In this paper, a novel optimal relay angle-based clustering routing protocol called RACR is proposed to balance nodes’ energy consumption and prolong the network lifespan of wireless sensor networks (WSNs). In RACR, each node considers the parameters residual energy, number of neighbors, and average distance with neighbors as a criteria to determine whether to become a cluster head (CH) which cooperates with its neighbors to form a cluster. Afterward, the optimal relay angle is calculated for each CH to find its best relay node according to a target function considering the least energy consumption, which is applied to reduce the search range for finding routing paths. Consequently, all the CHs can find their best next-hop nodes within the determined field according to their residual energy, distance to the next-hop CH, and loads. Iteratively, the CHs obtain their best routing paths to the BS in the end. Simulation results demonstrate its effectiveness of RACR in terms of energy consumption, standard deviation of residual energy, data communication delay, network throughput, and lifespan.

1. Introduction

WSN is a self-organizing multihop network composed of many miniature sensor nodes with sensing, computing, data processing, and communication capabilities. The nodes sense and deal with the data and transmit it to the base station (BS) [1, 2]. Now, WSNs are widely used for environmental monitoring, location tracking, healthcare, space exploration, etc. [3]. However, in WSNs, each node’s energy provided by battery is limited, and the number and range of nodes deployed in the industrial, military, and marine environments are so large that they cannot be supplemented with energy at any time [4]. If the node’s power is exhausted, it will cause some areas to not be monitored, resulting in blind spots [5]. Therefore, reducing energy loss and prolonging the service lifespan of the nodes have become a critical matter in the use of WSNs. Clustering routing protocol for WSNs proposed by many scholars is the primary way to solve such problems [6].

Clustering routing protocols still have difficulties to show their superiority [7]. Although the clustering routing algorithms have solved some of the problems faced in WSNs, the CHs far away from the BS consume more energy than those nearer ones, causing a new energy imbalance problem, even they forward messages in a multihop manner [8]. Moreover, the CHs close to the BS need to undertake the task of receiving data in the cluster and forwarding data from the other CHs, which causes them die prematurely due to energy exhaustion [9, 10]. Once these nodes are dead, other nodes far away from the BS will lose connection with the BS.

Accordingly, this paper proposes an optimal relaying angle-based clustering routing protocol called RACR in response to the problems. The relaying angle indicates the range where a CH selects its relay nodes. Most clustering routing protocols do not give a definite search range for CHs, but select relay nodes in the entire communication range of CHs. This will not only increase time complexity but also improperly select relay nodes, which will increase energy consumption. The remaining energy, the quantity of neighbors, and the average distance between neighbors are the main criteria for selecting CH in RACR. During the routing process, RACR determines the best relay angle of each CH according to the distance from the CH to BS. Then, within the range constrained by the relay angle, the CH with larger energy, smaller load, and smaller distance from the previous-hop CH is selected as the relay node. The contributions of this paper are listed as follows: (i)A new threshold function with parameters residual energy, number of neighbors, and average distance between neighbors is defined to select the best CHs. Moreover, the weights for these parameters are constantly adjusted to fit the dynamic changes of the network(ii)Optimal relay angle is calculated for the first time to narrow the range of relay nodes based on the distance from CHs to the BS. Besides, residual energy, distance to the next-hop CH, and loads are considered to find the best next-hop relay node. So, hot issues and balanced energy consumption are solved in the end(iii)Simulations are performed to verify the effectiveness of RACR, and the results show that RACR has an average increase of 29.75.13%, 48.61%, and 68.13% in energy saving compared with the other three relative protocols and an increase of 18.16%, 30.75%, and 38.14% in throughput, respectively. And in terms of prolonging the network lifespan, it increases by 31.19%, 25.94%, and 59.14%, respectively

What follows is the remaining of this paper. In Section 2, many clustering technologies and routing protocols of WSNs are investigated and briefly analyzed. Section 3 introduces the system model, and Section 4 describes RACR in detail. In Section 5, simulation and analysis are presented. Finally, we summarize the paper in Section 6.

This section will review the clustering routing protocols for WSNs in the following three parts: (1) clustering, (2) energy-efficient routing, and (3) angle-based dynamic clustering and routing.

2.1. Clustering

At present, a large number of scholars have proposed numerous clustering methods, which can effectively cut down the network energy consumption and extend the network lifespan [11, 12]. Heinzelman et al. proposed LEACH in 2000, which was the first clustering protocol [13]. In each round, every node would be allocated a random number between 0 and 1. When its random number was blow the preset threshold, the node was selected as CH and broadcasts a message announcing its CH identity, and all the non-CH nodes responded to their nearest CHs to form clusters. Still, it did not consider the energy when opting CHs, which caused some nodes with less energy to be selected as CHs. It is likely to die prematurely due to excessive loads. In 2006, Qing et al. proposed a heterogeneous WSN distributed energy-saving clustering algorithm (DEEC), in which the threshold was the same as the LEACH algorithm. However, in DEEC, two different types of nodes are set according to the remaining energy, and different formulas were used to calculate the probability of becoming CHs [14]. The heterogeneous WSN cluster head restricted energy-saving routing protocol (CREEP) proposed by Dutt et al. in 2018 made two improvements on DEEC [15]. First, the impact of node residual energy was considered when setting the threshold. Second, in order to improve the energy-saving effect in the CREEP algorithm, a distance factor was introduced into the CH probability calculation, which guaranteed that the probability of a node with a large distance from BS becoming a CH was small.

In addition to some traditional algorithms, fuzzy control is also widely used in the improvement of cluster routing protocols [16]. Baranidharan and Santhi proposed a clustering method using fuzzy control (DUCF) in 2016 [17]. The protocol used a fuzzy control system with three input parameters and two output parameters. The energy left by node, the density of node, and the distance to BS were applied to compute the possibility of the node becoming a CH. However, in DUCF algorithm, the best quantity of CHs cannot be determined. Wang et al. proposed an adaptive clustering algorithm (APSA) based on affinity propagation in 2019. The protocol used the AP algorithm to determine how many CHs to be selected and perform preliminary clustering [18]. As a centralized algorithm, AP would increase its time complexity with the expansion of the network scale.

In addition to fuzzy control, AP and other methods were often used in solving WSN problems, and intelligent bionic optimization algorithms have also been continuously applied to this field, such as gray wolf optimization algorithm, lion optimization algorithm, particle swarm optimization algorithm, and elephant swarm optimization algorithm [19, 20]. In 2019, the OMPFM protocol with enhanced genetic algorithm proposed by Mohammed et al. made two pretreatments in the selection of CH to improve the execution time of the genetic algorithm and the efficiency of chromosome quality [21]. In addition, the use of genetic algorithms to find the optimal transmission path depended on the following four parameters: the distance from CH to the relay node, the distance between the relay node and BS, the quantity of CH, and the member nodes. The multiobjective fractional particle lion algorithm (MOFPOL) proposed by Bhardwaj and Kumar in 2019, which mainly used a multiobjective function constructed by energy, time delay, distance, and cluster density to select the best CH [22]. In 2020, Prahadeeshwaran and Priscilla proposed a hybrid elephant herd optimization algorithm (NIUS-HEHOA) based on an individual update strategy for CH selection [23]. It took advantage of the characteristics of the clan update and separation of the elephant group, so that after each algorithm update, the node with the higher function value can be selected as the CH, and the lower node can be separated from the cluster. However, the algorithm had a low convergence speed in the optimization process.

2.2. Energy-Efficient Routing

Although many of the above clustering algorithms can avoid the distance of nodes directly transmitting information to the BS, they also significantly increased the load of CHs [24, 25]. In some WSNs with an extensive deployment range, the routing protocol of the two-layer structure can no longer meet the application demand. For example, the large-scale WSN energy aware clustering hierarchy protocol (EACHP) proposed by Barati et al. in 2015 chose the best CHs by establishing a multiobjective function [26]. The choice of CH in the EACHP protocol was affected by four parameters: energy, the distance to BS, the distance from the neighbors, and the quantity of neighbors. The EACHP protocol ignored the energy consumption in the routing process, which caused the energy of nodes far away from the BS is quickly exhausted. Therefore, many multilayer routing protocols are proposed to solve the abovementioned problem. Multilayer routing protocol meant that CHs need to look for one or more relay nodes in the process of sending messages to the BS. In 2018, Neamatollahi and Naghibzadeh proposed an unequal clustering approach using fuzzy logic called UCF [27]. The two parameters of node degree and distance to BS are input into fuzzy logic system to obtain the clustering radius. And UCF sent a message to BS through a multihop route during the data transmission stage. Since UCF only relies on energy to elect CHs, the distance within the cluster is likely to be too large. In 2020, Chanak et al. proposed a routing scheme based on green clustering for WSNs of different densities [28]. A load-balancing data routing algorithm is proposed in the protocol to balance the network energy cost and extend the network lifespan. The energy-saving algorithm for LS-WSNs proposed by Jawhar et al. in 2019 applied the collection of boundary to monitor application data [29]. In this algorithm, CHs were selected in the light of remaining energy in each cluster. Nodes with higher energy are opted as cluster coordinators (CCOs) in the clusters around the BS to balance loads, and in highly dense clusters, relay nodes (RN) are selected to share the burden of data traffic. The energy-aware cluster-based routing protocol (ECRP) proposed by Noureddine et al. in 2020 adopts a multihop communication model [30]. ECRP elected CHs in the same way as LEACH. In addition, ECRP also introduced energy and distance to construct a multilayer routing protocol to minimize and balance energy consumption. Although this protocol establishes optimal routing paths, the load imbalance of CH is prone to appear in the clustering stage.

2.3. Angle-Based Dynamic Clustering and Routing

The existing multipath protocols typically employ an on-demand routing approach to search for paths. The source node broadcasts a routing request message into the network, and the IDs of the visited intermediate nodes are appended into the messages. However, broadcast message is a wasteful approach of routing discovery, so relay angle is introduced accordingly. Similarly, in 2010, Taoy et al. proposed an adaptive energy-aware multipath routing protocol (AEMRP-LB) with load balancing [31]. It introduced a concept named direction-angle to overcome the deficiency of broadcast and took into account the trade-off between the residual energy and hop count to establish multiple node-disjoint paths. However, it was prone to CH loads imbalance during the clustering stage. In 2017, Jia et al. proposed two novel methods: a privacy preservation protocol for source location in WSN based on angles (APS) and an enhanced protocol for source location (EAPS), which dynamically adjusted emission radius during routing [32]. The APS protocol produced geographically dispersed phantom source nodes and utilizes the energy from the energy-abundant regions to make the routing path versatile among the entire network. In the EAPS protocol, a node adaptively adjusts its radius according to the number of adjacent nodes, energy, and distance to BS.

In addition to the introduction of the angle parameter in routing, the angle is also applied to the CH contention. In 2020, Zhang proposed a multihop routing protocol using cellular virtual grid in IoT environment [33]. In the CH selection stage, the node angle ratio, distance ratio, and throughput optimization threshold function were introduced to select CHs independently. In the routing protocol, the path cost from the intermediate node to the target node was calculated according to the distance and residual energy. However, this routing protocol may cause uneven load on relay nodes. In order to alleviate above problems to a certain extent, this paper proposed a novel optimal relay angle clustering routing protocol which balances the network energy consumption during both the clustering and routing processes by defining a new threshold function to select the best CHs and limiting the range of the relay nodes to quickly find the optimal routing paths.

3. System Model

3.1. Network Model

WSN is composed of many tiny nodes randomly distributed in the target area who are responsible for collecting the monitored data. The data is usually sent by the source nodes, received by the CHs as relay, and forwarded to the next-hop nodes. Finally, the data will be transferred to the BS.

In this paper, nodes are distributed in the rectangular area of , and the BS is situated inside the network. Simultaneously, in order to simply simulate and analyze this proposed protocol, we make the undermentioned hypothesis of WSN: (i)The nodes are freely distributed, and their positions are stationary after deposition(ii)Each node has the same composition and the same initial energy assignment(iii)Each node has own ID for identification(iv)The capabilities of the BS are unlimited

3.2. Energy Model

In a WSN, the energy consumption for completing a round of information collection comes from two parts: source node to CH and CH to the BS. On account of a mass of nodes and the large monitoring area, the system’s energy is mainly spent on data transmission. According to the different required transmission distances, two-channel transmission models like in UCF [26], EACHP [27], and ECRP [30] are used to calculate energy consumption. When the communication distance is below the threshold, the free space transmission model is used to calculate the node’s energy consumption to transmit information. When it is higher than the threshold , the multipath fading channel transmission model is used to calculate the energy consumption. The energy consumption model is shown in the following formula. where donates the size of the data package, represents the energy per bit loss sent or received by the wireless communication module, and and are the power amplifier parameters in the two-channel transmission models. is the threshold distance of the transmission information, which is determined by the parameters of the two power amplifiers [34], .

Each node also needs to spend some energy when receiving data. The energy consumed by getting -bit data packets is shown in the following formula.

After the CH receives the information of its members in the cluster, it needs to fuse the data. is the energy required to fuse 1-bit data. The energy consumed by the CH fusing -bit data is shown in the following formula.

The energy consumed in transmitting a -bit data packet from node to its CH can be expressed as

The energy consumed by each CH can be calculated by the energy consumption of the received data, the energy consumption of fusion, and the energy consumption of transmission to the relay node or BS, which can be expressed as formula (5), where represents how many nodes the CH receives data from.

Therefore, the total energy consumption of the network in the clustering and routing stages can be expressed as formula (6), where is the number of clusters.

4. Proposed Protocol

This section introduces RACR in detail, which mainly solves the problems of balancing network energy consumption, alleviating hot issues, and shortening the data transmission distance in the network. Then, the three stages of RACR are discussed in sequence: initiation, clustering, and data sending.

4.1. Initiation

In the system initialization phase, each node is deployed at random in the network. After all nodes are placed, the nodes exchange information with their neighbor nodes until all nodes receive the ID, location, energy, and some other information. After that, the system starts to cluster [35].

4.2. Clustering

In clustering stage, RACR selects its CHs based on three factors of energy, distance, and node density. Let be the target function of electing CH. In the formulation of the target function, we mainly consider the following three indicators.

The remaining energy of the node: during the operation of WSN, CHs need to receive the data sent by its member nodes and forward it to the BS. It can be seen that CHs need to consume more energy, so nodes with greater energy can be selected as CHs. The expression for energy is as follows: where is the remaining energy of the node, and denotes the initial energy of the node.

Node density: the quantity of neighbor nodes within the communication radius. Nodes with high density and more neighbor nodes have a greater probability of being selected as CHs. The expression of density is shown as follows:

In the above formula, is the amount of neighbor nodes between the communication range of the node, is the communication radius of the node, and is the amount of nodes deployed in WSN.

The average distance from neighbor nodes: since energy consumption is mainly in the transmission process, reducing the transmission distance is important in decreasing energy consumption. Therefore, in addition to selecting nodes with many of the neighbor nodes in CH, the average distance to neighbor nodes should be reduced as much as possible. The expression of average distance is given as follows: where is the distance from the node to its neighbor node.

Finally, considering the above parameters, the target function of all surviving nodes is defined as follows:

The target function comprehensively considers node energy, distance, and density in the clustering process. More importantly, the node density and distance are fixed during the clustering process, so increasing the influence of energy on the selection of CHs can prevent some nodes from being continuously selected as CHs, resulting in their premature death. The coefficient of energy in the target function should be greater than the coefficient of density and the coefficient of distance. It should satisfy and . In addition, as the frequency of network operations increases, the total remaining energy continues to decrease. Therefore, in order to improve the balance of energy consumption, the influence of energy factor in the CH election process should be continuously increased as the energy decreases. Similarly, the distance within a cluster plays a significant role in network, and this parameter needs to be assigned a greater weight. In this paper, we will set the coefficients of in the target function according to the remaining energy of the network as: , , and . is the proportion of all energy left by the node to initial energy in each round, .

Finally, the target function formulated is as follows:

After completing the calculation of the target function, the node with the largest function value is selected as the first CH, , then the message that it becomes a CH will be broadcast within the radius of the average distance from the neighbor node, and nodes that receives the message become common nodes. Next, choose the second CH with the largest value among the remaining nodes. Similarly, nodes within the radius of the average distance between and its neighbor nodes become common nodes; and so on, until all nodes have received the message of being selected as CHs or common nodes, the CH election is completed. After the CHs are selected, the remaining nodes send messages to their nearest CHs to apply for joining. When all CHs receive the information sent by their members, the clustering phase ends.

4.3. Data Sending
4.3.1. Intracluster Communication

After the cluster phase is completed, the cluster members will transmit their monitoring message to their corresponding CHs. Only the nodes being transmitting data are in the active mode, and the other nodes are in the sleep mode. This method can not only save energy but also effectively avoid direct conflicts between nodes. After CHs receive and merge the information from each member nodes, they forward the information to the BS.

4.3.2. Intercluster Communication

As shown in Figure 1, with the CH as the center and the distance from the CH to BS as the radius, each CH looks for relay node in the sector whose angle is with to transmit messages to BS. The calculation of the optimal relay angle is given as follows.

RACR uses the distance between CHs and BS to calculate the optimal range for finding relay nodes. Among them, it is assumed that the distance from the CH to the relay node is equal to the distance from the relay node to the BS, namely, .

If the relay node is not used, the CH sends the data straight to the BS. Under the circumstances, the energy consumed is shown in the following formula:

If a relay node is used for forwarding, the total energy consumed by the CH to forward information to the BS through the relay node is shown in formula (13).

Therefore, to reach the purpose of saving energy, the energy consumption of transmission using relay nodes should be lower than the energy consumed of direct communication, which is . Therefore, the optimal range angle for CH to choose the relay node can be obtained as shown in the following formula.

It can be clearly seen from formula (15) that the optimal search angle for each CH is determined by the distance between the CH and BS. When the angle between the line from the previous-hop CH to the next-hop CH and the line from the CH to the BS is less than , the CH is a candidate relay node.

In all the relay nodes that meet the condition, RACR uses the remaining energy, distance, and load to choose the appropriate one as the next hop node. The relay node not only needs to receive its own member information and forward it to the BS but also needs to transmit the message of the previous-hop CH to the BS. It can be seen that at the stage of sending information, the number of being selected as relay nodes of a CH determines its energy consumption. Therefore, in response to the demand for balancing energy consumed and prolonging lifespan in WSN, nodes with larger remaining energy and smaller loads should be selected while using relay nodes to reduce the transmission distance. According to the above points, RACR formulates a weight function of candidate relay nodes. The relay node selected has the maximum weight, and the weight function is expressed as follows: where represents the energy left by the next-hop CH, represents the distance between the previous-hop to the candidate relay node, and is the load of the candidate relay node.

4.4. Complexity Analysis

The time complexity of RACR is . For cluster formation, each non-CH node needs to process CHs to select its CH, where is the number of nodes. The RACR clustering algorithm sorts nodes according to their fitness values, and the worst-case computing power for sorting is (). For routing, each CH needs to calculate the cost value of the next-hop CH, and in the worst case, it needs to process -1 CHs, where indicates the number of CHs. Therefore, the time complexity of the RACR routing algorithm is (). Moreover, is less than , so the time complexity of RACR is .

In addition, the space complexity of RACR is a measure of the size of the temporary occupied space during the operation of the algorithm, which is represented by . In each round of clustering, if a node is selected as CH, the node broadcasts a message of becoming CH to its neighbor nodes, and if it is a non-CH node, it sends a message as acknowledge to join a CH. Therefore, for a network with sensor nodes, the message complexity is . Also in the routing process, each CH broadcasts a message to select a relay node to the CHs in the relay corner. The message complexity of selecting the next hop CH (relay node) is (). So the space complexity of RACR is .

5. Simulation Results

This section will conduct the simulation experiments of RACR in MATLAB 2019a. Four performance indicators, namely, the number of surviving nodes, the total energy consumption of the nodes, the network throughput, and the standard deviation of network remaining energy and delay are tested through many experiments and compared with EACHP, UCF, and ECRP clustering routing protocols proposed in [26, 29, 30], respectively.

5.1. Experimental Setup

In the experiments, the location of the BS and the number of nodes will affect the experimental results. Therefore, we set up four different scenarios to test the proposed protocol. In scenario 1, the area of the network is 400 m400 m, the coordinate of the BS is (200, 200), and 100 nodes are randomly distributed on the site. In the second scenario, there are 200 nodes randomly distributed. In scenario three, the coordinate of BS is (0,0), and the number of nodes is 100. Finally, in scenario 4, the area of the network is expanded to 600 m600 m, and the number of nodes is 200. The initial energy of the nodes is set to 2 J in all scenarios. Other values used for the simulation parameters are shown in Table 1.

5.2. Network Lifespan

The network lifespan directly relates to the quantity of surviving nodes. The number of surviving nodes in the network corresponds to the quantity of dead nodes in the network. The network lifespan can be directly measured by the rounds of death of the first node, half of nodes, and 80% of nodes in the network. Table 2 shows the simulation results of the first node death rounds (FND), half of nodes death rounds (HND), and 80% of nodes death rounds (END) for different protocols. In addition, the simulation results of the number of surviving nodes in the network are shown in Figure 2.

As shown in Table 2 and Figure 2, RACR, EACHP, ECRP, and UCF protocols have different lifespan due to their different working principles. EACHP adopts a single-hop routing model, and nodes far from the BS will die first in a large-area network. Although ECRP and UCF adopt multihop routing, the random rotation of CHs in ECRP will cause the CHs to be located at the network boundary or with low energy, which will accelerate the death of nodes. The UCF does not consider the energy of the next-hop CH in routing, so the overload of the relay node will cause the CHs close to the BS to die first. The RACR proposed in this paper adopts a multihop protocol, measures the distance and energy in the CHs selection, and selects the relay nodes through the relay angle to reduce the transmission and effectively balance the load of the CH, thereby extending the network lifespan as a whole.

It can also be seen from Figure 2 that the first node of EACHP dies faster, and the remaining nodes close to the BS have a longer lifespan. Nodes in ECRP die evenly, but at a higher rate than RACR. However, in the later stage of UCF operation, the nodes with a large distance from the BS die quickly due to the loss of relay nodes. Finally, the distance within the RACR cluster is small, the CH distribution is uniform, and the load balance causes each node to die evenly and have a long lifespan. In addition, from the experimental data, it can also be concluded that half of the nodes of RACR died in 1035, 1112, 309, and 563 rounds, respectively, which are 25.70%, 62.85%, 74.11%, and 73.89% than EACHP. The lifespan of RACR is 5.79%, 14.92%, 28.80%, and 54.52% higher than that of ECRP algorithm, and it is 27.43%, 35.43%, 18.77%, and 43.16% higher than that of UCF algorithm, respectively.

5.3. Network Energy Consumption

The lifespan of a network is also inversely proportional to the energy spending of nodes. The smaller the energy spending, the smaller the transmission distance used when the network is running. The change of total network energy consumption is shown in Figure 3.

As the frequency of CHs rotation in the network increases, the energy consumption of the network continues to increase. Among them, the CHs in EACHP communicate directly with the BS, so the energy consumption will increase rapidly with the expansion of the network area. The ECRP protocol uses the fitness function composed of the distance from CH to BS, the distance from node to CH and energy to cluster, which leads to the large transmission distance in the cluster and more energy consumption in each round. UCF can reduce the transmission distance from CHs to the BS through the relay node, but when the node closer to the BS dies, the energy consumption will rise rapidly. The RACR protocol selects nodes with higher node density as CHs to reduce the intracluster distance and selects relay nodes according to energy, distance, and load to ensure that relay nodes do not die prematurely due to excessive load. Therefore, the network energy consumption of RACR is small and can rise steadily.

Seen from Figure 3, when the dead nodes do not exceed 80%, the curve of EACHP, ECRP, and UCF algorithms remains above that of RACR. In the experiment, EACHP consumes half of the energy in 301, 235, 27, and 67 rounds, ECRP in 349, 336, 104, and 132 rounds, and UCF in 394, 444, 191, and 171 rounds. In RACR, the network energy consumption reaches 50% when CH rotations are 567, 538, 241, and 344, respectively. It can be drawn from these data, in terms of reducing energy consumption, compared with EACHP, ECRP, and UCF protocols, RACR has increased by 68.13%, 48.61%, and 29.75% on average, respectively. It can be seen that RACR has more evident advantages in saving energy.

5.4. Network Throughput

Network throughput indicates the amount of message transmitted by nodes to the BS. A higher amount of data transmission indicates a higher utilization rate of energy in the network. Figure 4 shows the changes in network throughput.

As can be seen from Figure 4, for all protocols, their network throughput increases continuously with the number of epochs the simulation runs. For EACHP, due to its single-hop communication method, the data will be lost when the CHs are too far from the BS. So its total network throughput is the worst. Moreover, RACR has a longer network lifetime compared with ECRP and UCF using multihop communication scheme, so the more alive nodes can send more data to the BS to increase the network throughput. Therefore, RACR achieved the highest total network throughput, which is 11.83%, 22.02%, 65.01%, and 53.72% higher than EACHP, 12.07%, 15.25%, 43.45%, and 52.23% higher than ECRP, and 16.56%, 14.82%, 18.93%, and 24.14% higher than UCF in the four scenarios, respectively. The results tell that RACR not only effectively saves energy consumption but also ensures the total amount of data transferred.

5.5. Standard Deviation of Residual Energy

The remaining network energy is a crucial indicator for measuring the network lifespan and energy expenditure, and its standard deviation can further calculate the balance of network energy consumption. Figure 5 gives the standard deviation of the remaining energy of the network.

As shown in Figure 5, the standard deviation of the remaining energy of RACR is smaller than that of EACHP, ECRP, and UCF. Moreover, the average residual energy standard deviation of RACR rises slower than those of the others. This can further illustrate that RACR balances the load and energy consumption of the nodes. In the four scenarios of the experiment, the average values of the residual energy standard deviation of EACHP are 0.4164, 0.4526, 0.6489, and 0.5530, respectively. The average values of ECRP are 0.3162, 0.3615, 0.5359, and 0.4947, respectively. The average values of UCF are 0.2145, 0.3278, 0.4459, and 0.3444. The average residual energy standard deviations of the RACR protocol proposed in this paper are 0.1566, 0.3044, 0.3244, and 0.2699, respectively, which are reduced by 20.3%, 37.41%, and 48.95% on average compared to EACHP, ECRP, and UCF.

5.6. Data Transmission Delay

The data transmission delay indicates the time difference between source nodes and the BS for data transmission. Moreover, data transmission delay is one of the important criteria to measure the performance of routing algorithm. The smaller the average delay, the better the efficiency of data transmission. The data transmission delay of RACR is shown in Figure 6.

Compared with ECRP, RACR has a smaller intra-cluster transmission distance. Compared with UCF, RACR can reduce the transmission distance and save some unnecessary multihop transmission through the relay angle, so the data transmission delay of RACR is smaller than that of ECRP and UCF. However, the delay of RACR is slightly larger than that of EACHP, because EACHP uses a two-layer protocol for data transmission, which reduces the processing delay of the network. The simulation results in Figure 6 also show that the transmission delay of RACR is 20.9%, 22.4, and 11.8% lower than ECRP in scenarios #1, #2, and #4, and 21.5%, 28.4%, and 11.9% lower than UCF in scenarios #1, #2, and #3, respectively.

6. Conclusion

In this paper, a novel relay angle-based clustering routing protocol called RACR is presented to balance the network energy consumption and improve the energy efficiency. RACR selects nodes with greater energy and density as the CHs to form clusters and calculates the best relay angle of each CH to find the optimal relay node according to distance, energy, and load within this angle range. Simulation results have shown RACR outperforms EACHP, ECRP, and UCF. In the aspect of network lifespan, RACR increases 59.14%, 25.94%, and 31.19% compared with EACHP, ECRP, and UCF, respectively. In terms of reducing energy consumption, RACR improves 68.13%, 48.61%, and 29.75%, respectively, compared with EACHP, ECRP, and UCF. In terms of network data throughput, RACR enhances 38.14%, 30.75%, and 18.16%, respectively, compared with EACHP, ECRP, and UCF. In a word, RACR has apparent advantages in equilibrating network energy and prolonging network lifespan and has a good scalability for the networks in different size.

Due to its excellent performance, RACR can be widely used in many applications such as indoors environmental monitoring, coal mine monitoring, hospital personnel and equipment monitoring, and earthquake monitoring. However, relatively idealized model and only simulation validation also restrict the practical application for RACR. Therefore, the future work will be carried out from the aspects of practical system model for real environments and live verification.

Data Availability

No data were used to support this study.

Conflicts of Interest

The authors declare that they have no conflicts of interest.


“This research was funded by the Foundation of the Department of Science and Technology of Jilin Province, China, with grant number 20210201051GX.”