An Energy-Efficient Data Collection Method for Wireless Multimedia Sensor Networks

Abstract

In wireless sensor networks (WSNs), some sensor nodes transmit their acquired data to a far distant base station directly or by linking several relay nodes. WSNs might not be able to operate if a sensor node completely consumes its battery power and does not perform its designated operation. Semiconductor technology drives sensor networks to collect video and audio data as well as environmental data. Therefore, energy is one of the most important resources in wireless multimedia sensor networks (WMSNs), and clustering is proposed to reduce energy consumption. When a mobile sink is used, a lot of energy is saved, and the lifetime of the network can be extended because the sensor nodes do not need to transmit multimedia data to a far distant base station. We propose an energy-efficient data collection method to extend the lifetime of networks that use a mobile sink. By using the neighborhood density clustering method and defining an optimal path for the mobile sink, the proposed method extends the life of the WMSN. Simulation results show that the proposed method increases the lifetime of WMSNs by up to 15 rounds, compared with LEACH. And data collection time is decreased owing to the predefined path for the mobile sink.

1. Introduction

Wireless sensor networks (WSNs) are deployed to collect data from the physical environment, process them, and transmit them to a base station (BS). A lot of sensor nodes are implemented in WSNs, and the number of sensor nodes varies with the application field. Typically, a sensor node consisted of a microprocessor with limited internal memory, a wireless communication device, sensors, and a limited power source. The power source may be batteries or solar-cell rechargeable batteries that operate for several months to several years. As semiconductor technologies advance, WSNs are applied to monitor and collect not only environmental data but also video and audio data. The introduction of inexpensive CMOS cameras has made it possible to monitor video data and to accelerate the emergence of wireless multimedia sensor networks (WMSNs) [1, 2]. These WMANs are able to monitor or sample video and audio data from the environment, animals, weather, and security systems [3, 4]. But the size of multimedia data requires more efficient energy consumption since the data size is larger than that of typical environmental data.

Recently, mobile sinks have been proposed as a solution to collect data from sensor nodes in WSNs. They are applied to reduce energy consumption, to balance energy levels among the sensor nodes throughout the WSN, and to deal with isolated regions [5, 6]. Since a mobile sink is a device external to the WSN, power consumption for its movement and operation does not affect the lifetime of the WSN. Periodically, the mobile sink returns to the base station to upload collected data and to recharge battery power for its next round [7]. When the mobile sink moves along the area of the WSN, the cost of path reconfiguration becomes much higher than that of a static network. Reconfiguration could result in sudden changes in the path. In addition, the mobile sink could be controllable or uncontrollable. An uncontrollable sensor node is attached to some moving object, such as a soldier, a public transportation vehicle, or an animal [8, 9]. A controllable mobile sink's position can be changed by adjusting its trajectory and speed [10–13]. Controlling and building an efficient trajectory for the mobile sink are also required for it to be preprogramed efficiently for the WMSN. Inapplicable speed or an ineffective trajectory for the mobile sink causes data loss and latency.

This paper presents an efficient method to reduce energy consumption in multimedia sensor networks and to extend the lifetime of the sensor nodes. An energy-efficient data collection method is proposed to extend the lifetime of networks that use a mobile sink. In particular, the proposed method uses the neighborhood density clustering method and defines an optimal travelling path for the mobile sink. In Section 2, related works are reviewed. In Section 3, the proposed model is introduced. In Section 4, several performance experiments for the proposed model are discussed. Finally, conclusions from this study are provided in Section 5.

2. Related Works

Mobility in a mobile sink has been studied in some of the literature to improve performance of WSNs [8, 14–20]. One study has deployed a system that uses mobile elements to carry data from sensor nodes to a base station, which is called mobile ubiquitous LAN extensions (MULEs) [8]. In MULEs, a mobile observer picks up data directly from sensor nodes, saves them in a buffer, and drops off data to wired access points. The movement of MULEs has been modeled as a 2D random walk. However, the main drawback of MULEs is increased data collecting latency, because the mobile observer has to travel across the transmission range of each sensor node in the network to collect data.

Zhao and Yang have presented polling-based data gathering using a mobile node and named it bounded relay hop mobile data gathering (BRH-MDG) [14]. They choose a subset of sensor nodes as polling points, and each node aggregates local data from its affiliated nodes within a certain number of relay hops. Those polling points temporarily save data in their caches and upload them to the mobile collector when it arrives. To find a set of polling points, they formulate two centralized and distributed algorithms. And they prove its NP-hardness (nondeterministic polynomial-time hardness). A disadvantage of this method is that the buffer of a sensor node might overflow before a mobile collector visits it because of many data polling points.

Wang et al. have studied a mobile sink at the edge of the sensing field. The movement of the mobile sink is either clockwise or counterclockwise [15]. Afterward, the entire network is divided into equal clusters, and, in each part, a cluster head is selected by residual energy. They formulated a scheme to send data from a normal node to the cluster head using multihop transmission. Member nodes decide to send data to the cluster head or a mobile sink by considering distance. While cluster heads aggregate and send data to the mobile sink, computing and data transmission overload could occur in these nodes. Data gathering routes are around the perimeter of the sensing area, and this indeed affects long data latency.

Gao et al. [17] have proposed a data collection method that improves the capacity of data buffers and reduces energy consumption in large-scale wireless sensor networks with mobile sinks. Mobile sinks move along fixed paths at a constant speed. In their method, the members within the multihop communications area (MCA) are assigned to the corresponding subsinks within the direct communications area (DCA) according to the length of the communication time between the mobile sinks and the subsinks, and this method improves network throughput. However, this method employs path-constrained sink mobility instead of path-controllable sink mobility.

Rao and Biswas [18] presented a non-real-time data collection method using mobile sinks. Their idea is that a mobile data harvester (MDH) moves along a desired trajectory while collecting data from a set of sensors that are commissioned as a designated gateway (DG) along the trajectory. A DG buffers the data generated in its multihop neighborhood till the MDH contacts it. Effectively, data from all nodes on the tree rooted at a DG are buffered at the DG before they are uploaded to the MDH. However, this method has the burden of data collection at a gateway that acts as a cluster head in LEACH [21].

Konstantopoulos et al. [19] have proposed a method called MobiCluster protocol, which minimizes the overall network overhead and energy expenditure associated with the data retrieval process from sensor nodes. It uses cluster structures consisting of member nodes that route their measured data to their assigned cluster head. The cluster heads perform data filtering on the raw data, exploiting potential spatial-temporal data redundancy and then forwarding the filtered information to their assigned rendezvous nodes (RNs) which are located in proximity to the trajectory of the mobile sink. This method has a limitation in that the mobile sink explores a fixed route.

Data collection methods using a mobile sink can be divided into the constrained (uncontrollable) path method and the changeable (controllable) path method, depending on whether the path of the mobile sink is fixed. Shah et al. [8], Zhao and Yang [14], and Rao and Biswas [18] have applied the constrained path method, whereas Wang et al. [15], Gao et al. [17], and Konstantopoulos et al. [19] have applied the changeable path method. In addition, there are a number of studies for data collection using mobile sinks. For example, Puthal et al. [20] have proposed a data collection method using a single hop, and Wang et al. [22] proposed a data collection method using multiple sinks. Many researchers have focused on collection methods for general sensor data, but collecting multimedia data has not been studied by many researchers.

3. Proposed Model

Since multimedia data is much larger than typical sensor data, a new data collection method is required. The proposed collection process is composed of three stages, which are (1) clustering, (2) deciding the mobile sink's path, and (3) gathering data, as shown in Figure 1. At the clustering stage, clustering is built by checking the density of sensor nodes, which is the neighborhood density clustering method. The optimal visitation path for the mobile sink is calculated using the minimum spanning tree (MST). Then, data is collected by the mobile sink using the time division multiple access (TDMA) protocol to avoid collisions and reduce energy consumption in the sensor nodes.

Figure 1

The process of the proposed data collection method.

3.1. Mobile Sink and Sensor Nodes

A mobile sink moves along a predefined trajectory and gathers sensed data from sensor nodes in a single hop. When it finishes gathering data at predefined points, it uploads the collected data to a base station. The velocity of the mobile sink is constant, and it is aware of all sensor node locations. The mobile sink is equipped with an unlimited energy source, a powerful processor, large buffers, a long-range transceiver, and a global positioning system.

All sensor nodes have the same limited energy source. When sensor nodes run out of energy, they are defined as dead nodes. The base station and sensor nodes are in fixed positions. Every sensor node is given its own identification number, with location information and sensor node IDs obtained at deployment.

3.2. Energy Model

Two different radio models are used. One is a free space model and the other is a multipath fading channel model. When the distance between sending and receiving nodes is less than the threshold value $d_{crossover}$ , in the free space model, the power loss, $d^{2}$ , is used. Otherwise, in the multipath fading channel model, the power loss $d^{4}$ is used. From the two models in (1), the amount of energy to send k-bits of data is calculated to distance d [2]:

\begin{matrix} E_{T x} = {\begin{cases} E_{elec} * k + ε_{f s} * k * d^{2}, & d < d_{crossover}, \\ E_{elec} * k + ε_{amp} * k * d^{4}, & d \geq d_{crossover}, \end{cases} \end{matrix}

(1)

\begin{matrix} E_{R x} (k) = E_{elec} * k, \end{matrix}

(2)

\begin{matrix} d_{crossover} = \sqrt{\frac{ε_{f s}}{ε_{amp}}} . \end{matrix}

(3)

From (1), $E_{T x}$ ( $k, d$ ) is the energy usage of the transmitter that sends k-bits of data to the receiver at distance d. In (2), $E_{R x}$ (k) is the energy consumption of the receiver node from receiving k-bits of data. $E_{elec}$ is the energy consumption of the wireless transmission and reception circuit. $ε_{f s}$ and $ε_{amp}$ define the energy consumption factor for amplification in the two radio models. In the experimental energy model, d means the distance between a mobile sink and a node in a cluster. By comparing it with the value of $d_{crossover}$ , one of the energy models of (1) is selected. The value of $d_{crossover}$ is calculated with (3).

3.3. Neighborhood Density Clustering

After sensor nodes are deployed in the sensing area, they are divided into clusters. Clustering sensor nodes reduces the number of internode communications by localizing data transmission within clusters and decreases the number of transmissions to the base station. One-hop neighbors of a node and their location data are used to construct clustering. After deploying sensor nodes in the field, a mobile sink obtains the location information of each node and classifies it by clustering. First, for every node, the number of one-hop neighbors is calculated and given as a degree for each node. Nodes with the highest degree are considered initial points for configuring clusters. Neighbor nodes with a low degree are merged with nodes with a higher degree. In Figure 2, sensor nodes are shown by degree and their neighbor nodes. In Figure 3, Node E and Node H are the highest degree nodes, and other neighbor nodes are merged with them. The result is two clusters.

Figure 2

Sensor nodes and their degrees.

Figure 3

Nodes are merged with higher degree nodes.

The next step is calculating the centroid location for each cluster. The centroid location of each cluster is calculated based on $(x, y)$ location information of cluster members. The centroid location of each cluster is the data collection point. Higher degree nodes are also taken to be member nodes. After calculating the distance of each node, the centroid location is calculated and grouped into the cluster.

These calculations continue until any node moves to another cluster. In Figure 4, the final result is given, and all nodes are grouped into two clusters.

Figure 4

The final result is reached.

3.4. Mobile Sink Path

To find the optimal data gathering path for a mobile sink, MST is applied to a weighted graph. Odd vertices are defined and found via minimum weights matching them. After odd vertices are matched, Euler tour is found in the resulting graph. A list of vertices is created that represents Euler tour, which crosses every edge exactly once without repetition. However, the Euler cycle path might pass through a vertex more than once, which is inefficient in our case. Therefore, it does not provide an optimal solution for our data gathering tour. Pseudocode 1 shows the algorithm for neighborhood density clustering.

Pseudocode 1: Pseudocode for neighborhood density clustering.

(1) Inputs:

(2) $S = {S_{1}, S_{2}, \dots, S_{n}}$ —Sensor nodes

(3) k—number of clusters

(4) Outputs

(5) $C = {C_{1}, C_{2}, \dots, C_{k}}$ —cluster centers

(6) $m : S \to C$ —cluster membership

(7) foreach $(S_{n} \in S)$

(8) Calculate node_degree $(D (S_{n}))$

(9) end

(10) Take k $D (S)$ .max nodes

(11) foreach Neighbors of $D (S)$ .max nodes

(12) Merge to $D (S)$ .max nodes

(13) end

(14) foreach notNeighbors of $D (S)$ .max nodes

(15) Merge to closest $D (S)$ .max

(16) end

(17) while m has changed

(18) for each $j \in {1, \dots, k}$

(19) recalculate the center of cluster

(20) end

(21) for each $S_{i}$

(22) $m (S_{n})$ = argmin distance $(S_{i}, C_{k})$

(23) end

(24) end

(25) return C

In Figure 5, data gathering points are defined and lines that connect them are shown. In Figure 6, solid edges represent the MST. In Figure 7, the path for the mobile sink is shown after all calculations.

Figure 5

Data gathering points and paths.

Figure 6

Finding MST and connecting odd-degree nodes.

Figure 7

Path for the mobile sink.

3.5. Data Gathering

A mobile sink starts its data gathering tour and moves along the defined path. When it gets to a data gathering point, it stops and broadcasts a “Hello” message to cluster members. This “Hello” message contains the ID numbers of cluster members. When sensor nodes receive this message, they check its content. Whenever each node finds its own ID number in the broadcast message, it sends a “Reply” message to the mobile sink and waits for a response. This reply message contains the ID of the sensor node. If a sensor node does not find its own ID, it ignores the message. The mobile sink receives several reply messages from the member nodes of the cluster. According to these reply messages, the mobile sink creates a TDMA schedule for the identified nodes in the cluster. In the TDMA schedule, every node has its own time slot, which is divided into three parts: Data, Acknowledge, and Period.

In the Data phase, nodes send their acquired data to the mobile sink. In the Acknowledge phase, the mobile sink verifies the data received from the sensor nodes. The Period phase is used to synchronize slots.

In its reply message, the mobile sink sends the TDMA schedule to the sensor nodes. In the TDMA time slots, the sensor nodes send their data to the mobile sink. A TDMA schedule is created for every cluster, because each cluster can have a different number of members. The TDMA schedule is created based on the number of members in the cluster. After sending the TDMA schedule to sensor nodes, the mobile sink gathers data from the sensor nodes and uploads the gathered data directly to the base station. After uploading data successfully, the mobile sink continues on its data gathering tour. The next data gathering point will be the centroid location of the upcoming cluster on its path.

4. Performance Evaluation

In this section, several experiments for performance of the proposed data collection method are described. First is an experiment for energy consumption according to the number of clusters in two types of protocol. Another experiment was conducted for comparison with other methods. For the simulations, network simulator 2 (NS-2) is used. Sensor nodes are distributed in a region of 250 m × 250 m. The number of sensor nodes was 100, with each sensor node in the network having the same 1 Joule of initial energy. In the simulation environment, 802.11MAC protocol, a wireless channel model, and the ad hoc on-demand distance vector (AODV) routing protocol are used. The base station is fixed outside the sensor area. The main focus is the energy consumption of the sensor nodes and the lifetime of the network. After each round, the energy in the network is calculated, with energy in the network defined as the sum of the sensor nodes’ energy percentages. Table 1 shows the simulation parameters and environment.

Table 1

Simulation parameters and environment.

Parameter	Value
Simulation area	250 m × 250 m
Number of sensors	100
Speed of the mobile sink	5 m/s
$E_{init}$	1 Joule
$E_{elec}$	50 nj/bit
$ε_{fs}$	10 pj/bit/m²
$ε_{amp}$	0.0013 pj/bit/m⁴
MAC protocol	802.11
Channel model	Wireless channel
Routing protocol	AODV

4.1. Energy Consumption

First, the proposed algorithm shown in Pseudocode 1 is analyzed with different numbers of clusters under two protocols: transmission control protocol (TCP) and user datagram protocol (UDP). Figure 8 shows energy consumption of the network through rounds with five clusters. When UDP is used, the energy consumption of the nodes is less than under TCP because, under TCP, the Acknowledge (ACK) message is used to inform the sender that the message was received. Although the size of the ACK is small, senders and receivers consume energy to transmit and receive this message. With sensor nodes divided into five clusters, nodes use more energy using TCP compared with UDP.

Figure 8

Energy consumption of nodes in five clusters.

Figure 9 illustrates the energy consumption of sensor nodes when they are divided into 8 clusters. Sensor nodes consume less energy when UDP is applied. Another noticeable point is that the lifetime of the network is increased in both cases. When the network is divided into eight clusters, the cluster size becomes smaller, and the number of member nodes decreases. As a result, the distance between the cluster center and the nodes is shorter. Therefore, sensor nodes can transmit multimedia data within a shorter distance. For TCP and UDP, network lifetime increases by 2 more rounds.

Figure 9

Energy consumption of nodes in eight clusters.

Figure 10 shows the energy consumption of sensor nodes when the network is divided into 11 clusters. In Figures 8 and 9, TCP uses more energy than UDP. However, Figure 10 shows that UDP and TCP have similar energy consumption. The reason is that when the number of clusters increases, the area of a cluster becomes narrower. herefore, data packets of nodes can be transmitted in a shorter distance, which decreases energy consumption. In this experiment, the network lifetime is extended by 38 rounds under UDP.

Figure 10

Energy consumption of nodes in 11 clusters.

According to the experiment results, UDP consumes less energy than TCP in the proposed method because UDP is not required to send an Acknowledge message. When the number of clusters increases, energy consumption is reduced such that energy consumption between UDP and TCP is similar.

4.2. Comparison with Other Methods

The proposed method was compared with LEACH [21] and the method of Wang et al. [15]. Figure 11 illustrates the comparison of the three methods, which are the proposed method, LEACH, and mobile sink at the edge of the area. In the proposed method, sensor nodes consume less energy compared with the other two methods. Compared with the LEACH clustering algorithm, the proposed method increases the lifetime of the network by 15 more rounds. This is more than twice the lifetime for multimedia data. Compared to Wang et al., the proposed method shows better performance and more than double the network lifetime.

Figure 11

Comparison of energy consumption.

The proposed method shows better performance than other methods in energy efficiency. The reason is analyzed below.

First, in the proposed method, sensor nodes do not dissipate energy to elect cluster heads (CHs). There is no energy wasted on cluster head nodes. But, in LEACH [21] and Wang et al. [15], a considerable amount of energy is consumed by selecting cluster heads and by relaying data from one node to the other for long-distance sensor nodes. Each cluster head consumes its energy to receive data from its member nodes, which is proportional to the number of its member nodes multiplied by the communications energy expressed by (1). For the proposed method, the energy of each cluster is saved in every data gathering round.

Second, one of the distinct points of the proposed method is communications configuration within a cluster, as seen in Figure 12. In other methods, a cluster head and member nodes could be placed on opposite sides of a cluster, and the distance between them might be very long, compared to other nodes, as shown in Figure 12(a). Therefore, the farther the distance between a cluster head and a member node is, the more they consume energy. Since a mobile sink in the proposed method stops at the centroid location of each cluster, as shown in Figure 12(b), the transmission distance is decreased.

Figure 12

Comparison of communications within a cluster.

5. Conclusion

This paper proposes a multimedia data collection method in clustered wireless multimedia sensor networks. Performance of the proposed method was evaluated and compared with other methods under the same conditions. The proposed method was evaluated with different numbers of nodes and different protocols. The simulation results show that UDP consumes less energy than TCP in the proposed method. Also, when the number of clusters is increased, energy consumption is reduced. In comparison with LEACH, the proposed method increases network lifetime by up to 15 rounds for WMSNs. And data collection time is decreased with a predefined path for the mobile sink. For future study, we are considering several experiments that take into account mobility in the sensor nodes.

Footnotes

Conflict of Interests

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

Acknowledgment

This work was supported by the 2012 Yeungnam University research grant.

References

Alyildiz

Weilian

Yogesh

Cayirci

A survey on wireless sensor networks

Computer Networks 2002 38 4 393 422

Milica

Zhilbert

Goran

Veljko

A survey of military applications of wireless sensor networks

Proceedings of the Mediterranean Conference on Embedded Computing (MECO '12)

2012

Rocha

A. R.

Pirmez

Delicato

F. C.

Lemos

É.

Santos

Gomes

D. G.

de Souza

J. N.

WSNs clustering based on semantic neighborhood relationships

Computer Networks 2012 56 5 1627 1645

10.1016/j.comnet.2012.01.014

2-s2.0-84858156766

Huiyong

Jingyang

Min

Building a smart home system with WSN and service robot

Proceedings of the 5th Conference on Measuring Technology and Mechatronics Automation (ICMTMA '13)

January 2013

Hong Kong

353 356

10.1109/ICMTMA.2013.90

2-s2.0-84876519337

Shah

R. C.

Roy

Jain

Brunette

Data mules: Modeling a three-tier architecture for sparse sensor networks

Proceedings of the IEEE Workshop on Sensor Network Protocols and Applications (SNPA '03)

2003

Chen

T.-C.

Chen

T.-S.

P.-W.

On data collection using mobile robot in wireless sensor networks

IEEE Transactions on Systems, Man, and Cybernetics A:Systems and Humans 2011 41 6 1213 1224

10.1109/TSMCA.2011.2157132

2-s2.0-80054820990

Zhao

Yang

mobile data gathering with space-division multiple access inwireless sensor network

Proceedings of the 27th IEEE INFOCOM

2008

Shah

R. C.

Roy

Jain

Brunette

Data MULEs: modeling and analysis of a three-tier architecture for sparse sensor networks

Ad Hoc Networks 2003 1 2-3 215 233

10.1016/S1570-8705(03)00003-9

2-s2.0-3242765057

Juang

Oki

Wang

Martonosi

Peh

L. S.

Rubenstein

Energy-rfficient computing for wildlife tracking: design tradeoffs and early experiences with zebranet

Proceedings of the 10th International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS '02)

2002

96 107

10.1145/605397.605408

10.

Somasundara

A. A.

Ramamoorthy

Srivastava

M. B.

Mobile element scheduling with dynamic deadlines

IEEE Transactions on Mobile Computing 2007 6 4 395 410

10.1109/TMC.2007.57

2-s2.0-33847756370

11.

Yang

Data gathering in wireless sensor networks with mobile collectors

Proceeding of the 22nd IEEE International Parallel and Distributed Processing Symposium (IPDPS '08)

April 2008

Miami, Fla, USA

1 9

10.1109/IPDPS.2008.4536269

2-s2.0-51049112060

12.

Kusy

Lee

Wicke

Milosavljevic

Guibas

Predictive QoS routing to mobile sinks in wireless sensor networks

Proceedings of the International Conference on Information Processing in Sensor Networks (IPSN '09)

April 2009

2-s2.0-71049154457

13.

Yang

SenCar: An energy-efficient data gathering mechanism for large-scale multihop sensor networks

IEEE Transactions on Parallel and Distributed Systems 2007 18 10 1476 1488

10.1109/TPDS.2007.1070

2-s2.0-34648846751

14.

Zhao

Yang

Bounded relay hop mobile data gathering in wireless sensor networks

IEEE Transactions on Computers 2012 61 2 265 277

10.1109/TC.2010.219

MR2931787

2-s2.0-84855458870

15.

Wang

Yin

Kim

J.-U.

Lee

Lai

C.-F.

A mobile-sink based energy-efficient clustering algorithm for wireless sensor networks

Proceedings of the 12th IEEE International Conference on Computer and Information Technology (CIT '12)

October 2012

Chengdu, China

678 683

10.1109/CIT.2012.142

2-s2.0-84872321812

16.

Sinha

Lobiyal

Performance evaluation of data aggregation for cluster-based wireless sensor network

Human-Centric Computing and Information Sciences 2013 3 13

17.

Gao

Zhang

Das

S. K.

Efficient data collection in wireless sensor networks with path-constrained mobile sinks

IEEE Transactions on Mobile Computing 2011 10 4 592 608

10.1109/TMC.2010.193

2-s2.0-79951913744

18.

Rao

Biswas

Analyzing multi-hop routing feasibility for sensor data harvesting using mobile sinks

Journal of Parallel and Distributed Computing 2012 72 6 764 777

10.1016/j.jpdc.2012.02.017

2-s2.0-84859915521

19.

Konstantopoulos

Pantziou

Gavalas

Mpitziopoulos

Mamalis

A rendezvous-based approach enabling energy-efficient sensory data collection with mobile Sinks

IEEE Transactions on Parallel and Distributed Systems 2012 23 5 809 817

10.1109/TPDS.2011.237

2-s2.0-84859727925

20.

Puthal

Sahoo

Sharma

Dynamic model for efficient data collection in wireless sensor networks with mobile sink

International Journal of Computer Science and Technology 2012 3 1

21.

Heinzelman

Balakrishnan

Chandrakasan

Low energy adaptive clustering hierarchy

Proceedings of Hawaii International Conference on System Science

January 2000

22.

Wang

Zuo

Shen

Lee

Multiple mobile sink-based routing algorithm for data dissemination in wireless sensor networks

Concurrency Computation: Practice Experience, 2014, http://wileyonlinelibrary.com