Enhanced Network QoS in Large Scale and High Sensor Node Density Wireless Sensor Networks Using (IR-DV-Hop) Localization Algorithm and Mobile Data Collector (MDC)

This paper poses new challenges, especially when designing routing protocols to improve the quality of service (QoS) criteria and the lifetime of large-scale wireless sensor networks (LS-WSNs) and high sensor node density WSNs (HSND-WSN). Some of these challenging problems for routing protocols are minimizing the distance between a base station (BS) and the cluster head (CH), latency in the transmission phase, power consumption in the clustering phase, and determining the precise SN location without error. Thus, for a dynamic WSN, it is important to support an intelligent mobile data collector (MDC) to continue data propagation despite the inevitable changes in the WSN topology. Considering these challenges, we propose a novel intelligent routing protocol based on the improved recursive distance vector-hop (IR-DV-Hop) localization algorithm to determine the accurate location of SNs without errors in LS-WSNs and determine the CHs and optimal route traversed by the MDC for energy efficiency, latency, and reliability. Specifically, the proposed mobile data collector-improved recursive distance vector-hop (MDC-IR-DV-Hop) protocol uses the MDC as an intermediate between the CH and BS to enhance the QoS of WSNs, reduce delays while collecting data, and improve the transmission phase of the routing protocol. Simulation results prove the performance of the proposed protocol compared with the energy-aware cluster based multi-hop (EACBM), tree clustering algorithm with mobile data collector (TCMDC), mobile data collector maximum residual energy low energy adaptive clustering hierarchy (MDC maximum residual energy leach), MDC minimum distance leach, mobile data collector-K-means (MDC-K), mobile data collector-traveling salesman problem-low energy adaptive clustering hierarchy-K-means (MDC-TSP-LEACH-K), a three-layer hierarchical architecture for optimized routing protocol based on clustering and the Khalimsky topology for mobile WSN (AMWSN), and node density-based clustering and mobile collection (NDCM) protocols. In addition, this article proposes a comparative analysis of the best clustering, CH election, localization and MDC by discussing and describing their functionalities, advantages, disadvantages and their main performance parameters.

SECTION I. Introduction

The growth of wireless sensor networks (WSNs) has recently accelerated owing to their capability to sense and compute data efficiently. This rapid growth of WSNs has contributed to their wide use in many applications, such as telecommunications, smart homes, disaster management, agriculture, drone applications, military services, and medical applications. As WSNs comprise small and low-cost sensor nodes (SNs), they suffer limitations, such as limited battery capacity, low memory, and a lower communication distance [1]. Energy use in WSNs is continuous because it is used while sensing, collecting, and transmitting data. However, energy consumption in the data transmission phase is higher on average. Although WSNs are low-cost, easily deployed, flexible, and efficient, there are challenges in energy efficiency and quality of service (QoS). The energy efficiency problem has recently gained more attention, as it is not easy to change or recharge the battery for large-scale networks [2]. One of these problems involves clustering, election of the cluster head (CH), and data transmission in WSNs. Clustering in WSNs is the most reliable solution for the challenges, where SNs are grouped into a few clusters, and a CH is selected for data aggregation. However, many challenges still exist in energy efficiency when designing routing protocols, such as minimizing the distance between SNs and the CH and between the base station (BS) and CH in the clustering phase. The announcement of the nearest CH location to the BS is one of these challenges and is a central problem for any routing protocol. Although much research has focused on reducing the energy consumption of routing protocols, few researchers have addressed other QoS criteria, such as stability, and throughput.

The clustering mechanism hierarchically provides SN grouping and ensures efficiency, scalability, and collaboration in the network. As a promising approach, it reduces the required total transmissions to the BS and reduces cluster energy because CHs aggregate the data from their SNs and transmit them to the BS. In addition, the dynamic topology incurs maintenance costs when using the clustering technique. Typically, reconfiguration is performed at the level of the CH, which does not affect the other SNs in the cluster. In summary, clustering can achieve several objectives, such as improved scalability, uniform network energy, efficient resource use, and enhanced QoS [3], [4]. These clustering algorithms for WSNs executed on data streams are divided into two main steps: the clustering and data transmission phases. In the clustering phase, by employing the density-based clustering algorithm [5], each cluster is formed by the same number of SNs [6]. Furthermore, other transmission algorithms have been suggested in the literature, such as the distance vector-hop (DV-Hop) algorithm and its improvements [7], [8] that reduce the localization error. However, the battery energy in a WSN is a limited resource for SNs. Therefore, the energy consumption of SNs determines their and the network’s lifetimes, which affects the network connectivity and coverage. A smart mobile data collector (MDC) can reduce energy consumption for data collection instead of transmitting multi-hop data to the BS. The MDC collects data from SNs and transfers it to the BS. Various MDC approaches have been explored for different assumptions and constraints [9], [10]. However, in all proposed models, the data latency is usually high due to the slow speed of the mobile nodes.

This paper proposes a new routing strategy based on clustering and a data collection mechanism based new model of MDC on wireless communication. Using clustering, IR-DV-Hop protocols and a MDC, we demonstrate that the delay can be reduced significantly without compromising the advantages of the MDC-based approach. Using extensive simulation studies, we analyzed the performance of the proposed approach and demonstrated that the packet delay was reduced by more than half compared to other existing approaches. Therefore, we propose a novel mobile routing protocol that minimizes significant data latency and improves several QoS criteria. This research focuses on combining the MDC and improved-recursive (IR) DV-Hop during in transmission phase using a combination of a novel equal-clustering-based round strategy and a wireless-based data collection mechanism, where the IR-DV-Hop protocol is an improvement of the original DV-Hop algorithm and updates the recursive least squares solution. We apply a recursive method for the SN localization process to help the BS determine the precise SN locations, elect the optimal CH, and determine the optimal MDC path to collect CH data. Thus, the proposed MDC-IR-DV-Hop-K protocol uses equal clustering to decrease energy consumption and the IR-DV-Hop algorithm to decrease the localization error, provide efficient reliability, and improve the CH election phase. It also uses a novel intelligent MDC based on the IR-DV-Hop results and a spiral path for efficient QoS. Specifically, the contribution of this paper is as follows:

• In this paper, We have proposed a new routing protocol called IR-DV-Hop-LEACH-K. We assume that the BS is not limited in energy and that the BS coordinates and field dimensions are known, at least in comparison to the energy of other SNs. We also assume that the SNs are uniformly distributed over the field and are not mobile. First, the BS applies the IR-DV-Hop algorithm in the WSN to determine the location of all SNs without error. Then, its fixed spiral is the MDC trajectory, beginning from the centroid of the area. After that, The BS assigns each CH to the SN that belongs to $spiral_{MDC}$ . In addition, we propose a threshold that balances the number of member SNs of all clusters. Each CH accepts the SN as a member if the distance between itself and the SNs is less than or equal to the minimal distance. The CH accepts this SN as a member SN whose number of member SNs is less than or equal to the threshold. In addition, CHs collect and aggregate data received from their cluster members. The intelligent MDC is based on the spiral trajectory starting from the centroid area as an interface between the CHs and BSs.

• This protocol uses the IR-DV-Hop algorithm to decrease the localization error, provide efficient reliability, and improve the CH election phase. Thus, it uses a smart MDC as an intermediary between the BS and CH. The MDC uses a spiral trajectory to collect data from all CHs for large-scale WSNs (LS-WSNs) and high SN density (HSND) WSNs, decreasing the latency and energy consumption. We study the effect of the area variation (from $100~m^ {2}$ to $10000~m^ {2}$ ) on QoS criteria.

• The simulation phase is divided into two scenarios: in the first, we justified our choice of localization protocol. In the second, we tested our protocol in the LS-WSNs and HSND-WSNs cases and demonstrated its higher performance.

The article is organized into the following sections. Section II discusses the related work. Next, Section III describes the IR-DV-Hop mechanism. Then, Section IV explains the proposed routing protocol. Section V presents the simulation results of the MDC-IR-DV-Hop mechanism. Finally, Section VI provides the conclusions.

SECTION II. Related Work

One of the main approaches to designing energy-efficient, robust, and highly scalable distributed networks is to organize WSNs into clustered architectures. In WSNs, the SNs are energy constrained. Therefore, CH selection among a cluster’s coordinators is a major problem in these network applications and can severely affect the energy dissipation of a network. Many studies have been conducted to evaluate the clustering algorithm to increase the lifetime of the WSN. The clustering in WSNs involves dividing the SNs into groups according to certain characteristics. Clustering can be formed based on the network topology, location, and residual energy. In each cluster, there is an SN that takes on more responsibilities, which is called the CH. The techniques of forming the CH are two-fold (distributed or centralized) in WSNs. In the former technique, each SN broadcasts information, such as residual energy, to its neighbor at a hop, and finally, the SN with the highest value becomes the CH. In the second technique, each SN is responsible for sending the information to the BS, which performs the calculation and selects CHs in the network. Subsequently, the BS notifies all nodes about their respective CHs. The clustering process consists of three steps (i.e., CH selection, cluster formation, and data transmission; Fig. 1).

FIGURE 1.

Clustering design process.

Several research studies have been performed to elect the CH in clustering protocols, focusing on minimizing the energy consumption of WSNs, whereas other research studies have evaluated the effect of various machine learning algorithms on clustering performance in WSNs. Table 1 summarizes some of the literature.

TABLE 1 Summary of Cluster Head (CH) Selection-Based Protocols

The clustering process and cluster number are very important factors in clustering protocols. The number of exchanged messages during the formation of clusters must be minimized, and the clusters must be well-balanced. In addition, the algorithm complexity must increase linearly. In clustering techniques, one must also consider that the designed clustering algorithm must be able to meet the requirements of different applications. Furthermore, another crucial factor is ensuring that the designed algorithm is secure enough and can be used in very sensitive data applications and large-scale applications, such as agricultural or military applications. Table 2 summarizes some of the literature on the cluster formation method.

TABLE 2 Summary of Cluster Formation Method Literature

The cluster formation techniques minimize the problem of hot spots in WSN deployments. Table 2 addresses various cluster formation techniques proposed by researchers in recent years. As Table 2 illustrates, density-based clustering is another clustering algorithm widely applied in various WSN applications. Recently, the MDC and clustering techniques have attracted growing interest. The three most important challenges are the network division into clusters, CH election, and building an optimal MDC path. Therefore, many research studies on MDC and clustering approaches have been suggested to extend the network lifetime. The objective of Table 2 is to survey and synthesize the MDC approaches of recent studies in routing protocols.

Many researchers in the existing WSN literature have demonstrated that hierarchical routing, particularly clustering, is the most appropriate approach to increase the network lifetime and throughput and minimize energy consumption in WSNs. This approach involves a process of smart Voronoi clustering of SNs using localization algorithms as a basis for classification. We distinguished various categories of clustering techniques and localization algorithms. The most popular are Voronoi and the IR-DV-Hop algorithm. However, the difficulty lies in choosing the CH, supervising the clusters, and improving the QoS of WSNs.

SECTION III. Improved Recursive DV-Hop Localisation Algorithm

The IR-DV-Hop algorithm enhances the DV-Hop algorithm [39] and implements a recursive solution of least squares. It uses a recursive approach for the location process based on a collection of anchor nodes chosen from a predefined population of anchors. The IR-DV-Hop algorithm concentrates on an algorithm for range-free localization in multihop homogeneous WSNs using a recursive calculation of the position of unknown SNs. It also concentrates on the second and third steps of the DV-Hop algorithm [8].

This approach employs a formulation optimized to compute anchor nodes’ average hop size to reduce the localization error in the predicted distance between the anchor and the unknown SN, resulting in higher localization accuracy. Algorithm 1 presents the pseudocode of the IR-DV-Hop algorithm. In the IR-DV-Hop algorithm, anchors are randomly created in the WSN with a determined location. Using these anchors, this algorithm randomly selects anchor candidates to provide an estimation reference to the location process. The initial position of each unknown node is $P_{0}$ = 0. Subsequently, The unknown nodes calculate their position $P_{n}$ using (Eq. 9) in [8].The major problem with WSNs in real applications is that certain anchor locations cannot be provided to continuously locate the unknown SNs for numerous reasons, such as lifetime, failure, and maintenance. Thus, the IR-DV-Hop algorithm considers a novel collection of anchor candidates chosen randomly and iteratively based on their availability within the anchor population for the position estimation process.

SECTION Algorithm 1 IR-DV-Hop

Input:

SNs distributed randomly in a (100, 100) m detection field, including anchoring SNs of coordinates ( $A_{i}$ , $B_{i}$ ) where i = 1…z, and unknown SNs ${n}$ to be located.

Output:

The $P_{n}$ estimation position of ${n}$ unknown SNs.

Begin:

$P_{n}$ (0) = 0; /* Initial position of unknown SN.

M = $\text{a}\times \text{I}$ ; /* Convariance matrix ${M}$ , where ${a}$ is a very high positive number and ${I}$ is the identity matrix.

Random selection of a set of candidate anchors for the localization process.

while The condition of arrest is not fulfilled do

1:

Process of trilateration and calculation of the lowest number of hops between the selected anchors.

2:

Improvement of the average distance of jumps between candidate anchors.

3:

Minimum number of hops between selected anchors and unknown SNs.

4:

Distance estimates between the unknown SNs and the closest selected anchors.

5:

Estimation of the position $P_{n}$ of an unknown SN utilizing Recursive Least Squares. $P_{n+1}(0)=P_{n}+M_{n+1}An_{n+1}^{T}(B_{n+1}-An_{n+1}P_{n})$ (Eq. 16) in

end while

$(x=P_{n}(1) $ , $x=P_{n}(2))$ /* Estimated coordinates of unknown SN ${n}$

end

SECTION IV. Mobile Data Collectors-Improved Recursive Distance Vector-Hop Localization K-Means Protocol

The intelligent MDC for smart LS-WSNs approach offers the MDC-IR-DV-Hop-K protocol, which is also based on rounds, where each round is divided into two phases: the initialization and transmission phases. Before moving to these phases, the SNs are distributed randomly. All SNs have the same initial energy and are homogeneous and fixed. Moreover, the BS is also fixed.

A. Initialization Phase

In this approach, we take the BS outside the deployment area to test the difficult cases because most protocols test protocols where the BS is located in the center of the area [9], [10], [19], [24], [40], [41]. First, the BS applies the IR-DV-Hop algorithm in the WSN to determine the location of all SNs without error. Then, its fixed spiral is the MDC trajectory, beginning from the centroid of the area. The BS assigns each CH to the SN that belongs to $spiral_{MDC}$ with coordinates ($x_{i}$ , $y_{i}$ ), with the minimum distance to the spiral MDC trajectory. The minimum distance is calculated as follows:\begin{align*} &D(SN, spiral_{MDC}) \\ &\quad =min \lbrace \!\! \sqrt {{\!(x_{spiral_{MDC}}\!\!-x_{i})}^{2}\!\!+\!\!{(y_{spiral_{MDC}}\!\!-\!\!y_{i}\!)}^{2}} \}\!. \tag{1}\end{align*}

We use the K-means machine learning algorithm to find the position of the centroid area. Then, the BS selects the SNs located on the path of the spiral as CHs. The number of CHs is calculated as follows:\begin{align*} Number_{CH} &= \sum _{i=0} (D(SN, spiral_{MDC}) \\ &=min \lbrace \!\! \sqrt {{\!(x_{spiral_{MDC}}\!\!-x_{i})}^{2}\!\!+\!\!{(y_{spiral_{MDC}}\!\!-\!\!y_{i}\!)}^{2}} \}\!). \tag{2}\end{align*}

Finally, each CH chooses the member SNs closest to it. However, in most cluster protocols, the clusters are very large, and no load balancing occurs between the clusters, which leads to energy waste. Therefore, we propose a threshold to limit and balance the number of member SNs in each cluster. This threshold is calculated as follows (3):\begin{equation*} Threshold = \frac {(Number_{living SN}- Number_{CH})}{Number_{CH}}. \tag{3}\end{equation*}

Each CH calculates the distance between itself and the other SNs to determine which SNs it accepts. If this distance is less than or equal to the minimal distance, the CH accepts this SN as a member SN whose number of members SNs is less than or equal to the threshold we calculated previously. Otherwise, this SN does not attach to this CH and attaches to another CH. Similarly, all clusters contain about the same number of SNs, and the CH assignment of a time division multiple access (TDMA) schedule for SN members is performed as illustrated in Fig. 3. After the cluster is created, each CH generates a TDMA schedule and sends these schedules to its member SNs in the cluster. The TDMA schedule prevents the data sent by the member SNs from colliding and allows the member SNs to enter sleep mode. The SN members transmit data to the CH during their time slot allocation. When an SN member sends data to the CH during its allocated time slot, another SN member in this cluster stays in the sleep state. This process is represented in Fig. 3 by a line connecting the red rectangle A. This feature of the MDC-IR-DV-Hop-K protocol reduces cluster collisions and energy consumption, which increases the lifetime of all SN members. In addition, CHs collect and aggregate data received from their cluster members, as presented in Fig. 3.

FIGURE 2.

Process of the MDC-IR-DV-Hop-K.

FIGURE 3.

Algorithm of the MDC-IR-DV-Hop-K.

B. Transmission Phase

With the proposed approach in the initialization phase of the MDC-IR-DV-Hop-K protocol, the energy consumption is reduced, and the stability is increased compared to other clustering protocols. Nevertheless, improving the QoS criteria and reducing energy consumption is still a significant problem. Thus, we used the intelligent MDC based on the spiral trajectory starting from the centroid area as an interface between the CHs and BSs, as depicted in Fig. 2.

The MDC intelligently performs by collecting data only from the CH that belongs to its path. The MDC collects the data, which are broadcasted by the CH to the MDC directly. Then, it completes the collection and provides data to the BS, as illustrated in Figs. 2 and 3. Fig. 3 illustrates the initialization and transmission phase of MDC-IR-DV-Hop-K data transmission between the BS and CH.

C. Energy Model of the MDC-IR-DV-Hop-K Protocol

The MDC-IR-DV-Hop-K adapts the energy model of the LEACH-K [42], MDC-K [10], and MDC-TSP-LEACH-K [19] protocols, as presented in Fig 4. According to this model, the transmitter or receiver circuits operate with an electronic dissipation energy of $E_{elec}$ = 50nJ/bit, and the transmitter amplifier uses $\epsilon _{Amp}$ = $0.0013 pJ/bit/m^{4}$ . However, when the distance is below a threshold of $T$ , we use the free space model (energy loss ${D}^{2}$ ); otherwise, we use the multipath model (energy loss ${D}^{4}$ ). Alternatively, for a short-distance transmission, like an intra-group communication, a transmission amplifier’s energy consumption is proportional to ${D}^{2}$ , whereas, for a longer-distance transmission, like an inter-group communication, a transmission amplifier’s energy consumption is proportional to ${D}^{4}$ . In this case, a threshold transmission distance $T$ is defined, in which:\begin{equation*} T= \sqrt {\epsilon _{fs}} / \sqrt {\epsilon _{Amp}}. \tag{4}\end{equation*} Therefore, if the sender sends a ${B}$ bit of data to the receiver up to a distance $D$ , the energy necessary to send ${B}$ bits of data is represented by:\begin{equation*} E_{Transmitting} = (E_{ele}B) +\epsilon _{Amp} BD^{4}. \tag{5}\end{equation*}

FIGURE 4.

Energy model of the MDC-TSP-LEACH-K.

Thus, the necessary energy to obtain ${B}$ bits of data is expressed by:\begin{equation*} E_{Receiving}(B) = E_{ele}B. \tag{6}\end{equation*}

Finally, the data collection energy consumed during a round is computed as follows:\begin{equation*} E_{Round} = E_{Transmitting} +E_{Receiving}. \tag{7}\end{equation*} where:

• $\epsilon _{fs}$ denotes the free space energy loss.

• $\epsilon _{Amp}$ represents the power of multi-path models.

• $E_{ele}$ indicates the energy dissipation per bit for transmission and reception.

SECTION V. Simulation Strategy, Performance Evaluation, and Result Discussion

This section evaluates the MDC-IR-DV-Hop-K protocol by comparing it to existing protocols that use MDC at the routing level. We compared the proposed protocol with others that have improved QoS metrics, such as latency, power consumption, and stability. We implemented it and tested its performance in MATLAB. We selected MATLAB due to the ease of its interface and because the required preprogrammed functions, such as the K-means and spiral trajectory functions, are available. In addition, it is easy to implement the mathematical model on MATLAB because MATLAB is a matrix language enabling the most natural expressions of mathematical computation. Different parameters and aspects were considered during a simulation to test the protocol performance in WSNs. Table 4 describes the simulation parameters.

TABLE 3 Summary of the MDC Approaches in the Data Transmission Phase

TABLE 4 Simulation Parameters

Several parameters and factors are considered in the simulation to study the efficiency of the proposed approach. The simulation is divided into three scenarios. In the first scenario, we justify why we chose IR-DV-Hop as the localization algorithm. We tested the performance of IR-DV-Hop against other algorithms, such as the DV-Hop, UDV-Hop, and advanced DV-Hop algorithms. In the second scenario, we assessed the simulation results of the proposed MDC-IR-DV-Hop-K protocol with a variety of protocols, including the MDC maximum residual energy leach [34], NDCM [35], TCMDC [31], MBEENISH [36], EEHPMDC [37], MDC minimum distance leach [38], MDC-K [9], MDC-TSP-LEACH-K [19], MDC-LEACH-K [10] and a three-layer hierarchical architecture for optimized routing protocol based on clustering and the Khalimsky topology for mobile WSN (AMWSN) protocols [43]. In the third scenario, we studied the performance of the MDC-IR-DV-Hop-K protocol in the case of LS-WSNs.

A. First Scenario

We justify choosing IR-DV-Hop as the localization algorithm and its advantages in this scenario. Therefore, we simulated and compared IR-DV-Hop with DV-Hop, UDV-Hop, and advanced DV-Hop. Within the simulation, the number of network SNs is 300 with a 15 m communication radius. The proportion of anchors randomly deployed in the detection field varies from 5% to 30% of the total SNs. Fig. 5 compares the localization error percentage obtained by the IR-DV-Hop algorithm and the DV-Hop, UDV-Hop, and advanced DV-Hop algorithms as a function of the percentage of anchor nodes.

FIGURE 5.

IR-DV-Hop, DV-Hop, UDV-Hop, and advanced DV-Hop algorithms’ localization error percentage as a function of the percentage of anchor nodes.

As presented in Fig. 5, the IR-DV-Hop algorithm is more precise than the DV-Hop, UDV-Hop, and advanced DV-Hop algorithms. It reached a localization error below 17% at 30% anchoring, whereas the localization errors were around 19% for the UDV-Hop algorithm, around 21% for the advanced DV-Hop algorithm, and around 36% for the DV-Hop algorithm. Fig. 6 displays the localization error obtained by the IR-DV-Hop, DV-Hop, UDV-Hop, and advanced DV-Hop algorithms as a function of the variation of the communication radius of the nodes.

FIGURE 6.

R-DV-Hop, DV-Hop, UDV-Hop, and advanced DV-Hop algorithms’ localization error percentage as a function of the percentage of anchor nodes.

According to Fig. 6, the IR-DV-Hop algorithm has about a 15% lower localization error than the DV-Hop algorithm, 5% lower than the advanced DV-Hop algorithm, and 3% lower than the UDV-Hop algorithm. In addition, in the case of a high communication radius value of around 40 m, the IR-DV-Hop algorithm has an approximately 14% error in localization, whereas the DV-Hop has about a 28% error in localization. Advanced DV-Hop has about a 19% error in localization, and UDV-Hop has an error of about 17%. Based on the simulation results in Figs. 5 and 6, IR-DV-Hop provides higher localization precision than the DV-Hop, UDV-Hop, and advanced DV-Hop algorithms. Consequently, we chose IR-DV-Hop as the localization algorithm for the proposed approach.

B. Second Scenario

To evaluate the performance of the MDC-IR-DV-Hop-K protocol, we simulated both MDC-IR-DV-Hop-K protocols with various protocols, including the EACBM, TCMDC, MDC maximum residual energy LEACH, MDC minimum distance leach, MDC-K, MDC-TSP-LEACH-K, AMWSN, and NDCM protocols using MATLAB. In this scenario, we analyzed the simulation results and adopted performance measures. These findings were analyzed for a trade-off between lifetime, throughput, stability, and latency over 10 500 rounds. Fig. 7 compares the MDC-IR-DV-Hop-K approach with the EACBM, TCMDC, MDC-TSP-LEACH-K, MDC maximum residual energy leach, MDC minimum distance leach, MDC-K, MhSa-LEACH, AMWSN, and NDCM protocols in terms of lifetime.

FIGURE 7.

Comparison of the MDC-IR-DV-Hop-K approach with the EACBM, TCMDC, MDC maximum residual energy leach, MDC minimum distance leach, MDC-K, MDC-TSP-LEACH-K, AMWSN, and NDCM protocols in terms of lifetime.

As observed in Fig. 7, the proposed MDC-IR-DV-Hop-K protocol enhances the network lifetime with better performance than the EACBM, TCMDC, MDC maximum residual energy leach, MDC minimum distance leach, MDC-K, MDC-TSP-LEACH-K, AMWSN, and NDCM protocols, but it is not as stable as the MDC-K protocol. According to the simulation results, MDC-IR-DV-Hop-K also improves the residual energy of SNs. The numerical results demonstrate that the proposed approach can reduce the energy consumption of the SNs. Fig. 8 demonstrates the stability of the proposed approach compared to the EACBM, TCMDC, MDC maximum residual energy leach, MDC minimum distance leach, MDC-TSP-LEACH-K, AMWSN, and NDCM protocols.

FIGURE 8.

Comparison of the MDC-IR-DV-Hop-K approach with the EACBM, TCMDC, MDC-TSP-LEACH-K, MDC maximum residual energy leach, MDC minimum distance leach, MDC-K, MhSa-LEACH, AMWSN, and NDCM protocols in terms of stability.

As depicted in Fig. 8, the stability improves a little bit from 733 (rounds) in AMWSN to 807 (rounds) in EACBM, 1200 (rounds) in MDC minimum distance leach, 1302 (rounds) in TCMDC, 1650 (rounds) in NDCM, 1913 (rounds) in MDC maximum residual energyleach, 1950 (rounds) in MDC-TSP-LEACH-K, 2001 (rounds) in MDC-IR-DV-Hop-LEACH-K, 2800 (rounds) in MDC-K, and 2000 (rounds). The proposed protocol is less stable than the MDC-K protocol but is more stable than MDC-TSP-LEACH-K, which is a routing protocol that uses the MDC for the LS-WSN with a very short period. Fig. 9 depicts the latency time of the proposed approach MDC-IR-DV-Hop-LEACH-K compared to the EACBM, TCMDC, MDC maximum residual energy leach, MDC minimum distance leach, MDC-TSP-LEACH-K, AMWSN, and NDCM protocols.

FIGURE 9.

Comparison of the MDC-IR-DV-Hop-K approach with EACBM, TCMDC, MDC-TSP-LEACH-K, MDC maximum residual energy leach, MDC minimum distance leach, MDC-K, MhSa-LEACH, AMWSN, and NDCM protocols in terms of latency time.

As observed in Fig. 9, using the MDC spiral movement based on the locations proposed by the IR-DV-Hop algorithm in this approach, the latency decreases to 26.08 ms from the MDC-IR-DV-Hop-K protocol compared to 35.16 ms from the MDC-TSP-LEACH-K protocol, 50.001 ms in the MDC-K protocol, 61.27 ms in the MDC maximum residual energy leach protocol, 67,77 ms in the NDCM protocol, 71.09 ms in the EACBM protocol, 79.87 ms in the AMWSN protocol, 85.76 ms in the MDC minimum distance leach protocol, and 89.98 ms in the TCMDC protocol. However, we conclude from the experimental results that the MDC-IR-DV-Hop-K protocol is the best solution to reduce the latency time compared to the MDC-TSP-LEACH-K and other routing protocols. Fig. 10 presents the throughput of the proposed approach (MDC-IR-DV-Hop-LEACH-K) compared to the EACBM, TCMDC, MDC maximum residual energy leach, MDC minimum distance leach, MDC-TSP-LEACH-K, AMWSN, and NDCM protocols.

FIGURE 10.

Comparison of the MDC-IR-DV-Hop-K approach with the EACBM, TCMDC, MDC-TSP-LEACH-K, MDC maximum residual energy leach, MDC minimum distance leach, MDC-K, MhSa-LEACH, AMWSN, and NDCM protocols in terms of throughput.

Fig. 10 illustrates the throughput simulation results of the MDC-IR-DV-Hop-K protocol versus EACBM, TCMDC, MDC-TSP-LEACH-K, MDC maximum residual energy leach, MDC minimum distance leach, MDC-K, AMWSN, and NDCM protocols. The plots in Fig. 10 demonstrate that the integration of the MDC using the spiral trajectory starts with the centroid of the area and increases the throughput by a significant amount. We observed that the throughput value of MDC-IR-DV-Hop-K at 10 000 rounds is equal to 32 060 packets/round compared to 29 110 packets/round for the EACBM protocol, 18 910 packets/round for the MDC-TSP-LEACH-K protocol, 18 300 packets/round for the MDC maximum residual energy leach protocol, 18 100 packets/round for the MDC-K protocol, 15 500 packets/round for the NDCM, 13 000 packets/round for the AMWSN protocol, 10 000 packets/round for the MDC minimum distance leach protocol, and 9001 packets/round for the TCMDC protocol. The proposed protocol increases the throughput value by minimizing the distance traveled by the MDC through the spiral movement starting from the simulation centroid area using the K-means algorithm and the precision of SN localization at the CH selection level using IR-DV-Hop algorithm. Fig. 10 demonstrates that the BS packet count for TCMDC is very low compared to other protocols. Comparing the MDC-IR-DV-Hop-K and NDCM reveals that MDC-IR-DV-Hop-K is better than NDCM from 1500 rounds. In 1500 rounds, the BS received packet count deviates significantly from the preceding round. The second scenario studies the proposed approach in LS-WSNs and HSND-WSNs to ensure the effectiveness of the approach.

C. Second Scenario

Improving the QoS criteria and prolonging the lifetime of WSNs is a difficult challenge, particularly for LS-WSNs, owing to the large geographic areas for data collection, high SN density, and high quantities of data to collect. This scenario is divided into two subscenarios. First, we study the efficiency of the proposed protocol in the LS-WSN compared to a protocol that uses an MDC in LS-WSNs. Second, we study the efficiency of the proposed protocol with density variation compared to a protocol that uses an MDC in HSND-WSNs.

1) Evaluation of the MDC-IR-DV-Hop-K in an LS-WSN

In this subscenario, we also test the performance of the proposed MDC-IR-DV-Hop-K protocol in LS-WSNs. More precisely, we estimate the QoS criteria values for different network sizes compared to a protocol that uses the MDC in LS-WSNs. Fig. 11 compares the MDC-IR-DV-Hop-K approach, MDC-TSP-LEACH-K, MDC maximum residual energy leach, MDC minimum distance leach, and MDC-K protocols in terms of lifetime, throughput, stability, and latency time in different area sizes.

FIGURE 11.

Comparison of the MDC-IR-DV-Hop-K approach, MDC-TSP-LEACH-K, MDC maximum residual energy leach, MDC minimum distance leach, and MDC-K protocols in terms of lifetime, throughput, stability, and latency time in different area sizes.

Fig. 11 demonstrates that as the area size is scaled up, the lifetime, throughput, and stability decrease a little, which is evidence that the advantages of the proposed protocol remain nearly stable with a sizable area. For instance, the latency also increases from 26.08 ms in a $100\times 100$ area to 29.09 ms in a $1500\times 1500$ area. Throughput changes from 32 060 packets/round in a $100\times 100$ area to 31 890 packets/round in a $1500\times 1500$ area. Lifetime alters from 8566 rounds in a $100\times 100$ area to 8471 rounds in a $1500\times 1500$ area. The stability changes from 2001 rounds in a $100\times 100$ area to 1813 rounds in a $1500\times 1500$ area. Based on these results, the MDC-IR-DV-Hop-K is the best solution for the LS-WSN regarding lifetime, throughput, latency, and reliability. However, in the case of stability, the MDC-K is still better than the MDC-IR-DV-Hop-K and MDC-TSP-LEACH-K protocols.

2) Evaluation of MDC-IR-DV-Hop-K in an HSND-WSN

We evaluated the effect of varying the SN density in the MDC-IR-DV-Hop-K protocol. The simulation consists of 100 to 2000 homogeneous SNs with an initial energy of 0.5 J, randomly dispersed in an SN field of $100\times 100$ m. The BS is located at (0.5, 125), at least 125 m. We compared the MDC-IR-DV-Hop-K protocol with MDC-TSP-LEACH-K, MDC maximum residual energy leach, MDC minimum distance leach, and MDC-K protocols. Fig. 12 compares the MDC-IR-DV-Hop-K approach, MDC-TSP-LEACH-K, MDC maximum residual energy leach, MDC minimum distance leach, and MDC-K protocols in terms of lifetime, throughput, stability, and latency time for different number of SNs.

FIGURE 12.

Comparison of the MDC-IR-DV-Hop-K approach, MDC-TSP-LEACH-K, MDC maximum residual energy EACH, MDC minimum distance leach, and MDC-K protocols in terms of lifetime, throughput, stability, and latency time in different area sizes.

The simulation results of this scenario revealed that varying the density of SNs maintains the same benefit and improvement in lifetime, throughput, stability, and latency. As illustrated in Fig. 12, as the number of nodes increases, the latency slightly increases but is still lower than that of the MDC-TSP-LEACH-K, MDC maximum residual energy leach, MDC minimum distance leach, and MDC-K protocols. In addition, the throughput of the MDC-IR-DV-Hop-K increases from 32 060 packets/wave in a density of 100 SNs to 34 169 packets/wave in a density of 2000 SNs in comparison with 11 302 packets/wave for MDC minimum distance leach in 2000 SNs, 19 868 packets/round for MDC maximum residual energy leach in 2000 SNs, 19 973 packets/round for MDC-K in 2000 SNs and 20 888 packets/round for MDC-TSP-LEACH-K in 2000 SNs. In contrast, the proposed protocol maintains its same lifetime benefits with the variation of SN density and is still the best compared to the MDC-TSP-LEACH-K, MDC maximum residual energy leach, MDC minimum distance leach, and MDC-K protocols. Thus, the MDC-IR-DV-Hop-K is the best solution for high SN density WSNs in terms of lifetime, throughput, latency, and reliability. However, in the case of stability, the MDC-K is still better than the MDC-IR-DV-Hop-K and MDC-TSP-LEACH-K protocols in terms of stability. Table 5 compares the proposed protocol and selected related protocols in the literature to evaluate the efficiency of the MDC-IR-DV-Hop-K protocol in enhancing the reliability of the routing protocol.

TABLE 5 The Comparison Analysis of Reliability for Cluster Head Election and Clustering of Our Protocol and Some Clustering Protocols in the Literature

SECTION VI. Conclusion

This paper proposes a new smart routing protocol called MDC-IR-DV-Hop, combining the IR-DV-Hop localization algorithm, K-means algorithm, and MDC. Specifically, this protocol uses the IR-DV-Hop localization algorithm to determine the accurate location of SNs without errors in LS-WSNs and determine the CHs. In addition, the MDC is used as an intermediate between the CH and BS to enhance the QoS criteria of WSNs, minimize time delays during data collection, and extend the WSN lifetime. The simulation results demonstrate that the MDC-IR-DV-Hop considerably influences energy consumption and QoS metrics. Particularly, this protocol significantly improves energy consumption, latency time, throughput, and stability gains compared to the EACBM, TCMDC, MDC maximum residual energy leach, MDC minimum distance leach, MDC-K, MDC-TSP-LEACH-K, AMWSN, and NDCM protocols. Moreover, the simulation results reveal that the MDC-IR-DV-Hop is more suited for LS-WSNs and HSND-WSNs. In future research, our interest is to investigate the MDC-IR-DV-Hop in large-scale mobile WSNs.