐶𝑝𝑧 )]
+ 𝑘[(𝑉𝑥 − 𝐶𝑝𝑥 )(𝑉𝑦 − 𝐷𝑦 ) − (𝑉𝑥 − 𝐷𝑥 )(𝑉𝑦 − 𝐶𝑝𝑦 )] Algorithm 1: Check for occurrence of void node on receiving the
packet.
The magnitude of these two vectors are:
Input: packet
|𝑉 ⃗𝐶𝑝 × 𝑉⃗𝐷| = Output: Start anchor discovery on occurrence of void
[{ }2
(𝑉𝑦 − 𝐶𝑝𝑦 )(𝑉𝑧 − 𝐷𝑧 ) − (𝑉𝑦 − 𝐷𝑦 )(𝑉𝑧 − 𝐶𝑝𝑧 ) − Procedure: checkForVoid (packet)
1: if (currentnode_sink.length > prevnode _sink.length) then
{ }2 2: packet.discovery = TRUE
(𝑉𝑥 − 𝐶𝑝𝑥 )(𝑉𝑧 − 𝐷𝑧 ) − (𝑉𝑥 − 𝐷𝑥 )(𝑉𝑧 − 𝐶𝑝𝑧 ) + 3: if (network_ structure == 2D) then
4: Duplicate the packet
{ }2 ] 21
5: Set directions as left and right
(𝑉𝑥 − 𝐶𝑝𝑥 )(𝑉𝑦 − 𝐷𝑦 ) − (𝑉𝑥 − 𝐷𝑥 )(𝑉𝑦 − 𝐶𝑝𝑦 ) (5)
6: Transmit packet in respective directions
7: else
and
√ 8: Make three copies of the packet
|𝑉⃗𝐷| = (𝑉𝑥 − 𝐷𝑥 )2 + (𝑉𝑦 − 𝐷𝑦 )2 + (𝑉𝑧 − 𝐷𝑧 )2 (6) 9: Set directions as top-left, top-right, bottom-left and bottom-
right
substituting the values of Eqs. (5) and (6) in Eq. (3)
10: Transmit packets in respective directions
𝑑(𝑉 𝐷 → 𝐶𝑝 ) = 11: end if
√ 12: anchorDiscovery (packet)
𝑄
(7) 13: end if
(𝑉𝑥 − 𝐷𝑥 )2 + (𝑉𝑦 − 𝐷𝑦 )2 + (𝑉𝑧 − 𝐷𝑧 )2

6
N.K. Gupta et al. Computer Networks 227 (2023) 109691

Algorithm 2: Anchor discovery process when packet stuck at void

node.
Input: packet
Output: arep message on discovering anchor node

Procedure: anchorDiscovery (packet)

1: if (packet.discovery == TRUE) then
2: Create anchor reply packet arep
Case1: Anchor node detected according to orthogonal distance
based procedure.
3: if (𝑑(𝑉 𝐷 → 𝑝𝑟𝑒𝑣𝑛𝑜𝑑𝑒) > 𝑑(𝑉 𝐷 → 𝑐𝑢𝑟𝑟𝑒𝑛𝑡𝑛𝑜𝑑𝑒) &&
(currentnode_sink.length < prevnode_sink.length)) then
4: packet.discovery = FALSE
5: Create anchor reply packet: arep
6: arep.anchorpos = selfposition
7: arep.isBoundary = FALSE
8: arep.direction = packet.direction
9: Transmit arep to void node
Case2: Anchor node not detected and reached to network
boundary.
10: else if (self_boundary_dist < comm_range && 𝛼 > 180) then
11: packet.discovery = FALSE
12: Create anchor error packet: aerr
13: aerr.direction = packet.direction
14: Transmit aerr to void node
15: end if
16: Transmit packet to next node
17: end if
Fig. 5. Perpendicular plane on the plane with source, sink and void nodes.

The anchor discovery process starts with forwarding of packet con- The anchor discovery process is again extended to cope coordinates.
tains discovery and direction fields to denote the status of packet. The extension includes condition selecting four directions, i.e., top-left,
The discovery field is initially set to false and it will be true on the top-right, bottom-left and bottom-right, for packet forwarding to dis-
occurrence of void node. Further, it becomes false on allocation of node cover the anchor nodes on the occurrence of void node. The equations
as anchor node. Algorithm 1 demonstrates the procedure to check the of base plane, containing source node, void node and sink node, and
occurrence of void node. After detecting the void node, start searching its perpendicular plane are given in Eqs. (9) and (12), respectively.
anchor nodes in four directions. These four directions are given as top- The values returned as four combinations of positive/negative numbers
left, top-right, bottom-left and bottom-right for 3D routing protocol. For on substituting the current node’s X, Y, and Z coordinates in these
clarity, we first discuss anchor discovery process for 2D plane where equations. The outcome of both equations may be positive–positive,
we can choose directions as left and right. The direction is calculated positive–negative, negative–positive or negative–negative, representing
based on equation of VD line i.e., line joining void node and sink. The top-left, top-right, bottom-left or bottom-right directions, respectively.
equation of VD line is given in Eq. (8). The value is returned as positive (
(𝑆𝑦 𝐷𝑧 + 𝑉𝑦 𝑆𝑧 + 𝐷𝑦 𝑉𝑧 − 𝑉𝑦 𝐷𝑧 − 𝑆𝑦 𝑉𝑧 − 𝐷𝑦 𝑆𝑧 )𝑋+
number or negative number when substituting the X, Y coordinates
of current node in this equation. The positive value indicates the left (𝑉𝑥 𝐷𝑧 + 𝑆𝑥 𝑉𝑧 + 𝐷𝑥 𝑆𝑧 − 𝑆𝑥 𝐷𝑧 − 𝑉𝑥 𝑆𝑧 − 𝐷𝑥 𝑉𝑧 )𝑌 +
direction, whereas the right direction to the line for negative value. (𝑆𝑥 𝐷𝑦 + 𝑉𝑥 𝑆𝑦 + 𝐷𝑥 𝑉𝑦 − 𝑉𝑥 𝐷𝑦 − 𝑆𝑥 𝑉𝑦 − 𝐷𝑥 𝑆𝑦 )𝑍+ (9)
This can be seen in Fig. 4 where on moving left direction path, we 𝑉𝑥 𝑆𝑦 𝐷𝑧 + 𝐷𝑥 𝑉𝑦 𝑆𝑧 + 𝑆𝑥 𝐷𝑦 𝑉𝑧 − 𝑉𝑥 𝐷𝑦 𝑆𝑧 − 𝑆𝑥 𝑉𝑦 𝐷𝑧
have reached node 9 via nodes 12, 19, and 7. Out of these nodes, node )
− 𝐷𝑥 𝑆𝑦 𝑉𝑧 = 0
5 has least orthogonal distance, while node 9 has maximum orthogonal
distance. Other nodes have distances between these two extremes.
The equation of perpendicular plane is derived as below:
(𝑉𝑦 − 𝐷𝑦 )𝑋 + (𝐷𝑥 − 𝑉𝑥 )𝑌 + 𝑉𝑦 (𝑉𝑥 − 𝐷𝑥 ) + 𝑉𝑥 (𝑉𝑦 − 𝐷𝑦 ) = 0 (8) The normalized form of equation of plane can be written as

Further, it is made clear that on following right direction path from 𝑎𝑋 + 𝑏𝑌 + 𝑐𝑍 + 𝑑 = 0 (10)
node 5, one has to backtrack up to node 3 and then start searching for Therefore, comparing Eq. (9) with Eq. (10) we have: 𝑎 = 𝑆𝑦 𝐷𝑧 +
another anchor node that is node number 17 via node 39, 24, 29, 16. 𝑉𝑦 𝑆𝑧 + 𝐷𝑦 𝑉𝑧 − 𝑉𝑦 𝐷𝑧 − 𝑆𝑦 𝑉𝑧 − 𝐷𝑦 𝑆𝑧 𝑏 = 𝑉𝑥 𝐷𝑧 + 𝑆𝑥 𝑉𝑧 + 𝐷𝑥 𝑆𝑧 − 𝑆𝑥 𝐷𝑧 −
Thus, we have two anchor nodes corresponding to void node number 𝑉𝑥 𝑆𝑧 − 𝐷𝑥 𝑉𝑧 𝑐 = 𝑆𝑥 𝐷𝑦 + 𝑉𝑥 𝑆𝑦 + 𝐷𝑥 𝑉𝑦 − 𝑉𝑥 𝐷𝑦 − 𝑆𝑥 𝑉𝑦 − 𝐷𝑥 𝑆𝑦 𝑑 =
5. The above process of finding out anchor nodes is given in Algorithm 𝑉𝑥 𝑆𝑦 𝐷𝑧 + 𝐷𝑥 𝑉𝑦 𝑆𝑧 + 𝑆𝑥 𝐷𝑦 𝑉𝑧 − 𝑉𝑥 𝐷𝑦 𝑆𝑧 − 𝑆𝑥 𝑉𝑦 𝐷𝑧 − 𝐷𝑥 𝑆𝑦 𝑉𝑧
2. So, the equation of perpendicular plane is the dot product
In three dimensional coordinate system, one has to search the ( )( )
anchor nodes at top-left, top-right, bottom-left and bottom-right di- 𝑉⃗𝐷 × (𝑎, 𝑏, 𝑐) . (𝑋 − 𝑉𝑥 ) + (𝑌 − 𝑉𝑦 ) + (𝑍 − 𝑉𝑧 ) = 0 (11)
rections. These directions can be obtained by considering two planes The simplified equation of plane, which is perpendicular to the
where one plane is plane coordinates: source node, void node and plane with source, sink and void nodes, is given as Eq. (12).
sink node, as shown in Fig. 5. Another plane is perpendicular to this
plane, passing through the line VD. The mathematical equation of the (𝑐𝑉𝑦 + 𝑏𝐷𝑧 − 𝑐𝐷𝑦 − 𝑏𝑉𝑧 )𝑋 + (𝑐𝐷𝑥 + 𝑎𝑉𝑧 − 𝑐𝑉𝑥 − 𝑎𝐷𝑧 )𝑌
plane passing through source node, void node and sink node is given + (𝑏𝑉𝑥 + 𝑎𝐷𝑦 − 𝑏𝐷𝑥 − 𝑎𝑉𝑦 )𝑍 + (𝑎𝑉𝑦 𝐷𝑧 + 𝑏𝐷𝑥 𝑉𝑧 + (12)
in Eq. (9). 𝑐𝑉𝑥 𝐷𝑦 − 𝑎𝐷𝑦 𝑉𝑧 − 𝑏𝑉𝑥 𝐷𝑧 − 𝑐𝐷𝑥 𝑉𝑦 ) = 0

7
N.K. Gupta et al. Computer Networks 227 (2023) 109691

Algorithm 2 considers two cases at the end of anchor discovery Algorithm 3: Advertise the location of newly discovered anchor node
process. Case 1: Discovered suitable anchor node in specific direction. to other nodes in the scope.
Case 2: The packet reaches and traverse the network boundary in Input: arep / aerr
specific direction but does not find anchor node. In case 1, anchor
node transmits arep (anchor reply) message to the void node. The Procedure: anchorAdvertisement (arep)
arep contains the anchor node’s details, e.g., location and direction. In 1: if (anchor detail is already exist) then
case 2, a node that detects the network boundary transmits the aerr 2: return
(anchor error) message to the void node. The aerr contains direction of 3: end if
void. The void node collects the arep/aerr messages and proceeds with 4: Compute 𝑀𝑙 AND 𝑀𝑟 as per Equations (14) and (13)
anchor advertisement process. Once the anchor nodes are detected, 5: if ((𝑀𝑙 > 0 && 𝑀𝑟 < 0) ‖ (𝑀𝑙 < 0 && 𝑀𝑟 > 0)) then
their use for other sources, which may communicate to the destination 6: Store anchor positions
node, must be ensured. This is achieved by advertising the anchor node 7: broadcast to neighbors
details to the other nodes in the anchor node’s communication zone. 8: end if
This is discussed in next subsection.

4.2. Anchor advertisement

The proposed protocol minimizes the packet transmissions using

anchor advertisement process. All other nodes do not need the anchor
discovery process. The anchor information is judicially shared with
other nearby nodes that may face void node problem. For such nodes,
anchors are intermediate destinations toward the sink node. Further,
these source nodes are said to be nodes lying in the scope of anchor
nodes. The decision about these nodes is done by means of (i) Projec-
tion lines in case of 2D coordinate system and (ii) Projection plane in
case of 3D coordinate system.
For 2D, lines joining the anchor node and sink node decide the
anchors’ scope. The equation of these lines are given in Eqs. (13) and
(14). We put the X, Y coordinate of a node in the line equation to
check the relative position of the node with respect to the line. If
the output value is positive, the node lies on one side of line and
negative output for another side. Therefore, from both line equations,
Fig. 6. Nodes in anchor forwarding zone.
the output (positive–positive, positive–negative, negative–positive &
negative–negative) can be combined as four directions. To ensure the
availability of node in the scope of anchor nodes, the direction of the
candidate node should be same as the direction of the source node. Example 4.2. This can be seen in Fig. 6 where nodes source, anchor𝑙𝑒𝑓 𝑡 ,
anchor𝑟𝑖𝑔ℎ𝑡 and sink node (D) with their coordinates are node 1 (0, 450,
𝑋(𝐷𝑦 − 𝐴𝑙𝑦 ) + 𝑌 (𝐴𝑙𝑥 − 𝐷𝑥 ) + 𝐷𝑥 𝐴𝑙𝑦 − 𝐴𝑙𝑥 𝐷𝑦 = 0 (13) 0), 9 (280, 640, 0), 17 (270, 230, 0), 15 (800, 400, 0) respectively.
Here, we have two lines: (i) line joining left anchor node and sink node
𝑋(𝐷𝑦 − 𝐴𝑟𝑦 ) + 𝑌 (𝐴𝑟𝑥 − 𝐷𝑥 ) + 𝐷𝑥 𝐴𝑟𝑦 − 𝐴𝑟𝑥 𝐷𝑦 = 0 (14) (i.e., node 9 & node 15) and (ii) line joining right anchor node and sink
node (i.e., node 17 & node 15).
It can be seen in Fig. 6 that some nodes are lying in the projection substituting the coordinates of node 9 and node 15 in Eq. (13), the
of these lines toward the considered source node. These nodes are equation of line joining anchor𝑙𝑒𝑓 𝑡 and sink node will be:
candidates for advertisement of anchor nodes’ information, i.e., node
numbers 1, 2, 3, 4, 5, 6, 7, 8, 12, 16, 19, 23, 24, 25, 26, 27, 28, 𝑋(400 − 640) + 𝑌 (280 − 800) + 800 × 640 − 280 × 400
(15)
29, 34, 35, 36, 37, 38, 39 are supposed to be user of anchor nodes. ⟹ −240𝑋 − 620𝑌 + 400000
First of all void node collects the information of anchor nodes. Then,
Similarly, the equation of line joining anchor𝑟𝑖𝑔ℎ𝑡 and sink node will be:
it starts advertising this information to the nodes between these two
lines. Anchor advertisement procedure is summarized as Algorithm
3. In case void node receives aerr message from all directions, it 𝑋(400 − 230) + 𝑌 (270 − 800) + 800 × 230 − 270 × 400
(16)
means network is disconnected and communication is not possible. This ⟹ 170𝑋 − 630𝑌 + 76000
message is forwarded to all connected nodes and stop routing process
putting the X, Y coordinates of node 1 in Eq. (15)
for disconnected segment.
This method can be extended to the case where we have four −240 × 0 − 620 × 450 + 400000 = 121000
anchor nodes for 3D coordinates. Here, we choose four planes: (i) plane and putting the X, Y coordinates of node 1 in Eq. (16)
containing top-left anchor, top-right anchor and sink node, (ii) plane
containing top-left anchor, bottom-left anchor and sink node, (iii) plane 170 × 0 − 630 × 450 + 76000 = −207500
containing bottom-left anchor, bottom-right anchor and sink node, and
Here, we found positive value with respect to line joining anchor𝑙𝑒𝑓 𝑡
(iv) plane containing top-right anchor, bottom-right anchor and sink
and sink node and negative value with respect to line joining anchor𝑟𝑖𝑔ℎ𝑡
node. All nodes (towards the source node) that lies within the scope
and sink node.
of these four planes will be candidate for advertisement. Let us check for node 36 (200, 500, 0) the output of equation of line
One node is detected as anchor node, and it is required to back- joining anchor𝑙𝑒𝑓 𝑡 and sink node
track from anchor node to source node via void node: when packet
is backtracking, it releases the resources captured during the forward −240 × 200 − 620 × 500 + 400000 = 42000
movement and frees all resources occupied between anchor nodes and
and the output of equation of line joining anchor𝑟𝑖𝑔ℎ𝑡 and sink node
source node. Further, source node start path searching with one anchor
node has least path length. 170 × 200 − 630 × 500 + 76000 = −205000

8
N.K. Gupta et al. Computer Networks 227 (2023) 109691

At node 30, the same pattern is followed and we got positive value Table 2
Experimental parameters.
for line joining anchor𝑙𝑒𝑓 𝑡 and sink node and negative value with
respect to line joining anchor𝑟𝑖𝑔ℎ𝑡 and sink node. Thus, it is in the scope Parameter Value

of anchor nodes. Simulator OMNET++ [44] (INET 3.6.3 framework)

Let us again check for node 39 (400, 470, 0) the output of equation MAC IEEE 802.11 (ad hoc mode)
of line joining anchor𝑙𝑒𝑓 𝑡 and sink node Data packet size 32 bytes (UDP datagram)

−240 × 400 − 620 × 470 + 400000 = 12600 Simulation time 300 s

Packet lifetime 1500 ms
and the output of equation of line joining anchor𝑟𝑖𝑔ℎ𝑡 and sink node (expiry time)
Target network area 400 m × 400 m × 300 m
170 × 400 − 630 × 470 + 76000 = −152100
Communication Varies for different nodes between 40
At node 39, the same pattern is followed and we got positive value range and 100 m (average and median near to
for line joining anchor𝑙𝑒𝑓 𝑡 and sink node and negative value with 70 m)
respect to line joining anchor𝑟𝑖𝑔ℎ𝑡 and sink node. Thus, it is also in the Number of nodes 1000 randomly deployed
scope of anchor nodes. Number of source 3 to 24
Let us check for node 10 (360, 630, 0) the output of equation of line nodes
joining anchor𝑙𝑒𝑓 𝑡 and sink node Void type (1) Cubic void with edge varies between
60 and 200 m.
−240 × 360 − 620 × 630 + 400000 = −77000 (2) Spherical void radiating between 30
and 100 m
and the output of equation of line joining anchor𝑟𝑖𝑔ℎ𝑡 and sink node
Packet rate 4 packets/s/source
170 × 360 − 630 × 630 + 76000 = −259700

In node 10, the same pattern is not followed and we got negative
values in both of the cases. Thus, this is not also in the scope of anchor so the performance of proposed protocol is compared with A3DR [41].
nodes. Additionally, to have in-depth performance analysis, it is also compared
with other existing methods including EDGR [11], GPSR-3D [29],
Now, the nodes in scope of anchors got the information about
BSMH [33] and RPL [39,40].
anchor nodes. So, the nodes within the scope of anchors, want to
communicate with the sink, compute the relative distance to sink 5.1. Experimental conditions
via all anchors. An anchor with minimum distance is treated as sub-
destination node. Above examples made it clear that the proposed The proposed protocol has been implemented on the INET frame-
method addresses the void node problem in sufficiently large volume work under the OMNET++ simulator by considering diversified sim-
of IoT network. If void region is very small then the proposed method ulation environment. OMNET++ is a discrete event simulator for In-
works like a traditional greedy geographical routing approach. In such ternet/network systems which is primarily used for research purposes.
cases, overhead of anchor selection and anchor advertisement processes It provides real wireless network environment to perform the exper-
gets reduced. Moreover, the proposed method is more suitable for imental work where network layer protocol stack is implemented.
multi-hop routing in broad geographical regions e.g., disaster prone The parameters taken for simulation work are given in Table 2. One
hilly area, step farming fields, and multi-floor buildings, and other large thousand IoT devices are randomly distributed in a 3D region of 400 m×
fields. In next section, we discuss the performance evaluation through 400 m×300 m. The void zone is created at center of the network. Cubical
results generated using simulated testbed and real field testbed. and spherical void structures are considered for the experimental work.
The sink node is fixed at one side of the network while source nodes are
5. Performance evaluation selected in such a way that each source has to pass the packet through
the void zone. Moreover, each node can directly communicate in its
Performance evaluation of the proposed method is carried out using communication range which radiates spherically. Further, considering
simulation study and real testbed results. The results concerning route the practical scenario, the communication range of each node may vary
length are obtained in terms of routing hop stretch. Hop stretch is due to its own environmental/physical conditions, so the communica-
the ratio of measured path length and minimum possible path length tion range of each node varies between 40 and 100 m. The average and
between source and destination. The measured path length is computed median of the communication range are near 70 m.
using proposed method which is a measure of hop count. The stretch Following three basic simulation scenarios are designed to evaluate
factor always becomes more than 1.0, but it should be close to 1.0 for the performance of the proposed protocol:
a good routing protocol. Further, its effect on other quality of service (i) Number of source node scenario: Here, number of source nodes
(QoS) parameters is accessed using packet delivery ratio (PDR), end- varies from 3 to 24 and a cubical void with 200 m edge is fixed at
to-end delay (E2ED) and routing overhead. The PDR is ratio of total center of the network. The 𝑋, 𝑌 , 𝑍 coordinates of center of void are set
number of packets received at the sink to the number of packets sent to (200, 200, 150). Other nodes may participate in the routing process.
from source node. End-to-end delay is the time difference between (ii) Void volume scenario: The number of source nodes is set to 10
packet generated at source node and delivered by the sink node. It is and center of spherical void is fixed at center of the network having
used to evaluate the timing performance of the protocol. Overhead is 𝑋, 𝑌 , 𝑍 coordinates (200, 200, 150). The void radius varies from 50 to
evaluated in concern with transmission ratio. The transmission ratio 120 m with an interval of 10 m.
is the ratio of total packets transmitted in the entire network over (iii) Multiple void scenario: The number of source nodes is set to
data packets generated from all source nodes. The number of packets 10 and two spherical voids with radii varying from 30 to 100 m. Two
becomes an overhead if a protocol requires additional packets due to spherical voids are placed so that both voids are separated on shorter
beacons, longer path and re-transmissions. In the simulation study, radii while these intersect each other on longer void radius. The 𝑋, 𝑌 , 𝑍
number of transmitted packets are fixed (i.e., according to the packet coordinates of centers of both of the voids are set to (150, 200, 150)
rate), so the traffic overhead is computed according to the average and (250, 200, 150), respectively.
number of transmissions occurred with respect to one packet from In the simulation study, each scenario is tested on 10 different
source node. The proposed AVDR protocol is an extension of A3DR, random networks (for same number of nodes) by changing 10 seed

9
N.K. Gupta et al. Computer Networks 227 (2023) 109691

Fig. 7. Effect of number of source nodes on packet delivery ratio.

Fig. 8. Effect of number of source nodes on routing stretch.

values. Further, for each network three repeated executions are carried
out. So, each reported result is an average of 30 outputs.

5.2. Number of source node scenario

In the proposed AVDR protocol, one source node discovers the

location of anchor nodes and forwards this information to other nearby
nodes. The boundary of these nearby nodes is decided by the position of
anchors and sink, which consists only nodes that may suffer from void
node problem. Therefore, the anchor discovery process requires only
once and is not affected by number of sources. On the other hand, EDGR
protocol requires anchor discovery with respect to each source node so
performance varies according to the number of source nodes. BSMH
protocol discovers and propagates void boundary nodes to the entire
network prior to data transmission. The packet rate is set fixed and
the number of source nodes is increasing linearly, so the total number
of packet transmissions is also increased. Such a scenario may cause
congestion in the network, resulting in packet loss, delay in packet
delivery, retransmissions, overhead etc. This effect is represented in
Fig. 9. Effect of number of source nodes on network overhead.
Figs. 7–10.
Fig. 7 depicts the delivery ratio of packets under number of source
nodes. When the number of sources increases, the packet delivery rate
of proposed A3DR protocol remains constant and provides guaranteed GPSR-3D protocols follow void boundary so it covers slightly longer
message delivery. There are four anchors available, so traffic divides path, which results higher route stretch. The EDGR protocol uses 2D
significantly and could not create congestion. Meanwhile, the A3DR plane by ignoring one dimension during anchor discovery process.
protocol follows the void boundary so initial packets are dropped due The ignorance of one dimension produces longer path which results
to the expiry of packet lifetime during covering the longer path. The greater path length. BSMH protocol assumes 2D network topology and
delivery ratio is still more than 98%. The GPSR3D protocol is showing traverses the void boundary so it is reporting highest routing stretch.
a downward trend but still maintaining a good delivery ratio of about The routing stretch of AVDR protocol is minimum and it is 10.89%,
95% in the worst case. EDGR and BSMH protocols rely on beaconing 13.42%, 12.07% and 21.89% better than EDGR, A3DR, GPSR-3D and
process and report high packet loss due to collision on high network BSMH protocols, respectively. This can be seen in Fig. 8. Next, we
traffic. Here, it can be seen that AVDR is providing 0.67%, 7.4%, 2.13% shall discuss about the effect of number of source nodes on network
and 13.66% more delivery ratio over A3DR, EDGR, GPSR-3D and BSMH overhead.
protocols, respectively. The following paragraph shows the impact on Data transmission amount is a good estimation of average energy
path length. consumption. It considers that most of the energy is consumed in data
The routing stretch depends on route length traveled by the packet. transmission. The number of packet transmitted is directly proportional
It is a ratio of measured path length traveled by packet and minimum to the number of sources because every source node generates the net-
possible path length in terms of hop count. It is also affected by work traffic. To cope with more number of packets, additional packets
increasing number of source nodes in view of higher congestion in the are also needed in terms of beacons, control packets, re-transmissions
network. The higher congestion causes an alternative longer path which etc. This effect is shown in Fig. 9. When number of source nodes were
leads to additional overall energy consumption. The hop stretch of set to 3, total of 3588 packets were generated from all the sources.
AVDR is minimum and varying between 1.10 and 1.15 because packets As per AVDR protocol, it requires 8 to 10 hops to deliver the packet.
are traveling through the shortest possible path. The 3D anchors bypass Hence, 34,912 packets are transmitted over the network to deliver
the void boundary traversal and provide direct shortest path between these 3588 packets with an average of 9.73 transmissions per packet
source and sink. Some initial packets travel through longer path during generated. The transmission ratio is reduced to 9.11 on 24 source
anchor discovery process so it is slightly higher than 1.0. The A3DR and nodes. It happened because anchor discovery and anchor advertisement

10
N.K. Gupta et al. Computer Networks 227 (2023) 109691

Fig. 10. Effect of number of source nodes on end-to-end delay.

Fig. 11. Effect of void radius size on packet delivery ratio.

is one time process and independent of number of sources. There is very

small additional traffic reported during initialization process. A3DR
protocol uses void boundary nodes without any periodic communica-
tion overhead. For three sources, the A3DR protocol transmitted 43,995
packets to the network with packet transmission ratio of 12.26. This
ratio becomes 12.81 on 24 source nodes. Here, trend is growing slowly
because number of re-transmissions is increasing due to congestion.
In case of EDGR protocol, when number of sources is set to 3,
it generated a total of 175,849 packets which included 42,813 data
packets and 100,000 beacons in the simulation time with the rate of 1
beacon per 3 s per node. Additionally, 33,036 are extra transmissions
due to anchor discovery, anchor list preparation and re-transmissions
etc. It generated 49.05 average transmissions per source data packets
during entire communication. When number of sources becomes 24
then packet transmission ratio decreases to 20.16 because beacons
are fixed. GPSR-3D protocol also follows the same style as EDGR but
unlike EDGR, it does not rely on creating and maintaining anchors.
Therefore, it has relatively smaller overhead. BSMH protocol uses too
many control packets (HELLO packets) exchanged to construct the void Fig. 12. Effect of void radius size on routing stretch.
boundary. The construction and maintenance of void boundary results
in excessive overhead. During the initialization phase, the packets get
congested and dropped in midway which causes re-transmissions. Ac- packets. Additionally, BSMH protocol suffers from high initial setup
cording to simulation results available in Fig. 9, AVDR has an average delay. Fig. 10 infers that AVDR protocol is performing better than
25.24%, 65.93%, 41.14% and 72.68% less overhead than A3DR, EDGR, other existing protocols in context of end-to-end delay with respect
GPSR-3D and BSMH, respectively. to varying numbers of source nodes and its performance is 11.03%,
The end-to-end delay consists of propagation delay, network con- 17.84%, 15.08% & 22.08% improved over A3DR, EDGR, GPSR-3D &
gestion delay and processing time. The processing time is fixed and BSMH protocols respectively.
propagation delay and processing time are affected by route length. Here, it can be concluded that AVDR protocol is performing better
The network congestion depends on number of sources as well as over- than other protocols. It has very less impact of increasing number
head. The effect on end-to-end delay is represented in Fig. 10 where of source nodes on performing parameters delivery ratio, end-to-end
results considered the delay of delivered packets only. Here, network delay, stretch and network overhead.
congestion increases with increasing the number of source nodes. In
case of AVDR protocol, the path length is minimized because of anchor 5.3. Void volume scenario
nodes. The congestion delay is also minimum which is slightly growing
with number of source nodes. As per simulation results, it has been In this subsection, we have evaluated the effect of increasing void
observed that each source demands about 9 to 10 hops to deliver the volume size. We fixed the number of source nodes 10, packet rate 4
packet and an average one hop travel time is about 5 ms with minor packets per second per source and manually set a void with various
congestion or queuing delay. Further, the average per hop travel time is sizes (radius) from 50 to 120 m in center of the network. The packet
slightly increasing with respect to number of source nodes. Therefore, takes longer path to deliver at sink node as void size increases. Unlike
end-to-end delay varies from 46 ms to 53 ms on variation of number Section 5.2, here we have considered spherical void structure to check
of source nodes. The A3DR protocol has to travel through boundary of the effect of various void shapes. Further, performance of a protocol
the void so it requires an average of 12 hops to deliver the packet, depends on void type and behavior of recovery policy. The effect of
which results in higher end-to-end delay. On the other hand GPSR- varying void size is shown in Figs. 11–14.
3D, EDGR and BSMH protocol suffers from higher congestion delay The void recovery from spherical structure is relatively easier be-
because of overhead of periodic beacons, longer path and other control cause routing path is smoother than other void structures. Fig. 11

11
N.K. Gupta et al. Computer Networks 227 (2023) 109691

it is able to conquer the shortest possible path. It has been observed

that AVDR is offering 11.83%, 7.63%, 10.66% and 18.2% improvement
over A3DR, EDGR, GPSR-3D and BSMH protocols, respectively.
The network overhead is calculated in terms of ratio of total packets
transmitted in the network during the simulation on total data packets
generated from different sources. The path length is always increased
on expanding the void size, which results more packet transmissions.
Apart from the path length, the control packets also affect the overhead.
This effect is depicted in Fig. 13. The plot shows the ratio for proposed
AVDR protocol and its comparison with existing A3DR, EDGR, GPSR-3D
and BSMH protocols where proposed protocol offers 17.95%, 67.26%,
49.6% and 75.01% improvement, respectively.
The increment in void size reflects longer path which results greater
delay. The condition becomes worse for the protocol, which follows the
void boundary. Additionally, boundary nodes get congested because
overall traffic passes through the void boundary, which creates extra
end-to-end delay. The timing performance of AVDR protocol along with
other existing protocols can be seen in Fig. 14. The results indicate
that anchor based protocols are able to follow smaller path lengths, but
Fig. 13. Effect of void radius size on network overhead.
network congestion may affect overall delivery delay. As per simulation
results, the AVDR protocol produced 36.18 ms delay when void radius
was 50 meters and increased to 46.57 ms on void radius of 120 m.
The A3DR is following boundary of void and producing end-to-end
delay of 40.54 ms to 56.12s ms when void radius increases from 50
to 120 m. The EDGR protocol has shorter path length but suffered
from congestion delay because of limited anchors and periodic control
packets. Thus, AVDR is producing an average of 19.42%, 14.19%,
22.68% and 25.37% less end-to-end delay over EDGR, A3DR, GPSR-3D
and BSMH protocols respectively when void radius length is increasing
from 50 to 120 m.
Here, it can be seen that AVDR protocol is outperforming other
existing approaches. It is working smoothly and has relatively smaller
impact of increasing void volume on stated parameters.

5.4. Multiple void scenario

We have analyzed the simulation results calculated on single void in

the network. The network may have multiple voids in two conditions:
(i) multiple separated voids (ii) consecutive overlapped voids. Consid-
Fig. 14. Effect of void radius size on end-to-end delay. ering both of the cases, we have created a new simulation scenario
where 10 source nodes are transmitting the packets to the sink node
with packet rate of 4 packets per second per source. Two consecutive
demonstrates the packet delivery ratio where packets from different spherical voids are placed inside the network. The distance between
source nodes are sent to the sink node. While void size increases, the centers of both voids is 100 m. The radius of void is varying between
remaining area of the network becomes denser as number of nodes 30 and 100 m. When void radius is 30, 40 or less than 50 m, the voids
is fixed. Therefore, each packet has to cross the void in the network. are separated while void boundary is touching each other on 50 m
Thus, it is clearly seen that the AVDR protocol produces guaranteed radius and overlap on larger radius size. In this way, all three cases
message delivery and there is no impact reported higher void volume. have been tested in this simulation scenario. The performance based on
On the other hand, existing protocols reported a trend of packet loss varying multiple void size is plotted in Figs. 15–18. Fig. 15 displays the
on increasing the void volume. A3DR and GPSR-3D protocols offer a packet delivery ratio of proposed protocol and other existing protocols
good delivery ratio facing a slight packet loss on increasing void size. with different void sizes in overlapping and non-overlapping mode. The
A3DR and GPSR protocols lose some packets during initial phase. AVDR delivery ratio is decreasing on larger size of void radius in case of non-
protocol has achieved an average of 4.1% and 11.9% improvement overlapping mode. On the other hand, overlapping void depicts single
over EDGR and BSMH protocols, respectively. The larger void zone view of void where delivery ratio is significantly increasing. Further,
causes longer path length, which results packet loss due to expiry of after certain points, it decreases with the effect of very larger voids.
lifetime of the packet. The expansion in network density along with This trend is followed by all of the approaches but variation depends
periodic beaconing increases the chance of collision, which results high on nature of protocol. The proposed AVDR protocol has minimal impact
packet loss in EDGR and BSMH protocols. On applying BSMH protocol, of such diversified conditions and able to maximize the delivery ratio.
additional losses are reported due to heavy congestion. Further, the The main reasons for packet drops are congestion due to (i) peri-
boundary discovery process in spherical void is relatively costlier than odic/setup control messages and (ii) tendency of following the same
cubical void shapes. path. The proposed protocol poses freedom from both of the problems.
In general, the hop distance increases with increasing the void size It does not rely on huge control packets and also divides the traffic
which affects routing stretch. As per results depicted in Fig. 12, it is among four different anchors. Thus, data collisions decrease to achieve
clearly shown that AVDR is producing minimum routing stretch among a better packet delivery ratio. The results indicate that the packet
existing geographical routing protocols. After completion of the 300 s of delivery ratio of AVDR fluctuates comparatively less than A3DR, EDGR,
simulation, it varies between 1.07 and 1.12 for AVDR protocol because GPSR-3D and BSMH protocols. It exploits separated paths, which are

12
N.K. Gupta et al. Computer Networks 227 (2023) 109691

Fig. 15. Effect of radius size of two consecutive voids on packet delivery ratio. Fig. 17. Effect of radius size of two consecutive voids on network overhead.

Fig. 16. Effect of radius size of two consecutive voids on routing stretch.
Fig. 18. Effect of radius size of two consecutive voids on end-to-end delay.

passing through different regions of the voids, to deliver data with paths, which cause more packet transmissions per source packet. Fur-
fewer collisions. Compared to A3DR, EDGR, GPSR-3D and BSMH, the ther, overhead of periodic control messages also plays a vital role in
proposed approach has an average of 1.61%, 8.25%, 5.72% and 16.01% generating the overhead. The proposed protocol follows a shorter path
higher packet delivery ratio, respectively, due to less data collisions. via anchor nodes and it does not consider any periodic control message.
When void radius size increases, it transforms from non-overlapping The average packet overhead for different void sizes is less in our
to overlapping form. It creates a type of deep valley structure by proposed scheme because other protocols either follow void boundary
combining two voids in overlapped manner (as shown in Fig. 2). or suffer from control message transmissions. This effect can be seen in
In the results, routing stretches are computed based on 2400 routes figure where our proposed scheme performs 16.27%, 36.06%, 29.49%
between each source and to the sink node. We have observed that and 39.11% better than A3DR, EDGR, GPSR-3D and BSMH protocols
the routing stretch of AVDR is the lowest and slightly lower than that respectively, in terms of average packets transmitted per source data
of EDGR for every void size. The boundary based protocols traverse packet.
the entire boundary and follow a longer path around the boundary On increasing the void size, the packets have to follow the longer
whereas anchor based protocols directly connect the extreme nodes path, which increases end-to-end delay. This fact is supported by results
of different voids. The AVHR and EDGR protocols are anchor based plotted in Fig. 18. The figure indicates the delivery delay of AVDR,
hence reported good routing stretch. Further, EDGR relies on 2D based A3DR, EDGR, GPSR-3D and BSMH with two consecutive voids on
structure, hence, it may not cover optimal paths in real 3D networks. different sizes. The results show that the end-to-end delay of proposed
In comparison, the average routing stretch in AVDR is substantially approach is 12.04%, 10.60%, 21.34% and 28.09% lower than A3DR,
higher, especially for longer voids radius. Fig. 16 shows that the routing EDGR, GPSR-3D and BSMH protocols, respectively, as the voids size
stretch of A3DR protocol is very high because of traversal of void increases. The AVDR and EDGR are anchor based protocols and min-
boundary and it is highest on radius of 50 m. The overall performance imize path length, which causes shorter end-to-end delay. The EDGR
of AVDR is improved by 18.52% from A3DR, 6.24% from EDGR, projects 3D network to 2D so, due to posture errors. It does not always
20.30% from GPSR-3D and 31.99% from BSMH protocols. find an optimal anchor node. As a result, it could not minimize delivery
Fig. 17 shows the effect of different sizes of various voids on delays. On the other hand, the A3DR follows void boundary and suffers
network overhead. The boundary traversing protocols follow longer from high delivery delay. Here, we can observe that there is a sudden

13
N.K. Gupta et al. Computer Networks 227 (2023) 109691

increase in end-to-end delay of A3DR protocol when void radius is Table 3

Variations in the packet delivery ratio, end-to-end delay, routing stretch, and overhead
moving from 40 m to 50 m. This happened because of beginning of the
with varying one source–sink pair over a real field testbed.
void overlapping, which creates deep valley type structure for packet
Protocol used Packet End-to- Routing Network
traversing. Additionally, A3DR protocol does not generate any periodic delivery end delay stretch overhead
overhead, so it delivers data with minimum delay. When voids radius ratio (%) (ms) (total
increases continuously, then we can visualize these voids as single one. transmis-
In this case, there will be less difference in anchor and non anchor based sions by
all nodes)
approaches and majority part of delay will be affected by network
congestion. This effect is depicted beyond radius 80 m in Fig. 18. AVDR (proposed) 100 34.14 1.14 5147
A3DR [41] 99.33 38.27 1.43 5873
In the simulation, we tested the proposal on different network
EDGR [11] 97.83 40.52 1.29 18421
structures under various network conditions. It is important to note GPSR-3D [29] 96.82 41.48 1.35 13277
that a delay of 36.18 ms is observed using proposed method when void BSMH [33] 93.81 45.47 1.49 35498
radius is 30 m and it increases to 46.57 ms on void radius of 100 m. RPL [39,40] 98.16 45.81 1.51 21746
Hence, it is concluded that the end-to-end delay using proposed method
is approximately 14.18%, 19.42%, 22.68% and 25.37% less than A3DR,
EDGR, GPSR-3D and BSMH protocols, respectively. Moreover, it is also
observed that proposed approach claims an improvement of 11.83%,
7.63%, 10.66% and 18.20% in routing stretch and 17.95%, 67.26%,
49.60% and 75.08% in network overhead over A3DR, EDGR, GPSR-
3D, and BSMH protocols, respectively while maximizing the delivery of
data. In comparison with existing protocols, we found that the overall
performance of proposed AVDR protocol is better than related protocols
on given parameters.

5.5. Real field testbed results

The results are also obtained using real field data set [12]. The de-
tails of real field data set in form of testbed are available at FIT-IoT Lab,
Strasbourg, France. In this work, the data set has been generated using
the information available in [12,45]. It utilizes IPV6 based addressing
scheme and routing protocol for low power and lossy networks (RPL)
as baseline protocol [39,40,46]. We have considered 88, IoT devices for
experimental work deployed over 120 m × 100 m × 30 m of volume in
ICube building at Strasbourg. A detailed description of the real testbed
used in this work is given in [47]. All of the IoT devices use Arduino-
compatible microcontroller platform architecture Zigduino board, and
IEEE 802.15.4 radio chip ATmega128RFA1. For simplicity, the nodes’
actual 3D locations are normalized to the network dimensions such
that x, y, z coordinates of all nodes lie between (0, 0, 0) and (120,
100, 30). The experimental work is carried out under two different
test cases. The first test case is depicted in Fig. 2 where node 74 is
considered as source and node 47 is the sink. The first test case focuses
on void creation and anchor discovery process. However, the second
test case considers the varying number of sources while sink is fixed
at node 74, shown in Fig. 19. This test case illustrates the anchor
advertisement process and represents the effect of varying sources on
network performance. Similar to the simulation work, the experimental
results are also obtained in terms of packet delivery ratio, end-to-end
delay, routing stretch and routing overhead. The numerical values of
the experimental results are presented in Tables 3 and 4. Each source
generates 2 packets per second and results are collected for 300 s of
data transmission.
In first test case, 3D coordinates of sink node 47 are fixed at one
end at (111.21, 75.36, 12.6) of the network, whereas source node 74
is configured at another end with 3D coordinates (8.9, 11.54, 21.20).
The source node initiates the data packet and forwards it towards the Fig. 19. An illustration of network scenarios for FIT-IoT lab dataset [12] with (a) void
and (b) anchors under various combination of sources existing in the network.
sink node. Initially, packet follows the greedy approach and passes
through nodes 85, 84, 37 and then stuck at node 19. However, node
19 does not have any neighbor nearer to the sink node, it creates void
in the network and node 19 becomes a void node. This scenario is discovery process, source node approaches anchors rather than sink
depicted in Fig. 2(a). At this point, proposed method starts anchor node. Further, anchor node forwards the packet to the sink node. This
discovery process and packet follows the void boundary nodes. The procedure is illustrated in Fig. 2(b).
boundary nodes compute orthogonal distance over line connecting void The source node generates 2 packets per second and total com-
node and sink node. A node with longest orthogonal in a specified munication duration is 300 s. Initial communication for one second
direction becomes an anchor node. Following such procedure, node 81 time is ignored for anchor discovery process hence, we collected the
and node 18 are declared as anchor nodes. On completing the anchor data of 299 s. A total of 598 packets were initiated from the source

14
N.K. Gupta et al. Computer Networks 227 (2023) 109691

Table 4
Variations in the packet delivery ratio, end-to-end delay, routing stretch, and overhead with varying number of source nodes over a real field
testbed.
Protocol used Number of source Packet delivery End-to-end delay Routing stretch Network overhead
nodes ratio (%) (ms) (number of packets
per transmission)
1 99.33 26.49 1.14 5896
2 98.58 27.16 1.17 9925
3 98.49 28.55 1.19 14098
AVDR (proposed)
4 98.45 29.68 1.21 17837
5 98.36 31.09 1.22 21825
6 98.24 31.57 1.22 25959
1 98.83 30.01 1.28 6002
2 98.16 30.94 1.32 12793
3 97.94 31.52 1.35 19278
A3DR [41]
4 97.58 32.71 1.38 26021
5 97.39 33.86 1.42 32964
6 97.24 35.15 1.45 39913
1 97.32 31.07 1.21 20522
2 95.48 32.14 1.22 28630
3 94.26 33.24 1.24 34785
EDGR [11]
4 93.39 34.81 1.26 40742
5 92.31 36.43 1.29 46818
6 91.44 38.25 1.32 53058
1 96.15 31.21 1.27 14525
2 95.57 31.78 1.27 23118
3 95.04 32.92 1.27 29995
GPSR-3D [29]
4 94.44 33.79 1.28 35301
5 94.15 35.11 1.28 42516
6 94.90 36.84 1.29 49741
1 93.48 32.41 1.34 37917
2 92.64 33.42 1.35 55629
3 90.47 35.35 1.39 61932
BSMH [33]
4 89.30 37.68 1.43 69943
5 87.69 40.59 1.46 79296
6 86.68 43.74 1.52 86274
1 97.83 33.47 1.29 22354
2 95.74 34.12 1.31 29415
3 94.70 37.04 1.34 34786
RPL [39,40]
4 93.77 39.36 1.38 41251
5 92.71 42.59 1.43 48637
6 91.89 45.81 1.47 55972

node. The proposed method is able to deliver all of the packets at the sources get the information of available anchors. Now, all the sources
sink node whereas other methods have certain packet loss. Further, utilize the anchor information to find out direct route from source
the proposed method delivers the packet after going through 8 hops to anchor and anchor to the sink node. This scenario is illustrated
but using manual process, data can be delivered in 7 hops. Hence, in Fig. 19(b) where network void does not affect the communication
average routing stretch of the proposed method is 1.14 (i.e., 8/7). Other between source and sink.
existing methods are delivering data using more than 8 hops and have In this test case, the proposed method achieves more than 99%
bigger value of routing stretch. Moreover, path length also affects the delivery ratio in the best case and it is more than 98% when number
end-to-end delay. The average time consumed per hop is about 4 ms. of source nodes increases to 6. Here, a few packets are lost in the
Hence, proposed method consumes an average of 34.14 ms of time to beginning due to anchor discovery and anchor advertisement process.
deliver the packet at sink node. This delay is smaller than other existing However, it is noted that the performance of the proposed method
routing protocols. Additionally, proposed method consumes very less increases by 0.74%, 4.83% 3.90%, 9.48% and 4.38% when compared
number of additional transmissions for data delivery. Thus, the network to A3DR, EDGR, GPSR-3D, BSMH, and RPL protocols, respectively.
overhead is minimized. The beacon based methods suffer from network In addition to the delivery ratio, it is also noted that the proposed
overhead. So, it can be concluded that proposed method outperforms AVDR protocol has reported the least average end to end delay when
existing geographical routing protocols in the first test case. compared to other existing geographical routing protocols. The pro-
The second test case illustrates the effect of varying number of posed method has an average end-to-end delay about 31 ms when
source nodes on routing performance. This test case is depicted in all sources are transferring the data simultaneously. However, the
Fig. 19(a). In this test case, 3D coordinates of sink node 74 are fixed average end-to-end delay in case of conventional A3DR, EDGR, GPSR-
at one end at (8.9, 11.54, 21.20) of the network where as six source 3D, BSMH and RPL routing protocols is greater than 35 ms under
nodes are configured at another end of the network. A void is available similar conditions. The reason for the improved end-to-end delay of the
between source and sink such that each source packet has to pass proposed method is due to its ability to select the direct path through
through void. Initially, node 50 initiates the data packet and is stuck at anchor nodes during data delivery to the sink node. It is lacking in the
void node 28 in the route of the sink node while following the greedy conventional geographical routing and existing two dimensional based
approach. This can be seen in Fig. 19(a). The anchor discovery process anchor discovery methods. Recall that the proposed method takes into
starts at this point. The discovery process discovers node 7 and node 17 account path length as one of the parameters and minimizes the data
as anchor nodes for upcoming communications. Further, the location transmission delay during the data transfer. The proposed method ob-
of these two anchors is advertised to the other nodes in scope of tains 10.12%, 15.25%, 13.44%, 21.80% and 24.89% improved results
these anchors. Here, anchor advertisement process completes and other over existing A3DR, EDGR, GPSR-3D, BSMH and RPL methods.

15
N.K. Gupta et al. Computer Networks 227 (2023) 109691

The average routing stretch for all the packet transmissions over been considered on various void conditions in the network. The pro-
the testbed is also given in Table 4. A closer examination of the posed scheme is efficiently able to detour the network void and follows
experimental results shows that the proposed method achieves routing smaller path in shorter duration. Thus, it has been proven that the
stretch of 1.14 in the best case. This is possible in case of less number experimental results of the proposed AVDR protocol is outperforming
of source nodes as very less traffic is generated in the network. When other existing anchor-based and non anchor-based geographical routing
number of sources increases, the respective traffic also increases which schemes.
degrades the path length performance. Also, the average routing stretch
is less than 1.22 in all other cases using proposed method. Hence, CRediT authorship contribution statement
the proposed method is able to deliver the data using the shortest
path length when compared to other existing methods. The proposed The authors are obliged to Digital India Corporation (formerly
method achieves more than 5% and 12% average improvement over media lab asia) under the Ministry of Electronics and Information
existing anchor based and traditional non-anchor based approaches, Technology, Govt. of India, for providing remarkable set-up in Depart-
respectively. ment of Computer Sc. & Engg., Motilal Nehru National Institute of
This study also considers the effect of varying source nodes on Technology Allahabad, Prayagraj, 211004, India.
the network overhead. The overhead increases due to additional pack-
ets e.g., beacons, control packets and re-transmissions. This effect is Declaration of competing interest
illustrated in Table 4. A total of 598 packets are sent from source
when only one source is generating the data packets. Existing protocols The authors declare that they have no known competing finan-
EDGR, GPSR-3D, BSMH and RPL protocols require more than 14,500 cial interests or personal relationships that could have appeared to
transmissions to deliver these packets. On the other hand, A3DR and influence the work reported in this paper.
proposed AVDR protocol require near to 6000 transmissions. When
number of source nodes increases to 6 then total 3588 packets gen- Data availability
erate for data delivery. In this case, proposed method utilizes 25,959
transmissions but non-anchor based methods needed more than 39,700 The authors are unable or have chosen not to specify which data
transmissions due to longer path. On the other hand, the beacon based has been used.
methods require more than 53,000 transmissions including beacons,
control packets and re-transmissions. Further, it is noted that the net- References
work overhead analysis is also important because it affects the average
[1] J. Marietta, B.C. Mohan, A review on routing in internet of things, Wirel. Pers.
energy consumption. Each transmission and reception consume energy
Commun. 111 (1) (2020) 209–233.
of IoT devices. Therefore, the discussion on energy consumption and [2] O.J. Pandey, R.M. Hegde, Low-latency and energy-balanced data transmission
its comparative analysis are given in next paragraph. over cognitive small world WSN, IEEE Trans. Veh. Technol. 67 (8) (2018)
As part of energy consumption analysis using different methods, 7719–7733.
[3] M.-S. Pan, S.-W. Yang, A lightweight and distributed geographic multicast routing
BSMH method reports higher energy consumption and lesser network
protocol for IoT applications, Comput. Netw. 112 (2017) 95–107.
lifetime. This method wastes a lot of energy to deal with network [4] H. Huang, H. Yin, Y. Luo, X. Zhang, G. Min, Q. Fan, Three-dimensional
overhead due to packet loss, re-transmissions and longer paths. The geographic routing in wireless mobile ad hoc and sensor networks, IEEE Netw.
RPL method also utilizes more energy but less than the BSMH. Further, 30 (2) (2016) 82–90.
EDGR and GPSR-3D methods consume relatively less energy because [5] N.K. Gupta, R.S. Yadav, R.K. Nagaria, 3D geographical routing protocols in
wireless ad hoc and sensor networks: an overview, Wirel. Netw. 26 (4) (2020)
these methods are anchor-based methods and require smaller paths
2549–2566.
during void conditions. A3DR also consumes moderate energy be- [6] O.J. Pandey, A. Mahajan, R.M. Hegde, Joint localization and data gathering over
cause it does not rely on periodic beacons but follows a longer path a small-world WSN with optimal data mule allocation, IEEE Trans. Veh. Technol.
around the boundary of the void. The path length, overhead, and losses 67 (7) (2018) 6518–6532.
[7] S.F.A. Rubeaai, M.A. Abd, B.K. Singh, K.E. Tepe, 3D real-time routing protocol
are minimized in the proposed novel AVDR method. As a result, the
with tunable parameters for wireless sensor networks, IEEE Sens. J. 16 (3) (2016)
novel AVDR method performs better than existing geographical routing 843–853.
methods. Hence, it can be treated as an energy-efficient one. [8] A.E. Abdallah, Low overhead hybrid geographic-based routing algorithms with
smart partial flooding for 3D ad hoc networks, J. Ambient Intell. Humaniz.
Comput. 9 (1) (2018) 85–94.
6. Conclusion
[9] K. Redjimi, M. Boulaiche, M. Redjimi, IEGGR: Improved energy-aware and
delivery guarantee geographic routing, J. Commun. Softw. Syst. 12 (2) (2021)
In this paper, a novel anchor-based void detouring routing (AVDR) 143–153.
method is proposed to detour the void with reduced path length and [10] M. Król, E. Schiller, F. Rousseau, A. Duda, WEAVE: Efficient geographical routing
in large-scale networks, in: EWSN, 2016, pp. 89–100.
improved data transmission delay. The method utilizes anchor nodes
[11] H. Huang, H. Yin, G. Min, J. Zhang, Y. Wu, X. Zhang, Energy-aware dual-path
towards routing of data packets in a 3D IoT network. The anchor node geographic routing to bypass routing holes in wireless sensor networks, IEEE
selection is based on orthogonal distance between anchor candidate Trans. Mob. Comput. 17 (6) (2018) 1339–1352.
and line joining between void node and sink node. The anchor nodes [12] C. Adjih, E. Baccelli, E. Fleury, G. Harter, N. Mitton, T. Noel, R. Pissard-
Gibollet, F. Saint-Marcel, G. Schreiner, J. Vandaele, et al., FIT IoT-LAB: A
are judiciously distributed to other nodes lying inside the scope of an-
large scale open experimental IoT testbed, in: 2015 IEEE 2nd World Forum on
chor nodes. Therefore, all source nodes do not require anchor discovery Internet of Things, WF-IoT, IEEE, 2015, pp. 459–464, URL: https://www.iot-
process again and again which helps in minimizing the overall network lab.info/docs/deployment/strasbourg/.
overhead. The performance of the proposed method has been illustrated [13] K. Cai, Z. Yin, H. Jiang, G. Tan, P. Guo, C. Wang, B. Li, OnionMap: A scalable
with diagrams, mathematical analysis, examples, test cases, simulated geometric addressing and routing scheme for 3D sensor networks, IEEE Trans.
Wireless Commun. 14 (1) (2015) 57–68.
and real testbed results. [14] I. Ghareeb, M. Yahya, Three-dimensional wireless ad hoc networks with random
The simulation results are obtained using INET framework under nodes distribution, Int. J. Electron. (2020).
OMNET++ simulator. The performance of the proposed method is also [15] F. Yu, S. Pan, G. Hu, Hole plastic scheme for geographic routing in wireless
validated using real field datasets in different test cases. The results sensor networks, in: 2015 IEEE International Conference on Communications,
ICC, IEEE, 2015, pp. 6444–6449.
illustrate that the AVDR protocol efficiently enhances the performance
[16] B.-H. Liu, V.-T. Pham, B.-Y. Hou, S.-W. Chiu, Virtual-coordinate-based delivery-
of packet delivery ratio, path length in terms of routing stretch, network guaranteed routing protocol in three-dimensional wireless sensor networks,
overhead and end-to-end delay. Multiple experimental scenarios have Wirel. Commun. Mob. Comput. 15 (2) (2015) 215–227.

16
N.K. Gupta et al. Computer Networks 227 (2023) 109691

[17] S.M. Ghoreyshi, A. Shahrabi, T. Boutaleb, Void-handling techniques for routing [46] J. Nassar, M. Berthomé, J. Dubrulle, N. Gouvy, N. Mitton, B. Quoitin, Multiple
protocols in underwater sensor networks: Survey and challenges, IEEE Commun. instances QoS routing in RPL: Application to smart grids, Sensors 18 (8) (2018)
Surv. Tutor. 19 (2) (2017) 800–827. 2472.
[18] H. Jodeh, A. Mikkawi, A. Awad, O. Othman, Comparative analysis of routing [47] FIT-IoT Lab real field testbed data, 2022, URL: https://github.com/
protocols for under-water wireless sensor networks, in: Proceedings of the 2nd naveengupta119/avdr. (Online accessed 13 October 2022).
International Conference on Future Networks and Distributed Systems, ACM,
2018, p. 33.
[19] B. Karp, H.-T. Kung, GPSR: Greedy perimeter stateless routing for wireless
networks, in: Proceedings of the 6th Annual International Conference on Mobile Naveen Kumar Gupta is currently working as Assistant
Computing and Networking, ACM, 2000, pp. 243–254. Professor in Jaypee Institute of Information Technology,
[20] B. Leong, B. Liskov, R. Morris, Geographic routing without planarization, in: Noida, India. He completed his doctorate from Motilal
NSDI, Vol. 6, 2006, p. 25. Nehru National Institute of Technology Allahabad, Praya-
[21] Y. Mao, F. Wang, L. Qiu, S.S. Lam, J.M. Smith, S4: Small state and small stretch graj, India in 2020. He received his B.Tech. degree in
routing protocol for large wireless sensor networks, in: NSDI, 2007. Computer Science and Engineering from Inderprastha En-
[22] P. Le Nguyen, Y. Ji, K. Le, T.-H. Nguyen, Load balanced and constant stretch gineering College Ghaziabad, India in 2011 and M.Tech.
routing in the vicinity of holes in wsns, in: 2018 15th IEEE Annual Consumer in Information Technology from Madan Mohan Malaviya
Communications & Networking Conference, CCNC, IEEE, 2018, pp. 1–6. University of Technology Gorakhpur, India in 2014. He
[23] P. Bose, P. Morin, I. Stojmenović, J. Urrutia, Routing with guaranteed delivery has two years of teaching experience. He has published
in ad hoc wireless networks, Wirel. Netw. 7 (6) (2001) 609–616. various research papers in reputed international journals
[24] B. Leong, S. Mitra, B. Liskov, Path vector face routing: Geographic routing and conferences. His main research interest lies in Wire-
with local face information, in: Network Protocols, 2005. ICNP 2005. 13th IEEE less Sensor Networks, IoT, Ad hoc Routing protocols and
International Conference on, IEEE, 2005, pp. 12–pp. Blockchain Technology. He served as a reviewer for various
[25] F. Kuhn, R. Wattenhofer, A. Zollinger, Asymptotically optimal geometric mobile reputed journals including international journal of com-
ad-hoc routing, in: Proceedings of the 6th International Workshop on Discrete munication systems, computer communications, journal of
Algorithms and Methods for Mobile Computing and Communications, ACM, supercomputing, ad hoc & sensor wireless networks.
2002, pp. 24–33.
[26] A.E. Abdallah, T. Fevens, J. Opatrny, High delivery rate position-based routing
Rama Shankar Yadav is currently a professor at Motilal
algorithms for 3D ad hoc networks, Comput. Commun. 31 (4) (2008) 807–817.
Nehru National Institute of Technology, Allahabad, India.
[27] J. Zhou, Y. Chen, B. Leong, P.S. Sundaramoorthy, Practical 3D geographic routing
He received his Ph.D. degree from the Indian Institute
for wireless sensor networks, in: Proceedings of the 8th ACM Conference on
of Technology (IIT), M.S. degree from Birla Institute of
Embedded Networked Sensor Systems, ACM, 2010, pp. 337–350.
Technology and Science (BITS) Pilani, and B.Tech. degree
[28] S.S. Lam, C. Qian, Geographic routing in d-dimensional spaces with guaranteed
from the Institute of Engineering and Technology (IET), Luc-
delivery and low stretch, IEEE/ACM Trans. Netw. 21 (2) (2013) 663–677.
know, India. Dr. Yadav has extensive research and academic
[29] J. Fu, B. Cui, N. Wang, X. Liu, A distributed position-based routing algorithm
experience. He has worked in leading institutions such as
in 3-D wireless industrial internet of things, IEEE Trans. Ind. Inform. 15 (10)
Govind Ballabh Pant Engineering College (GBPEC), Pauri,
(2019) 5664–5673.
Garhwal, and Birla Technical Training Institute (BTTI),
[30] F. Cadger, K. Curran, J. Santos, S. Moffett, A survey of geographical routing in
Pilani. He has authored more than 70 research papers in na-
wireless ad-hoc networks, IEEE Commun. Surv. Tutor. 15 (2) (2013) 621–653.
tional/international conferences, refereed journals, and book
[31] D.-G. Zhang, P.-Z. Zhao, Y.-y. Cui, L. Chen, T. Zhang, H. Wu, A new method of
chapters. Dr. Yadav’s areas of interest are real time systems,
mobile ad hoc network routing based on greed forwarding improvement strategy,
embedded systems, fault-tolerant systems, energy aware
IEEE Access 7 (2019) 158514–158524.
scheduling, network survivability, computer architecture,
[32] M. Heissenbüttel, T. Braun, T. Bernoulli, M. WäLchli, BLR: beacon-less routing
distributed computing, and cryptography.
algorithm for mobile ad hoc networks, Comput. Commun. 27 (11) (2004)
1076–1086.
[33] P.L. Nguyen, T.H. Nguyen, K. Nguyen, A path-length efficient, low-overhead,
Rajendra Kumar Nagaria received B.Tech. and M.Tech.
load-balanced routing protocol for maximum network lifetime in wireless sensor
degree in electronics engineering from KNIT Sultanpur,
networks with holes, Sensors 20 (9) (2020) 2506.
India in 1988 and 1996 respectively and the Ph.D.(Engg.)
[34] H. Huang, G. Hu, F. Yu, Energy-aware multipath geographic routing for detouring
degree from Jadhavpur University Kolkata, India in 2004.
mode in wireless sensor networks, Eur. Trans. Telecommun. 22 (7) (2011)
He is currntly working as a full Professor in the department
375–387.
of Electronics & Communication Engineering, Motilal Nehru
[35] G. Han, J. Jiang, C. Zhang, T.Q. Duong, M. Guizani, G.K. Karagiannidis, A survey
National Institute of Technology Allahabad, India since
on mobile anchor node assisted localization in wireless sensor networks, IEEE
January 2009. He has more than 30 years of teaching
Commun. Surv. Tutor. 18 (3) (2016) 2220–2243.
and research experience and contributed more than eighty
[36] S. Kim, T. Yang, C. Kim, H. Cho, S.-H. Kim, Dynamic anchors based void
research articles. He is fellow of professional bodies like The
avoidance scheme for real-time application in wsns, in: 2017 IEEE 86th Vehicular
Institution of Engineers (India), Indian Society for Technical
Technology Conference, VTC-Fall, IEEE, 2017, pp. 1–5.
Education and member of IEEE. His main research interest
[37] H. Huang, G. Hu, F. Yu, Energy-aware geographic routing in wireless sensor
includes Analog/Mixed-mode circuits and Low power VLSI
networks with anchor nodes, Int. J. Commun. Syst. 26 (1) (2013) 100–113.
circuits and systems.
[38] P.-L. Nguyen, Y. Ji, K. Le, T.-H. Nguyen, Routing in the vicinity of multiple
holes in wsns, in: 2018 5th International Conference on Information and
Communication Technologies for Disaster Management, ICT-DM, IEEE, 2018, pp. Deepak Gupta currently working as Assistant professor
1–8. in Faculty of Engineering and Technology (Mathematics)
[39] T. Winter, P. Thubert, A. Brandt, J. Hui, R. Kelsey, P. Levis, K. Pister, R. Struik, at University of Lucknow, Lucknow,UP, India. His Ph.D.
J.-P. Vasseur, R. Alexander, RPL: IPv6 Routing Protocol for Low-Power and Lossy awarded from Bundelkhand University, Jhansi in 2013.
Networks, Technical Report, 2012. He has published research papers in various journal. His
[40] T. Tsvetkov, A. Klein, et al., RPL: IPv6 routing protocol for low power and lossy research area is Queuing theory, optimization techniques
networks, Network 59 (2011) 59–66. and Fuzzy logic.
[41] N.K. Gupta, R.S. Yadav, R.K. Nagaria, D. Gupta, An angular 3D path selection
protocol in wireless sensor networks, Open Comput. Sci. 11 (1) (2021) 190–207.
[42] S. Wu, K.S. Candan, Power-aware single-and multipath geographic routing in Achyut Mani Tripathi received the Bachelor of Engineering
sensor networks, Ad Hoc Netw. 5 (7) (2007) 974–997. in Information Technology from the Chhattisgarh Swami
[43] T.J. Rouphael, Chapter 4 - High-level requirements and link budget analysis, Vivekanand Technical University (CSVTU), Bhilai, Chhattis-
in: T.J. Rouphael (Ed.), RF and Digital Signal Processing for Software-Defined garh, India, in 2012. He obtained his Ph.D. degree from
Radio, Newnes, Burlington, 2009, pp. 87–122, URL: http://www.sciencedirect. the Department of Computer Science & Engineering at the
com/science/article/pii/B9780750682107000047. Indian Institute of Technology Guwahati, Assam, India, in
[44] A. Varga, OMNET++ website, 2019, URL: https://omnetpp.org/. 2021. His research interests include Deep Learning, Knowl-
[45] J. Munoz, F. Rincon, T. Chang, X. Vilajosana, B. Vermeulen, T. Walcarius, edge Distillation, Adversarial Attacks, Environmental Sound
W. Van de Meerssche, T. Watteyne, OpenTestBed: poor man’s IoT testbed, in: Classification and Anomaly Detection in Time Series.
IEEE INFOCOM 2019-IEEE Conference on Computer Communications Workshops,
INFOCOM WKSHPS, IEEE, 2019, pp. 467–471.

17
N.K. Gupta et al. Computer Networks 227 (2023) 109691

Om Jee Pandey received the B.Tech. degree in electronics Engineering College and Research Center, Jaipur, from 2013
and communication engineering from Uttar Pradesh Techni- to 2014. He is currently working as an Assistant Professor
cal University, Lucknow, India, in 2008, the M.Tech. degree with the Department of Electronics Engineering, Indian
in digital communications from the ABV-Indian Institute of Institute of Technology (BHU), Varsnasi, Uttar Pradesh.
Information Technology and Management, Gwalior, India, His research interests include wireless sensor networks,
in 2013, and the Ph.D. degree from the Department of low-power wide-area networks, unmanned aerial vehicle
Electrical Engineering, Indian Institute of Technology Kan- networks, mobile and pervasive computing, cyber physical
pur, Kanpur, India, in 2019. He worked as a Postdoctoral systems and Internet of Things, cloud and fog computing,
Fellow with the Communications Theories Research Group, UAV-assisted optical communications, and social networks.
Department of Electrical and Computer Engineering, Univer- He is a senior member of IEEE and serves as a regular
sity of Saskatchewan, Saskatoon, SK, Canada. From 2008 reviewer for various reputed journals including computer
to 2011, he worked with Escorts Ltd., and FANUC India communications, computer networks, pervasive & mobile
Private Ltd. He was a Senior Lecturer with the Jaipur computing and Ad Hoc Networks.

18