Cooperative Sensor Selection Optimization and Power-Efficient Gathering for Multihop Wireless Sensor Networks

Abstract

Since a single sensor may be in a fading dip, cooperative sensing among multiple sensors which experience uncorrelated fading is required to guarantee reliable sensing performance. This paper proposes a cooperative sensor selection and power-efficient gathering (CSSPEG) strategy in multihop wireless sensor networks to avoid the “hot spot” problem. Intracluster communication radiuses have different sizes, in which the clusters closer to the base station have smaller size. Furthermore, intracluster member allocation scheme uses cooperative sensor selection and power-efficient gathering strategy. Simulation results show that the proposed CSSPEG can efficiently balance the energy consumption of cluster heads and prolong the lifetime of the network.

1. Introduction

Wireless sensor networks (WSNs) are usually composed of hundreds or thousands of sensing devices, which are deployed over a geographical area for the purpose of target tracking and monitoring physical phenomena [1]. In recent years, WSNs architectures involve many power-limited sensor nodes to enable a large class of applications, such as environmental and building monitoring, pollution monitoring, agriculture, health care, energy management, and eruption monitoring. However, wireless sensor nodes are highly energy-constrained devices, and the sensor is usually powered by battery which is undesirable or even impossible to be recharged or replaced. Collaborative techniques and energy consumption are both critical issues in WSNs.

For reducing energy dissipation and enhancing system lifetime effectively, many clustering algorithms have been proposed. In WSNs, the network is divided into several clusters and one node is selected as the CH (cluster head) in each cluster [2]. The members in each cluster transmit data to their respective CH and the CH in turn forwards the data after aggregating/fusing to sink node through single-/multihop transmission. When the nodes use single-hop communication to reach the sink node located farther away, it will lead to the high burden which is due to long-range communication. A power-efficient solution aims to minimize the communication distance [3], which makes multihop communication essential in case of real life deployment. Data is relayed hop-by-hop by the intermediate node rather than communicating directly with a possibly distant sink node. However, with regard to the multihop transmission, it increases the load on those located near the sink and spares the energy of distant sensors. Therefore, if the energy consumption of sensor nodes is not managed properly, it may result in “hot spot” problem in the network [4].

The dynamic sensor selection optimization strategy refers to the optimization of the tradeoff between energy consumption and effective coverage rate. Moreover, it is usually adopted to enhance energy efficiency and prolong the lifetime of WSNs. The purpose of this paper is to reduce energy consumption through the use of collaborative technologies. Energy consumption and collaborative techniques are two of the most severe constraints of WSNs due to their intrinsic nature. WSNs using cooperative sensor selection optimization technology and power-efficient gathering strategy can effectively reduce energy consumption. In this paper, we propose a dynamic sensor selection optimization and power-efficient gathering strategy for multihop wireless sensor networks to avoid the “hot spot” problem. Intracluster communication radiuses have different sizes; those closer to the base station (BS) have smaller size. On the other hand, a dynamic sensor selection strategy is presented for intracluster member allocation, which restricts the number of nodes with higher burden. Therefore, the CHs of these clusters can preserve more energy for the intercluster relay traffic. Simulation results show that the proposed CSSPEG can efficiently balance the energy consumption of CHs, decrease the dead speed of the nodes, and prolong the lifetime of the network.

The main contributions of this paper include the following: (1) a brief review about the advantages and disadvantages of various existing cluster-based routing protocols in WSNs is presented; (2) a cooperative sensor selection and power-efficient gathering strategy in multihop wireless sensor networks is proposed; (3) an effective cooperative sensor selection optimization model is proposed.

The remainder of the paper is organized as follows. In Section 2, we briefly review related works. Section 3 describes the power-efficient gathering model. Section 4 presents the details of CSSPEG and argues the choice of its parameters. Section 5 shows the experimental simulation results. Finally, Section 6 concludes the paper and discusses some future research directions.

2. Related Works

In recent years, a number of protocols have been proposed to manage and reduce the energy consumption of sensor nodes. In this section, we give a brief review about various existing cluster-based routing protocols in WSNs.

Some protocols require the CH in any cluster to be able to communicate with its neighboring CHs in one hop. Thus, a virtual grid is built, such as BP [5] or EEPA [6]. It is important for the above algorithms to keep the number of clusters as low as possible and the optimal clustering is often defined as the one that minimizes the number of clusters. HEED [7] protocol proved that it can minimize the control overhead and prolong the lifetime of network which is due to the well distribution of CHs. Also, HEED protocol considered the residual energy of sensor nodes and the cost of communication within the cluster during the selection of CHs.

An energy-efficient heterogeneous clustered scheme was proposed in [8], which was effective in prolonging the lifetime of WSNs. The impact of heterogeneity of sensor nodes was studied in terms of their energy, and the residual energy of each node is chosen as a key factor in the process of calculating the weighted election probabilities for CH selection. A novel distributed clustering algorithm was proposed in [9], in which CH nodes were selected following a three-way message exchange. The eligibility to be elected CH was based on its residual energy and degree. An energy-efficient clustering protocol (EECPL) was proposed in [10], which organized sensor nodes into clusters and used the ring topology to send data, so that each sensor node can receive data from a previous neighbor sensor and can transmit data to a next neighbor.

An energy-efficient routing technique, called energy-aware intracluster routing (EAICR), was proposed in [11], which has increased energy efficiency up to 17% and increased the network lifetime up to 12%. EAICR focused on the energy consumed by the packets sent in the network and the residual energy level of sensor nodes at specific time. An efficient approach for clustering aggregation based on data fragments was proposed in [12], which can improve the network performance and reduce the computational complexity.

To minimize the total energy consumption in the network, the expressions for the required CH densities have been studied in a multihop wireless sensor network [13]. However, they did not provide any justification for choosing multihop mode for communication between the sensor nodes and the CHs. It was one of the rare efforts to match the size and shape of clusters to the gathered sensor data: nodes form clusters only if their data were similar and can be aggregated with little or no data loss [14]. In terms of the cluster form, the algorithm adopted a traditional k-hop clustering with random cluster sizes due to the requirement of data similarity. Since the intermediate CH of the routing path spent more time listening or sending data and then consuming more energy, it would exhaust batteries more rapidly than other nodes.

In order to balance the energy level or energy consumption of CHs, an unequal clustering model was first proposed by Soro and Heinzelman in 2005, which was based on unequal clustering size (UCS) [15]. In the case of multihop networks, it was demonstrated that UCS was 10–30% better than existing equal clustering models.

In the energy-efficient unequal clustering (EEUC) algorithm, Li et al. proposed a novel unequal clustering algorithm in which nodes joined clusters of unequal size [16]. In EEUC, even clusters equidistant from the BS may have different number of member nodes, neither too many nor too few. However, EEUC may produce lone nodes since the CH election was random [17]. In [17], a multihop routing protocol with unequal clustering (MRPUC) was proposed. The experimental results demonstrated that MRPUC outperformed an equal clustered version of itself by extending the network lifetime by 34.4%. Zhao and Wang proposed an unequal layered clustering approach (ULCA) for large scale wireless sensor network, which assumed a BS at the center of the grid [18]. The layers closer to the base station were smaller in size giving the inner layers more residual energy for intercluster traffic. When compared to EEUC, ULCA had a better network lifetime and the overhead for clustering the network was much lower, because of the inherent local join and local broadcast mechanism. Bagci and Yazici proposed an energy-aware fuzzy unequal clustering algorithm (EAUCF) which used 2 parameters in order to calculate the competition range of the CH [19]. From the experimental results, EAUCF outperformed EEUC, ULCA, and LEACH for all the performance measures. Liu et al. proposed IEEUC, which was similar to the study in EEUC, since it created unequal sized clusters as they were further away from the base station [20]. The main difference between IEEUC and EEUC lied in the competition radius calculation. IEEUC used the node degree factor, which was based on the number of hops to the base station, to calculate the competition radius.

A new approach was proposed by Jaichandran et al. to mitigate the hot spot problem in WSNs in which an area S is divided into N cells, each of them having at least 1 sensor node, and the sensor nodes cooperated with neighbor nodes to forward sensed data to the base station [21]. Sensor nodes near the base station acted as gateway nodes (G nodes) which tended to die earlier as they had to relay heavy traffic. In [22], an energy-driven unequal clustering (EDUC) algorithm was proposed which discussed the rotation of the role of CH based on either time-driven CH rotation or energy-driven CH rotation approach. In EDUC unequal clusters were formed by having unequal competition ranges. One drawback of EDUC is the used single-hop transmission for intracluster communication. To solve the hot spot problem, Wan et al. proposed an energy-efficient unequal clustering algorithm (EUCA) with the ideal of unequal clustering in the circle area in which the CHs were in charge of different geographical scope according to different distance to the base station [23]. It could reduce the number of cluster members and led to the proportional energy dissipation in each layer.

3. The Power-Efficient Gathering Model

In this section, the power-efficient gathering model in wireless sensor networks is described. The proposed model consists of three parts: the transmitter, the power amplifier, and the receiver.

The energy spent for transmission of an l-bit packet from the transmitter to the receiver at a distance (d) is defined as

\begin{matrix} E_{Tx} = \{\begin{cases} l E_{elec} + l ε_{fs} d^{2}, & d < d_{0}, \\ l E_{elec} + l ε_{mp} d^{4}, & d \geq d_{0}, \end{cases} \end{matrix}

(1)

where the electronics energy

E_{elec}

depends on the digital coding, modulation, filtering, and spreading of the signal.

ε_{fs} d^{2}

and

ε_{mp} d^{4}

are the amplifier energies that depend on the distance to the receiver and the acceptable bit-error rate. The crossover distance

d_{0} = \sqrt{ε_{mp} / ε_{fs}}

Energy consumption to receive procedure and energy for data fusion can be calculated as

\begin{matrix} E_{Rx} = l E_{elec}, \\ E_{fx} = m l E_{DA}, \end{matrix}

(2)

in which m is the number of messages and

E_{DA}

is the energy dissipation per bit for fusion.

Assume that there are N sensor nodes, which are uniformly dispersed within a circular region of radius R. There is a sink without energy constraint in the geometric center of the area.

The network is organized into a clustering hierarchy, in which a set of sensor nodes is grouped and coordinated by a CH. The CHs collect the data and execute fusion function to reduce correlated data produced by the sensor nodes within the clusters [24]. Hence, CHs are required to possess high energy source. When the network becomes large, most of the CHs in the network may not reach the sink directly. Then they transmit packets to the sink in a multihop path via other CHs. To avoid the frequent change of the topology, we assume that the nodes are stationary as supposed in [25].

In [26], simulation results clearly showed that CHs nearer to the BS had lower residual energy compared to CHs further away. To maintain the connectivity of the entire network, it is very important that the CHs closer to the BS should keep alive as long as possible for the intercluster communication. So the number of the nodes in the clusters closer to the BS ought to be smaller than those farther away from the BS. Therefore, we assume that the circular area is divided into M rings $P_{1}, P_{2}, \dots, P_{M}$ , and the nodes are deployed with uniform density ρ (nodes/m²). $d_{i} (1 \leq i \leq m)$ is a one-hop distance for intercluster communication. The sensor network model is shown in Figure 1. We assume the sink is at $(x_{s}, y_{s})$ and the Euclidean distance with any node $(x_{i}, y_{i})$ can be calculated by $d = \sqrt{(x_{s} - x_{i})^{2} + (y_{s} - y_{i})^{2}}$ approximately.

Figure 1

The deployment of sensors in the network.

4. The Proposed CSSPEG Strategy

We characterize key properties of the transmission ranges of CHs by numerical analysis and quantitative methods. The CHs near the sink node are responsible for forwarding the data to the sink in combination with the remoter cluster and their own; thus they usually consume more energy than remote heads. As a result, the nodes near the sink becoming invalid are earlier than the remote ones and lead to network segment. To achieve energy evenly in the network, we divide the circle into concentric rings of thickness R (see Figure 2) and partition all nodes into clusters with unequal size, in which the clusters closer to the base station have smaller size. To balance the energy consumption between all CHs, we propose an intramember allocation scheme, which restricts the number of nodes in some clusters with higher burden. We also propose a communication model which is cost-effective for intercluster transmission.

Figure 2

An illustrative example of different partitions.

4.1. Cooperative Sensor Selection Optimization

The task of CH is to receive the data from the various member nodes of the cluster and send the data to the next hop and forwards messages from neighboring CH. The CHs have to spend extra energy for aggregating data and performing long-range transmission to the distant sink via the intercluster forwarding. Energy consumption for CH is as follows:

\begin{matrix} E_{ch} = (n - 1) l E_{rec} + n E_{aggr} + E_{trans} + E_{rel}, \\ E_{trans} = l E_{elec} + l ε_{mp} d_{next-ch}^{4}, \\ E_{rel} = l E_{rec} + l E_{tran} + l ε_{mp} d_{next-ch}^{4}, \end{matrix}

(3)

where n denotes the number of nodes in a cluster,

E_{rec}

is the energy consumed by receiving data from the nodes in the cluster,

E_{aggr}

is the energy consumed by fusing data,

E_{trans}

denotes the energy dissipated by transmitting intracluster data to the CH in the next-hop routing, and

E_{rel}

is the energy dissipated by forwarding the messages from the upstream neighbor CH.

In [27], Bhardwaj et al. proved that the hops must be equidistant for convex radios and proposed a solution of optimal routing in a linear sensor network. The question is how to relay data from a sensor, located at distance D from the BS. The optimal number of hops $(k_{opt})$ is always one of

\begin{matrix} k_{opt} = ⌈\frac{D}{d_{char}}⌉ or ⌊\frac{D}{d_{char}}⌋, \end{matrix}

(4)

where

d_{char}

is the characteristic distance given by

d_{char} = \sqrt[γ]{(2 E_{elec} + E_{DA}) \times (γ - 1) / ε_{mp}}

The optimal routing path is shown as an iterative sequence in Figure 3. Note that sensors do not necessarily transmit to their direct downstream neighbor; some nodes can be left out between successive relays. In Figure 3, the distance to the next-hop node is the characteristic distance.

Figure 3

The routing path with the optimal number of hops.

Therefore, the $d_{next-CH}$ in 3 can be obtained from the following equation:

\begin{matrix} d_{next-CH} = \frac{d_{toSink}}{k_{opt}} . \end{matrix}

(5)

Hence, the energy depletion for CH in partition i which relays data from the upstream neighbors to the next-hop node is

\begin{matrix} E_{rel} (i) = \sum_{j = 1}^{ch (i)} n_{i, j} \times E_{rel}, \end{matrix}

(6)

where

n_{i, j}

denotes the number of upstream node j of CH in partition i and

E_{rel}

is the energy consumption for data forwarding. The average energy consumption of the CHs in partition i is defined as follows:

\begin{matrix} {\hat{E}}_{ch} (i) = E_{ch} + E_{rel} (i) . \end{matrix}

(7)

To ensure cluster sensor network connectivity, the transmission radius of CH is larger than member nodes. The approximate ratio of $E_{rel}$ and $E_{non-ch}$ can be considered as

\begin{matrix} \frac{E_{rel}}{E_{non-ch}} = η, 1.5 < η < 2.5 . \end{matrix}

(8)

Therefore, the total energy consumption by the CH in partition i is given by

\begin{matrix} E_{ch} (i) \approx E_{ch} + \frac{η \sum_{j}^{{ch}_{i}} n_{i, j} \times E_{non-ch}}{n_{i}}, \end{matrix}

(9)

where

n_{i}

is the number of CHs in partition i.

Since CHs far away from the sink transmit their data packets via intermediate CHs in a multihop fashion to the sink node, these intermediate CHs get overloaded and drained out sooner, especially for CHs which are closer to the sink node. In order to handle the load and increase the network lifetime, the transmission range of the CH could be adjusted adaptively. The corresponding transmission range of the cluster heads near the sink is rather short and the range increases gradually as the distance between the sink and CHs of other partitions increases. Due to the number of partitions determined by the transmission radius, the transmission radius of all partitions may be satisfied $Tx (1) < Tx (2) < \dots < Tx (M)$ . It is assumed that the sensors should be evenly deployed within the environment; the density of the sensors is equal to

\begin{matrix} ρ = \frac{N}{π R^{2}} . \end{matrix}

(10)

The number of neighboring nodes in the range of the normal nodes can be calculated as

\begin{matrix} n^{'} = ρ \times π r^{2} = \frac{N r^{2}}{R^{2}} . \end{matrix}

(11)

Therefore, the transmission radius of the CH in partition i is given by the following equation:

\begin{array}{l} Tx (i) & = \sqrt{\frac{ρ}{n^{'} - (η / n_{i}) \sum_{j = 1}^{{ch}_{i}} n_{i, j}}} \\ = \sqrt{\frac{N}{π (N r^{2} - (η R^{2} / n_{i}) \sum_{j = 1}^{{ch}_{i}} n_{i, j})}} . \end{array}

(12)

4.2. Power-Efficient Gathering Strategy

To increase the lifetime of WSNs by efficiently balancing energy among nodes, we mainly focus on maintaining a balance of energy consumption of all CHs in different partition. We determine the number of members in each cluster in accordance with the CH node energy consumption equal as much as possible. The resolution for the above question is to allocate a number of limited nodes within each cluster and then append additional nodes according to the burden of CH adaptively.

Let $σ_{i}$ denote the compression ratio of the CH in the ith partition. For convenience, we assume that the periodical overloads produced by all nodes are equal and can be defined as λ (bit/second).

Then the data from cluster i which will be forwarded to the next CH can be expressed as

\begin{matrix} Data (i) = λ N_{i} σ_{i} + λ \sum_{j = i + 1}^{M} N_{j} σ_{j} . \end{matrix}

(13)

Therefore, the energy consumption of the ith CH during a round is equal to

\begin{array}{l} E ({ch}_{i}) & = E_{elec} (λ N_{i} + λ \sum_{j = i + 1}^{M} N_{j} σ_{j}) \\ + (E_{elec} + ε_{mp} d_{next-ch}^{4}) λ \sum_{j = i}^{M} N_{j} σ_{j}, \end{array}

(14)

in which

E_{elec}

is the electronics energy and

ε_{mp} d_{next-ch}^{4}

is the amplifier energy of the next CH.

For convenience, we suppose $σ_{i} = σ, 0 < σ < 1$ and $σ_{i}$ is a constant; then the amount of energy $E ({ch}_{i})$ is

\begin{array}{l} E ({ch}_{i}) & = λ N_{i} [E_{elec} - σ (E_{elec} + ε_{mp} d_{next-ch}^{4})] \\ + λ σ (2 E_{elec} + ε_{mp} d_{next-ch}^{4}) \sum_{j = i}^{N} N_{j} . \end{array}

(15)

In order to balance the energy consumption between all CHs, we propose an intracluster member allocation scheme, which restricts the number of member nodes of some CHs with higher burden. We divide sensor nodes into the normal groups and the additional groups. In this way each CH will have a maximum set of member nodes it can support without much burden [28]. In order to ensure that the energy consumption of all CHs in different partitions in a round is equal, there exists equivalence relation $E ({ch}_{i}) = E ({ch}_{i + 1})$ .

Hence, we can obtain the following equation:

\begin{matrix} \frac{N_{i + 1}}{N_{i}} = \frac{(1 + σ) E_{elec}}{E_{elec} - σ (E_{elec} + ε_{mp} d_{next-ch}^{4})} . \end{matrix}

(16)

According to the above assumption and analysis, $N_{i}$ is the sum of normal nodes and the additional nodes, which can be expressed as $N_{i} = norm (N_{i}) + add (N_{i})$ . This relationship is given by the following equation: $norm (N_{i + 1}) = N_{i}$ :

\begin{array}{l} \frac{norm (N_{i + 1}) + add (N_{i + 1})}{N_{i}} \\ = 1 + \frac{σ (2 E_{elec} + ε_{mp} d_{next-ch}^{4})}{E_{elec} - σ (E_{elec} + ε_{mp} d_{next-ch}^{4})} . \end{array}

(17)

Then, we can obtain the relationship between the normal nodes of each cluster in partition i and the additional nodes in partition $i + 1$ :

\begin{matrix} add (N_{i + 1}) = \frac{σ (2 E_{elec} + ε_{mp} d_{next-ch}^{4})}{E_{elec} - σ (E_{elec} + ε_{mp} d_{next-ch}^{4})} N_{i} . \end{matrix}

(18)

This can guarantee that the energy consumption between all CHs balances per round.

4.3. Intercluster Communication

In the process of dynamic sensor selection optimization, each CH should build a multihop path to communicate with the sink. During the process of selecting the next-hop nodes for the CHs, we can find three rules that satisfy the conditions: the low energy consumption, the link distance as short as possible, and the minimal hops to the sink. We assume that the communication distance between CHs is less than $d_{0}$ , and the distance between most of CHs and the sink is larger than $d_{0}$ . According to energy consumption formula, the energy required by any CH forwarding k-bit message to the sink is

\begin{array}{l} E_{Total} & = E_{Tx} ({ch}_{i}, {ch}_{j}) + E_{Rx} ({ch}_{i}, {ch}_{j}) + E_{Tx} ({ch}_{j}, b s) \\ = k E_{elec} + k ε_{fs} d^{2} ({ch}_{i}, {ch}_{j}) + k E_{elec} + k E_{elec} \\ + k ε_{mp} d^{4} ({ch}_{j}, \sin k) \\ = 3 k E_{elec} + k ε_{fs} d^{2} ({ch}_{i}, {ch}_{j}) + k ε_{mp} d^{4} ({ch}_{j}, b s) . \end{array}

(19)

Here, we focus on the objective of selecting the next hop for data forwarding, which increases the lifetime of WSNs. In many systems, the energy of node and link quantities are usually the main factors. However, they should not be the primary factor. We take into consideration not only the energy of node, but also the energy consumption of communication and the maximum and minimum hop count between the neighboring nodes and the sink. The weight function is defined as

\begin{array}{l} w_{next-hop} ({ch}_{i}) \\ = \sqrt{\frac{h_{\max}}{h_{\min}}} * [α \frac{E_{res} (j)}{E_{res} (i)} + (1 - α) \\ \times \frac{d^{4} ({ch}_{i}, \sin k)}{d^{2} ({ch}_{i}, {ch}_{j}) + d^{4} ({ch}_{i}, \sin k)}], \end{array}

(20)

where

h_{\max}

and

h_{\min}

denote the maximum and minimum hop count, respectively.

CH i lays neighbor node j to the neighboring CH set and calculates the weight value of the path from j to sink. Then, all weight values are sorted in ascending order and the maximum one could be chosen and its representative CH will be the next-hop node for data forwarding.

Due to the dynamic sensor selection optimization, some CHs may act as intermediate nodes many times, which could tend to die much faster and lead to the “hot spot” problem. To avoid this phenomenon, the threshold $ma x_{relay}$ is set as $α (E_{res} (i) / E ({ch}_{i}))$ , which denotes the maximum number of data forwarding. Its value can be changeable adaptively according to the residual energy of node, and the node will not act as intermediate if the limit is reached.

4.4. Update Scheme for Routing

Definition 1.

The number of hops ranges from 0 to 255, and the max number of hops is used for outgoing probe packets. During initialization, the number of hops from the sink node is set to 0. The default number of hops for normal nodes is set as 255, which indicates unreachable state.

Definition 2.

The simplest way to route queries is to flood queries from the sink over the entire WSNs and set up the reverse paths for desired data to be sent back to the sink within intercluster communication. Each CH sets gradients between neighboring ones and reinforces the best route for real data while transferring the exploratory events on the reverse query path. Initially, the sink acts as a data server to send query packets in the way of flooding. The signal contains its identity (ID), the smallest hop count (HC), and the weight value for the next hop.

Definition 3.

To update the hop count, the intermediate CH determines by verifying the packets from the ones much nearer to the sink. The progress depends on its neighboring hop-count values. If the value of HC in the header of the packets plus 1 is greater than the previous one preserved in its own buffer, the node will update the routing tables and broadcast to its one-hop neighbors. Otherwise, the packet will be ignored.

In the establishment phase, the hop value of sink is set to 0. The initial value of the normal nodes is set as 255, which means the host is unreachable and the list of its parent node is empty. To route a data packet toward the destination, a node broadcasts a route request (RREQ) broadcast to its neighborhood. Each node receiving the RREQ keeps track of it until delaying expiration. When the neighbor node receives RERQ, it will check whether the list of parent node is empty. If a route toward the destination exists, it sends a route reply (RREP) packet to the node that has sent the RREQ. The RREP contains the source ID and sequence number from the RREQ.

When a RREQ is received by a CH, it will make the value of HC in the header of the packet plus 1 as a new hop count and compare with the HC value in its route table. If the new value is less than its HC, it proves that the node has sent the RREQ as its parent node. Otherwise, the node discards the RREP packets without any treatment. Additionally, the new HC value is stored instead of the original value.

The above steps will need to be repeated until all CHs in the network record the minimal HC from the sink for completion of routing.

5. Simulation Results and Analysis

The performance of our proposed model via simulation is evaluated in this section. NS2 (Network Simulator version 2) is used as a simulation test tool. NS2 is a publicly available tool for network simulations, built by various researches including LBL, Xerox PARC, UCB, USC/ISI, and many other contributors as a variant of the “Real Network Simulator.” NS2 is a discrete-event driven and object-oriented network simulator. NS2 is intended to simulate and research computer networks, using a discrete-event simulation technique. For simplicity, we assume the probability of signal collision and interference in the wireless channel is ignorable. Since our CSSPEG is based on multihop mode and unequal clustering for sensors, the transmission radius of the CHs in different ring-shaped areas will exert great impact on network. To examine the overall performance and scalability, we will firstly analyze the trends in terms of communication radius, the number of CHs, and the energy consumption of all areas under the condition of the network size being varied. Secondly, we will review the characteristics of CSSPEG by comparing with cluster algorithms such as UCS [15] and EUCA [23]. The specific experimental parameters are shown in Table 1.

Table 1

Experimental parameters.

Parameter	Value
r	50 m
Node number	200–400
$d_{0}$	87
Sink position	(200, 200)
$E_{elec}$ (nJ/b)	0
$ε_{fs}$ (pJ/b/m²)	10
$ε_{mp}$ (pJ/b/m⁴)	0.0013
$E_{DA}$ (nJ/b/m²)	5
Data packet size (bits)	1000
Initial energy (J)	2
M	5

In our scenario, we set $M = 5$ and then partition the region into 5 ring-shaped areas, and the network size (terrain radius) is varied from 200 meters to 400 meters. Figure 4 shows the number of CHs varying in different network sizes; meanwhile the value η is set from 1.5 to 2.5. In 8, we defined η as the approximate ratio of $E_{rel}$ and $E_{non-ch}$ . Obviously, when its value is the upper limit, the CH node communication radius is much larger than the normal nodes so as to enlarge the coverage of a single CH. This leads to a few numbers of clusters in the network relatively. Conversely, the number of CHs will increase properly.

Figure 4

The number of CHs varying in different network size.

The transmission radius of the CHs at different area is shown in Figure 5. Each curve corresponds to a particular network size, where the network size is varied from 200 meters to 400 meters. As the network size increases, the transmission range of the CHs increases synchronously. However, we can notice that the transmission range of the inner area is comparatively shorter than those in the outer area; the reason is that smaller size of the clusters nearer to the sink than the remote ones can balance the energy depletion of sensors in the whole network. The CHs near the sink are usually responsible for forwarding the data to the sink from remoter cluster. Thus, they consume more energy than remote heads. The simulation results demonstrate that our strategy can preserve some more energy for the intercluster relay traffic.

Figure 5

The transmission radii of the CHs in the ith ring-shaped area.

As mentioned in the previous section, the energy hole problem may rise in multihop wireless sensor networks if the CHs which act as intermediate nodes in the routing path are burdened with heavy relay traffic. We analyze the average residual energy of all nodes in different partitions as the network size is 400 m² specially. As shown in Figure 6, the average residual energy of the nodes in different partitions drops almost simultaneously and demonstrates stability and volatility. Therefore, the proposed strategy can alleviate relay traffic burden on the CH nodes effectively by allocating the member nodes in the clusters and adjusting the transmission radius according to the hop counts.

Figure 6

The average residual energy of nodes in $P_{i}$ area.

Next, we verify the lifetime of network and energy efficiency for our strategy. Figure 7 shows the number of active nodes and energy consumption during their lifetime. From Figure 7, it can be seen that CSSPEG is better than the other algorithms, and the lifetime extends about 27.5% and 19.8%. In addition, all the nodes almost died till the round is 500 in UCS and 520 in EUCA; however CSSPEG algorithm requires 560.

Figure 7

The number of active nodes when $N = 200$ .

Figure 8 shows the total energy consumption over the number of rounds. CSSPEG significantly reduces energy consumption compared to UCS and EUCA because CSSPEG uses an alternative method to adjust the CH's transmission radius based on its location. The method of division into quadrants and the use of multihop for transmission data between the clusters also lead to more efficient energy usage and result in less consumption of energy for both intracluster and intercluster data transmission. The graphs curve shows that the dissipation varied between rounds of CSSPEG is higher than UCS and EUCA. It is evident that CSSPEG has better performance than the other two algorithms in terms of energy efficiency.

Figure 8

The total energy consumption when $N = 200$ .

Finally, we carry out the experiment in the environment with more densely deployed nodes in order to investigate the scalability of CSSPEG. The terrain radius is set to 300 and $N = 400$ . As can be seen in Figure 9, the number of surviving nodes in CSSPEG is much more than UCS and EUCA. Meanwhile we can observe that the EUCA does not perform well when the network size increases. Compared with the results of the number of active nodes in UCS shown in Figure 7, the advantage of EUCA is not obvious.

Figure 9

The number of active nodes when $N = 400$ .

Figure 10 demonstrates that the total energy consumption of CSSPEG is better than the other algorithms when $N = 400$ and the terrain radius set is 300. Through the above experimental results, it can be observed that CSSPEG can be satisfied with the requirement of different sizes of networks, and it has better performance in terms of energy efficiency.

Figure 10

The total energy consumption when $N = 400$ .

Through the above experimental results, it can be observed that the proposed CSSPEG can efficiently balance the energy consumption of CHs, decrease the dead speed of the nodes, and prolong the lifetime of the network. Thus, the CSSPEG shows effectiveness to improve energy efficiency of WSNs and can be successfully applied to the field of wireless sensor networks.

6. Conclusions and Future Work

Since a single sensor may be in a fading dip, cooperative sensing among multiple sensors which experience uncorrelated fading is required to guarantee reliable sensing performance. This paper proposes a cooperative sensor selection and power-efficient gathering strategy in multihop wireless sensor networks to solve the energy consumption problem. The performance evaluation has been done via different simulations, which shows the proposed CSSPEG strategy can efficiently balance the energy consumption of CHs, decrease the dead speed of the nodes, and prolong the lifetime of the network. In the future, we will try to improve the CSSPEG and try to meet the other requirements in WSNs, such as effective coverage rate of the monitored area.

Footnotes

Conflict of Interests

The authors declare that there is no conflict of interests regarding the publication of this paper.

Acknowledgment

The authors would like to thank the anonymous reviewers for their insightful comments and constructive suggestions that have improved the paper.

References

Akyildiz

I. F.

Sankarasubramaniam

Cayirci

A survey on sensor networks

IEEE Communications Magazine 2002 40 8 102 114

10.1109/MCOM.2002.1024422

2-s2.0-0036688074

Yick

Mukherjee

Ghosal

Wireless sensor network survey

Computer Networks 2008 52 12 2292 2330

10.1016/j.comnet.2008.04.002

2-s2.0-46449122114

Lindsey

Raghavendra

Sivalingam

K. M.

Data gathering algorithms in sensor networks using energy metrics

IEEE Transactions on Parallel and Distributed Systems 2002 13 9 924 935

10.1109/TPDS.2002.1036066

2-s2.0-0036766616

Perillo

M. A.

Cheng

Heinzelman

W. B.

An analysis of strategies for mitigating the sensor network hot spot problem

Proceedings of the 2nd Annual International Conference on Mobile and Ubiquitous Systems: Networking and Services (MobiQuitous ′05)

July 2005

IEEE Computer Society

474 478

10.1109/MOBIQUITOUS.2005.10

2-s2.0-33749526900

Zhu

Shu

Mao

Deng

Enhanced MAC protocol to support multimedia traffic in cognitive wireless mesh networks

Multimedia Tools and Applications 2013 67 1 269 288

10.1007/s11042-011-0942-7

2-s2.0-84881174187

Leung

K. K.

Malvankar

A dynamic clustering and energy efficient routing technique for sensor networks

IEEE Transactions on Wireless Communications 2007 6 8 3069 3079

10.1109/TWC.2007.06003

2-s2.0-34548062119

Younis

Fahmy

HEED: a hybrid, energy-efficient, distributed clustering approach for ad hoc sensor networks

IEEE Transactions on Mobile Computing 2004 3 4 366 379

10.1109/TMC.2004.41

2-s2.0-10944266504

Kumar

Aseri

T. C.

Patel

R. B.

EEHC: energy efficient heterogeneous clustered scheme for wireless sensor networks

Computer Communications 2009 32 4 662 667

10.1016/j.comcom.2008.11.025

2-s2.0-59849092261

Chamam

Pierre

A distributed energy-efficient clustering protocol for wireless sensor networks

Computers and Electrical Engineering 2010 36 2 303 312

10.1016/j.compeleceng.2009.03.008

2-s2.0-75949120144

10.

Bajaber

Awan

Energy efficient clustering protocol to enhance lifetime of wireless sensor network

Journal of Ambient Intelligence and Humanized Computing 2010 1 4 239 248

10.1007/s12652-010-0019-x

2-s2.0-79952443096

11.

Minhas Akhtar

A. A.

Jabbar

Energy aware intra cluster routing for wireless sensor networks

International Journal of Hybrid Information Technology 2010 3 1 29 48

12.

Maybank

S. J.

Zhu

Efficient clustering aggregation based on data fragments

IEEE Transactions on Systems, Man, and Cybernetics Part B: Cybernetics 2012 42 3 913 926

2-s2.0-84862783105

10.1109/TSMCB.2012.2183591

13.

Bandyopadhyay

Coyle

E. J.

An energy efficient hierarchical clustering algorithm for wireless sensor networks

Proceedings of the IEEE Societies 22nd Annual Joint Conference of the IEEE Computer and Communications (INFOCOM '03)

March-April 2003

1713 1723

10.1109/INFCOM.2003.1209194

14.

Xia

Vlajic

Near-optimal node clustering in wireless sensor networks for environment monitoring

Proceedings of the 21st International Conference on Advanced Information Networking and Applications (AINA ′07)

May 2007

Niagara Falls, NY, USA

632 641

2-s2.0-34548782126

10.1109/AINA.2007.97

15.

Soro

Heinzelman

W. B.

Prolonging the life-time of wireless sensor networks via unequal clustering

Proceedings of the 19th IEEE International Parallel and Distributed Processing Symposium (IPDPS ′05)

April 2005

236 243

10.1109/IPDPS.2005.365

2-s2.0-33746310074

16.

Chen

An energy-efficient unequal clustering mechanism for wireless sensor networks

Proceedings of the IEEE International Conference on Mobile Ad-hoc and Sensor Systems Conference (MASS ′05)

November 2005

Washington, D.C., USA

597 604

2-s2.0-33750328982

10.1109/MAHSS.2005.1542849

17.

Gong

Wang

Zhou

Multihop routing protocol with unequal clustering for wireless sensor networks

Proceedings of the ISECS International Colloquium on Computing, Communication, Control, and Management (CCCM ′08)

August 2008

552 556

10.1109/CCCM.2008.99

2-s2.0-55749088640

18.

Zhao

Wang

An unequal layered clustering approach for large scale wireless sensor networks

Proceedings of the 2nd International Conference on Future Computer and Communication (ICFCC ′10)

May 2010

Wuhan, China

V1750 V1756

10.1109/ICFCC.2010.5497328

2-s2.0-78049347553

19.

Bagci

Yazici

An energy aware fuzzy unequal clustering algorithm for wireless sensor net-works

Proceedings of the IEEE International Conference on Fuzzy Systems (FUZZ ′06)

2006

97 108

20.

Liu

Huang

T.-L.

Zhou

X.-Y.

G.-X.

An improved energy efficient unequal clustering algorithm of wireless sensor network

Proceedings of the IEEE International Conference on Intelligent Computing and Integrated Systems (ICISS ′10)

October 2010

930 933

10.1109/ICISS.2010.5657032

2-s2.0-78651466813

21.

Jaichandran

Irudhayaraj

A. A.

Raja

J. E.

Effective strategies and optimal solutions for Hot Spot Problem in wireless sensor networks (WSN)

Proceedings of the 10th International Conference on Information Sciences, Signal Processing and their Applications (ISSPA ′10)

May 2010

389 392

2-s2.0-78650274491

10.1109/ISSPA.2010.5605518

22.

Wang

An energy-driven unequal clustering protocol for heterogeneous wireless sensor networks

Journal of Control Theory and Applications 2011 9 1 133 139

10.1007/s11768-011-0232-y

MR2835100

2-s2.0-79952483688

23.

Wan

Lei

Gou

An energy-efficient unequal clustering algorithm in wireless sensor networks

Advanced Materials Research 2013 629 801 807

2-s2.0-84872892904

10.4028/www.scientific.net/AMR.629.801

24.

Zhu

Wang

Power-efficient spatial reusable channel assignment scheme in WLAN mesh networks

Mobile Networks and Applications 2012 17 1 53 63

10.1007/s11036-011-0302-x

2-s2.0-84861808267

25.

Flushing

E. F.

di Caro

G. A.

An analytical model of IEEE 802.15.4 non-beacon enabled CSMA/CA in multi-hop wireless sensor networks

2011 05-11

Lugano, Switzerland

IDSIA

26.

Ever

Luchmun

Mostarda

Navarra

Shah

UHEED: an unequal clustering algorithm for wireless sensor networks

Proceedings of the 1st International Conference on Sensor Networks (SENSORNETS ′12)

February 2012

185 193

2-s2.0-84862174776

27.

Bhardwaj

Garnett

Chandrakasan

A. P.

Upper bounds on the lifetime of sensor networks

Proceedings of the International Conference on Communications (ICC ′01)

June 2001

785 790

2-s2.0-0034865944

28.

Shu

Wang

An optimized multi-hop routing algorithm based on clonal selection strategy for energy-efficient management in wireless sensor networks

Sensors and Transducers 2013 22 8 14

2-s2.0-84888990565