MEASUREMENT SCIENCE REVIEW, 24, (2024), No.

3, 95-99

Journal homepage: https://content.sciendo.com

Improved Grey Wolf Optimization Based Node Localization

Approach in Underwater Wireless Sensor Networks
Salem Jeyaseelan WR1, Vinoth Kumar K2*, Jayasankar T3, Ponni R4
1
Department of IT, PSNA College of Engineering and Technology, Dindigul, India, [email protected]
2
Department of ECE, Vivekanandha College of Engineering for Women, Tiruchengode, India, [email protected]
3
Department of ECE, University College of Engineering (BIT Campus), Trichirappall, India, [email protected]
4
Department of ECE, Kings College of Engineering, Thanjavur, India, [email protected]

Abstract: Underwater Wireless Sensor Networks (UWSNs) are established by Autonomous Underwater Vehicles (AUVs) or static Sensor
Nodes (SN) that collect and transmit information over the underwater environment. Localization plays a vital role in the effective
deployment, navigation and coordination of these nodes for many applications, namely underwater surveillance, underwater exploration,
oceanographic data collection and environmental monitoring. Due to the unique characteristics of underwater transmission and acquisition,
this is a fundamental challenge in underwater networks. However, localization in UWSNs is problematic due to the unique features of
underwater transmission and the harsh underwater environment. To address these challenges, this paper presents an Improved Grey Wolf
Optimization Based Node Localization Approach in UWSN (IGWONL-UWSN) technique. The presented IGWONL-UWSN technique is
inspired by the hunting behavior of grey wolves with the Dimension Learning-based Hunting (DLH) search process. The proposed
IGWONL-UWSN technique uses the Improved Grey Wolf Optimization Based (IGWO) algorithm to calculate the optimal location of the
nodes in the UWSN. Moreover, the IGWONL-UWSN technique incorporates the DLH search process to improve the convergence and
accuracy. The simulation results of the IGWONL-UWSN technique are validated using a set of performance measures. The simulation
results show the improvements of the IGWONL-UWSN method over other approaches with respect to various metrics.

Keywords: Underwater Wireless Sensor Networks (UWSN), localization, sensor nodes, Grey Wolf Optimizer (GWO), localization accuracy.

1. INTRODUCTION an underwater environment is very large. Non-negligible

The concept of Underwater Wireless Sensor Networks node mobility can lead to general fluctuations in the network
(UWSNs) has attracted a lot of attention recently. Underwater topology due to water currents [5]. The tracking and detection
of the intrusion object should rely on the location of the nodes
Sensor Networks (USNs) can be used for a range of
in the marine military defense field [6].
applications. Each implementation is critical in its field [1],
Localization is a problem of estimating node locations and
but some could enhance ocean exploration to fulfill the
can be done globally using altitude, latitude and longitude
number of underwater applications, namely, assisted
data or locally using position data with respect to other nodes
navigation, natural disaster warning systems (e.g., seismic [7]. Location data is required for data tagging because the
and tsunami tracking), ecological applications (e.g., pollution data received from the sink node cannot be recognized
tracking, biological water quality), underwater monitoring, without node location information and becomes worthless for
industrial applications (e.g., marine research), oceanographic the application. Nodes with location data are called beacon or
information collection, and so on [2]. For offshore Anchor Nodes (ANs), while nodes without location data are
engineering applications, for example, sensors could evaluate called ordinary or blind nodes [8]. Furthermore, localization
a number of parameters, namely, mooring tension and base may be required not only for tagging objects, but also for
intensity, to monitor the structural quality of the mooring finding the best routes in geographic routing and for optimal
environment [3]. In addition to the essential features of coverage of an area. Location information can be used to
typical Wireless Sensor Networks (WSNs), such as limited develop effective management and networking protocols [9].
energy and large-scale deployment, UWSNs have certain The nodes can be deployed using an Autonomous
differences from terrestrial systems. First, underwater Underwater Vehicle (AUV) or manually, depending on the
transmission was detected only by the acoustic signal, which area and network size. In manual deployment, the network is
has higher error rates and lower bandwidth [4]. In addition, accessible to humans and the nodes register their location
beacon nodes are sparser and the scale of node deployment in [10].

DOI: 10.2478/msr-2024-0013 *Corresponding author: [email protected] (Vinoth Kumar K)

95
 MEASUREMENT SCIENCE REVIEW, 24, (2024), No. 3, 95-99

Localization Approach in UWSNs (IGWONL-UWSN)

method.
2. RELATED WORKS
A mobile AN-based Received Signal Strength Indicator 3. PROPOSED SYSTEM
(RSSI) localization method in UWSN is presented in [11]. A In this paper, we present a novel IGWONL-UWSN
Support Vector Regression (SVR) related to the interpolation technique to determine the optimal location of SNs in the
approach was developed to predict the prediction of Sensor UWSN. The presented IGWONL-UWSN technique is
Nodes (SNs) on the linear trajectory of mobile ANs. Next, a inspired by the hunting behavior of grey wolves with the
curve matching approach was developed to determine the Dimension Learning-based Hunting (DLH) search process.
perpendicular distances of SNs to the linear trajectory of the The proposed IGWONL-UWSN technique utilizes the
mobile ANs. Finally, the simulation results prove that the IGWO algorithm to calculate the optimal location of nodes in
presented method enables more precise SN localization in the UWSN. Fig. 1 shows the working procedure of the
less time than the current methods. IGWONL-UWSN method, which employs UWSN with 𝑛
A Cross-Layer Protocol with Lower Interference (CLIC) nodes used in the 2D space of 𝑍 2 , and 𝑚 ANs. There are 𝑛 −
and congestion depending on directional reception is 𝑚 unknown nodes. where 𝑚 < 𝑛. The distance of all nodes
presented in [12]. In the CLIC method, a combined routing- to their near neighbors within their ranging distance was
MAC model can be developed to utilize the directional beams evaluated and then a network was used. All effective distance
for creating high capacity and low interference data measurements are transmitted to the Base Station (BS)
transmission links and balance the main aspects affecting the together with the node conditions using multi-hop routing. A
network performance to obtain low congestion and low graph is created and each evaluated distance is transmitted to
collision routes. A Geodesic Search Algorithm (GSA) is the BS. This graph for WSN is modeled as 𝐺 with (𝑉, 𝐸),
related to target localization that minimizes the Localization where 𝑉 and 𝐸 denote the group of vertices and edges. A
Error (LE) using the phase-space constancy of the UWSN to group of SNs is denoted by the vertices 𝑉 with {𝑣1 , 𝑣2 , . . 𝑣𝑛 }.
efficiently triangulate the targeted nodes regardless of their The connection of vertices is denoted as a group of edges 𝐸
mobility. A malicious AN approach was introduced in [13]. with {𝑒1,2 , 𝑒1,3 , . . 𝑒𝑖,𝑗 , . . 𝑒𝑛−1,𝑛 }. If the connected component
A better localization approach for mobile aquaculture of 𝐺, 𝐺1 = (𝑉1 , 𝐸1 ) does not have three or more ANs, then any
WSN related to the Improving Dynamic Population Monte SN from the subgraph 𝐺1 cannot be localizable. It assumes
Carlo Localization (I-DPMCL) method is presented in [14]. that each connected element of the graph 𝐺 has at least 3
According to these localization behaviors, specific delays anchors.
were predicted depending on the statistical point of view. A
performance comparison of I-DPMCL with other Sliding
Mode Control (SMC)-related methods was also presented. In
[15], a precise range-based method was modeled and the need
to utilize the power of SNs expeditiously is a different feature
of underwater WSNs. An improved analysis for underwater
localization is developed by providing a general localization
approach and then installing a normal beacon node to
determine the accuracy and error of sensor localization. The
author presented two localization methods, the angle-based
method and the distance-based method.
A range-free Radial Basis Function Network (RBFN) and
a Kalman Filtering (KF) based technique called RBFN+KF is
presented. Compared to other techniques, the simulation
results show lower location estimation errors [16]. Moreover, Fig. 1. Working process of the IGWONL-UWSN approach.
the RBFN-oriented approach is more energy-efficient than
Multilayer Perceptron (MLP)-based localization and A. Design of the Improved Grey Wolf Optimization Based
trilateration methods. An energy-free Heuristic Neural (IGWO) algorithm
Network (HNN) localization method with Deep Learning
(DL) algorithm for locating the dead Mobile Sensor Nodes The Greedy Wolf Optimization (GWO) algorithm is a
(MSN) in a largescale Underwater Acoustic Sensor Network Swarm Intelligence (SI) optimization technique in which the
(UASN) is presented. The HNN localization achieves high process of finding the global optimum is motivated by the
accuracy and minimum LE compared to the presented DL hunting behavior of the Greedy Wolf (GW) population. There
algorithms [17]. is a strict hierarchy in the GW population, and a few GWs
The problem of localization in the UWSNs poses a great with absolute discourse power guide a population of GWs
challenge due to the unique characteristics of underwater towards the prey. GWs are usually divided into 4 groups: 𝛼,
transmission and the harsh underwater environment. 𝛽, 𝛿, and 𝜔 wolves. The rights are from larger to smaller to
Conventional localization techniques often struggle to simulate the leadership class. Collective hunting is a social
provide accurate and efficient localization solutions under behavior of GWs. It mainly consists of three phases: (1)
such conditions. To overcome these challenges, this paper approaching, harassing and tracking the target; (2) encircling
proposes an Improved Grey Wolf Optimization Based Node and hunting the prey until it stops moving; and (3) attacking

96
 MEASUREMENT SCIENCE REVIEW, 24, (2024), No. 3, 95-99

the target. First, create a mathematical process for the social 𝑋3 = 𝑋𝛿 − 𝐴3 ⋅ (𝐷𝛿 )
hierarchy of the GW and the model of the social hierarchy of
𝑋1 +𝑋2 +𝑋3
GW. The 𝛼 wolf is used as the optimal solution, i.e. the fitness 𝑋𝑧+1 = (7)
3
of the individual is optimal, the suboptimal solution is the 𝛽
wolf, and the fittest solution is the 𝛿 wolf, i.e. the global B. Process involved in the IGWONL-UWSN technique
optimum or the local optimum solution of the main function,
with the minimum or maximum values of the main function. The proposed IGWONL-UWSN technique utilizes the
The remaining candidate solution is called the 𝜔 wolf. The IGWO algorithm to calculate the optimal location of nodes in
hunting strategy is based on the 𝛽, 𝛿, and 𝜔 wolves following the UWSN. The goal of IGWONL-UWSN localization in a
the above three wolves. More specifically, look for the three UWSN is to find the coordinates of 𝑛 − 𝑚 unknown nodes
best solutions first and then search around the area to find the and use the previous data about the locations of 𝑚 ANs. The
best solution and improve the 𝛽 and 𝛿 wolves later. The GW presented main function for node localization contains
strategy of prey hunting can be described using (1): 2 phase processes. The primary one is a ranging system in
which the nodes determine their distances in ANs based on
𝐷 = |𝐶 ⋅ 𝑋𝑝 (𝑧) − 𝑋(𝑧)| (1)
the signal propagation time or RSSI, and the secondary one is
The equation for the position update of GW is given below: the position estimation of the nodes, i.e. using the ranging
data. The LE was minimized by applying optimization
𝑋(𝑧 + 1) = 𝑋𝑝 (𝑧) − 𝐴 ⋅ 𝐷 (2) techniques. Initially, all ANs in the application estimate their
distance to each of their neighboring target nodes. RSSI
Coefficient vector: ranging technology is used to determine the distance between
the nodes.
𝐴 = 2𝛼 ⋅ 𝑟1 − 𝛼 (3)

𝐶 = 2 ⋅ 𝑟2 (4)

From the equation, 𝑋𝑝 stands for the vector prey location,

𝑋 represents the place vector of the GWs, 𝑧 denotes the
number of iterations, 𝐷 shows the distance vector between the
individuals and the hunt, 𝑟1 and 𝑟2 are the random vector
numbers from the interval of zero and one and 𝛼 denotes the
convergence factor (decreases linearly in two to zero with the
iteration number).
GWs could find the location of the prey and encircle it.
Once the GW has identified the location of the prey, it leads
the wolf population to encircle the prey in the guidance of 𝛽
and 𝛿. The tracking of the prey location can be
mathematically modeled as follows:

𝐷𝛼 = |𝐶1 ⋅ 𝑋𝛼 − 𝑋|
Fig. 2. Steps involved in GWO.
𝐷𝛽 = |𝐶2 ⋅ 𝑋𝛽 − 𝑋| (5)
The distance between the unknown nodes 𝑜(𝑥, 𝑦) is
𝐷𝛿 = |𝐶3 ⋅ 𝑋𝛿 − 𝑋| denoted as 𝑑1 , 𝑑2 , 𝑑𝑛 and AN was obtained by the hop count
and the average hop distance between the nodes. The ranging
From the expression, 𝐷𝛼 , 𝐷𝛽 and 𝐷𝛿 signify 𝛼, 𝛽 and 𝛿 and
error is 𝜀1 , 𝜀2 , 𝜀𝑛 , the estimation coordinates (𝑥, 𝑦) satisfy the
the distance to other individuals, respectively; 𝑋𝛼 , 𝑋𝛽 and 𝑋𝛿
following inequalities:
show the existing location of 𝛼, 𝛽 and 𝛿, respectively; 𝐶1 , 𝐶2
and 𝐶3 represent a random vector and X indicates the existing 𝑑12 − 𝜀12 < (𝑥 − 𝑥1 )2 + (𝑦 − 𝑦2 )2 ≤ 𝑑12 + 𝜀12
location of the GW. Equation (6) determines the step length 𝑑22 − 𝜀22 < (𝑥 − 𝑥2 )2 + (𝑦 − 𝑦2 )2 ≤ 𝑑22 + 𝜀22
and direction of 𝜔 individuals from the wolf pack near 𝛼, 𝛽 .... (8)
and 𝛿, and (7) describes the final location of 𝜔. The steps of ....
GWO are shown in Fig. 2. {𝑑𝑛2 − 𝜀𝑛2 < (𝑥 − 𝑥𝑛 )2 + (𝑦 − 𝑦𝑛 )2 ≤ 𝑑𝑛2 + 𝜀𝑛2

𝑋1 = 𝑋𝛼 − 𝐴1 ⋅ (𝐷𝛼 ) where 𝑑 refers to the actual distance between 2 nodes and 𝜀

denotes the ranging error. The localization problem has been
𝑋2 = 𝑋𝛽 − 𝐴2 ⋅ (𝐷𝛽 ) (6) changed to searching for coordinates (𝑥, 𝑦) that minimize the
objective function 𝑓(𝑥, 𝑦). This optimizer 𝑓(𝑥, 𝑦) guarantees

97
 MEASUREMENT SCIENCE REVIEW, 24, (2024), No. 3, 95-99

minimum total error. IGWONL-UWSN system achieves optimal performance on

50 anchors with a lower LE value of 0.053, while the SS-DE,
𝑓(𝑥,𝑦) = ∑𝑚 𝑛 2 2 SS-NL, CS-NL and GWO-NL methods achieve maximum
𝑗=1 ∑𝑖=𝑚+1 |√((𝑥𝑖 − 𝑥𝑗 ) + (𝑦𝑖 − 𝑦𝑗 ) ) − 𝑑𝑗 | (9)
LE values of 0.281, 0.344, 0.356 and 0.445, respectively.

where (𝑥𝑖 , 𝑦𝑖 ) and (𝑥𝑗 , 𝑦𝑗 ) are the coordinates of the locations

of nodes 𝑖 and 𝑗. 𝑑𝑗 denotes the distance between unknown
nodes to AN 𝑗.

4. RESULTS AND DISCUSSION

In this section, the simulation results of the IGWONL-
UWSN method are examined in detail. Fig. 3 shows a
comparison of the IGWONL-UWSN system in terms of the
number of Localized Nodes (LN) under different anchors.
The results show that the IGWONL-UWSN method achieves
higher LN values. For example, on 10 anchors, the IGWONL-
UWSN system achieves a superior LN value of 163, while
the Smell Sensing (SS)-Differential Evolution (DE), SS-
Network Lifetime (NL), Cuckoo Search (CS)-NL and GWO-
NL techniques achieve a lower LN value of 142, 139, 128 and
119, respectively. On 50 anchors, the IGWONL-UWSN Fig. 4. LE analysis of the IGWONL-UWSN approach under
distinct anchors.
approach achieves an improved NL value of 198, while the
SS-DE, SS-NL, CS-NL and GWO-NL systems achieve a A comprehensive LE result of the IGWONL-UWSN
lower LN value of 182, 177, 165 and 147, respectively. technique with other methods under distinct transmission
ranges. The simulation values show that the IGWONL-
UWSN method has resulted in improved performance with
minimum LE values. At a 10 m transmission range, the
IGWONL-UWSN technique achieves better performance
with a minimum LE value of 0.134, while the SS-DE, SS-NL,
CS-NL and GWO-NL approaches achieve higher LE values
of 0.205, 0.286, 0.397 and 0.518, respectively. At the same
time, the IGWONL-UWSN system achieves the best
performance at a 30 m transmission range with a minimum
LE value of 0.016, while the SS-DE, SS-NL, CS-NL and
GWO-NL methods achieve the maximum value.
The computational complexity of IGWONL-UWSN is
balanced against its performance in terms of localization
accuracy. While the GWO algorithm used in IGWONL-
UWSN may require computational resources, especially for
large-scale networks, it has the advantage of optimizing the
node coordinates to efficiently minimize the LE. The trade-
offs between computational complexity and performance
Fig. 3. LN analysis of the IGWONL-UWSN approach under depend on factors such as network size, environmental
distinct anchors. conditions and the desired localization accuracy. Therefore,
it is important to evaluate the computational requirements of
A detailed LE assessment of the IGWONL-UWSN method IGWONL-UWSN in relation to its localization performance
compared to other systems under distinct number of anchors to achieve an optimal balance.
can be found in Fig. 4. The simulation values show that the
5. CONCLUSION
IGWONL-UWSN method resulted in improved performance
with lower LE values. On 10 anchors, the IGWONL-UWSN In this paper, we have developed a new IGWONL-UWSN
method achieves better performance with a minimum LE algorithm to determine the optimal location of SNs in the
value of 0.181, while the SS-DE, SS-NL, CS-NL and GWO- UWSN. The presented IGWONL-UWSN technique is
NL approaches achieve higher LE values of 0.319, 0.461, inspired by the hunting behavior of grey wolves with the DLH
0.477 and 0.563, respectively. At the same time, the search process. The proposed IGWONL-UWSN technique

98
 MEASUREMENT SCIENCE REVIEW, 24, (2024), No. 3, 95-99

utilizes the IGWO algorithm to calculate the optimal location [8] Ismail, A. S., Wang, X., Hawbani, A., Alsamhi, S.,
of the nodes in the UWSN. In addition, the IGWONL-UWSN Abdel Aziz, S. (2022). Routing protocols classification
technique incorporates the DLH search process for greater for underwater wireless sensor networks based on
convergence and accuracy. The simulation results of the localization and mobility. Wireless Networks, 28 (2),
IGWONL-UWSN technique are validated using a set of 797-826. https://doi.org/10.1007/s11276-021-02880-z
performance measures. The simulation results illustrate the [9] Muthulakshmi, K., Sundar Prakash, B., Stephe, S.,
developments of the IGWONL-UWSN method over other Vijayalakshmi, J. (2024). Adaptive wind driven
systems in terms of various metrics. In the future, node optimization based energy aware clustering scheme for
mobility can be considered in the development of the wireless sensor. Technical Gazette, 31 (2), 466-473.
https://doi.org/10.17559/TV-20230610000715
IGWONL-UWSN technique in the UWSN. Future directions
[10] Nain, M., Goyal, N. (2022). Energy efficient
in this area could include the development of adaptive
localization through node mobility and propagation
algorithms capable of adapting to changing environmental
delay prediction in underwater wireles sensor network.
conditions, the integration of machine learning techniques to
Wireless Personal Communications, 122 (3), 2667-
improve localization accuracy, and the exploration of 2685. https://doi.org/10.1007/s11277-021-09024-8
applications in emerging areas such as underwater robotics [11] Sun, Y., Ge, W., Li, Y., Yin, J. (2023). Cross-layer
and autonomous systems. protocol based on directional reception in underwater
acoustic wireless sensor networks. Journal of Marine
REFERENCES
Science and Engineering, 11 (3), 666.
[1] Luo, J., Chen, Y., Wu, M., Yang, Y. (2021). A survey https://doi.org/10.3390/jmse11030666
of routing protocols for underwater wireless sensor [12] Prateek, Reddy, T. S., Chandra, S., Arya, R., Verma, A.
networks. IEEE Communications Surveys & Tutorials, K. (2022). Malicious anchor node extraction using
23 (1), 137-160. geodesic search for survivable underwater wireless
https://doi.org/10.1109/COMST.2020.3048190 sensor network. Scientific Reports, 12 (1), 13691.
[2] Patel, M., Saxena, P., Panchal, C. (2019). Target https://doi.org/10.1038/s41598-022-17956-9
localization scheme for underwater acoustic sensor [13] Lv, C., Zhu, J., Chen, G. (2023). A localization scheme
network. International Journal of Applied Engineering based on Improving Dynamic Population Monte Carlo
Research, 14 (8), 2002-2008. Localization method for large‐scale mobile wireless
[3] Hong, Y., Wang, S., Kang, H., Hu, Y. (2022). Iterative aquaculture sensor networks. IET Wireless Sensor
virtual force localization based on anchor selection for Systems, 13 (2), 58-74.
three-dimensional wireless sensor networks. Technical https://doi.org/10.1049/wss2.12053
Gazette, 29 (3), 1048-1058. [14] Balakrishnan, S., Vinoth Kumar, K. (2023). Hybrid
sine-cosine black widow spider optimization based
https://doi.org/10.17559/TV-20220219093303
route selection protocol for multihop communication in
[4] Han, G., Wang, H., Ansere, J. A., Jiang, J., Peng, Y.
IoT assisted WSN. Technical Gazette, 30 (4), 1159-
(2020). SSLP: A stratification-based source location
1165. https://doi.org/10.17559/TV-20230201000306
privacy scheme in underwater acoustic sensor [15] Krishnamoorthy, V. K., Duraisamy, U. N., Jondhale, A.
networks. IEEE Network, 34 (4), 188-195. S., Lloret, J., Ramasamy, B. V. (2023). Energy-
https://doi.org/10.1109/MNET.001.1900478 constrained target localization scheme for wireless
[5] Thiruppathi, M., Vinoth Kumar, K. (2023). Seagull sensor networks using radial basis function neural
optimization-based feature selection with optimal network. International Journal of Distributed Sensor
extreme learning machine for intrusion detection in fog Networks, 2023, 1426430.
assisted WSN. Technical Gazette, 30 (5), 1547-1553. https://doi.org/10.1155/2023/1426430 Jayamala, R.,
https://doi.org/10.17559/TV-20230130000295 Sheryl Oliver, A., Jayanthi, J., Nithya, N. (2024).
[6] Ahmad, M. F., Isa, N. A. M., Lim W. H., Ang, K. M. Enhanced secured and real-time data transmissions in
(2022). Differential evolution with modified wireless sensor networks using SFRT routing protocol.
initialization scheme using chaotic oppositional based Technical Gazette, 31 (2), 420-425.
learning strategy. Alexandria Engineering Journal, 61 https://doi.org/10.17559/TV-20230617000742
(12), 11835-11858. [16] Ponni, R., Jayasankar, T., Vinoth Kumar, K. (2023).
https://doi.org/10.1016/j.aej.2022.05.028 Investigations on underwater acoustic sensor networks
[7] Kumar, M., Goyal, N., Qaisi, R. M. A., Najim, M., framework for RLS enabled LoRa networks in disaster
Gupta, S. K. (2022). Game theory based hybrid management applications. Journal of Information
localization technique for underwater wireless sensor Science and Engineering, 39 (2), 389-406.
https://doi.org/10.6688/JISE.202303_39(2).0009
networks. Transactions on Emerging
Telecommunications Technologies, 33 (11), e4572.
https://doi.org/10.1002/ett.4572 Received September 17, 2023
Accepted April 17, 2024

99