Procedia Computer Science 46 ( 2015 ) 1140 – 1147
1877-0509 © 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of organizing committee of the International Conference on Information and Communication Technologies (ICICT 2014) doi: 10.1016/j.procs.2015.01.026
ScienceDirect
International Conference on Information and Communication Technologies (ICICT 2014)
Grid Based Adaptive Sleep for Prolonging Network Lifetime in
Wireless Sensor Network
Kumar Nitesh
a,*, Prasanta K. Jana
ba,bDepartment of Computer Science and Engineering
Indian School of Mines, Dhanbad-826004, India
bIEEE Senior Member
Abstract
Energy conservation of the sensor nodes is a burning issue and a central problem to the development of a large scale wireless sensor network. Many schemes have been developed to address this problem. However, duty cycling is one of the most efficient schemes for energy saving, which is especially applied for a densely deployed network. In this paper, we propose a novel scheme for duty cycle, which implements an adaptive sleep / wake up strategies in efficient manner. This is an adaptive algorithm which takes decisions on the basis of the current status of the sensor network with the aim of covering each and every activity in the area of interest at each instance of time. The simulation results demonstrate the effectiveness of the algorithm.
© 2014 The Authors.Published by Elsevier B.V.
Peer-review under responsibility of organizing committee of the International Conference on Information and Communication Technologies (ICICT 2014).
Keywords:Duty Cycle; Adaptive algorithm; Coverage ratio; Network lifetime.
1. Introduction
A wireless sensor network (WSN) is a spatially dispersed self-sufficient network1, where a large number of sensor
nodes are deployed to monitor a target area with respect to several environmental entities, such as vibration, temperature, motion, pressure and so on. The sensor nodes read local data autonomously in unattended manner and
*Kumar Nitesh. Tel.: +91-8235261235
E-mail address- kumarnitesh.ism@gmail.com
© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of organizing committee of the International Conference on Information and Communication Technologies (ICICT 2014)
send it to a remote base station called sink. Initially the development of WSNs was initiated for military applications such as intrusion detection and battlefield surveillance. However, WSNs have gradually become popular for their uses in industry, underground mine, disaster management2, health monitoring for machines3, environment, habitat
monitoring4 and health care5. However, the major bottleneck of the WSN is that the sensor nodes are constrained to
energy as they are operated on small batteries. Therefore, many researches have been carried out for energy saving of the sensor nodes, which include energy efficient clustering6,7, routing protocols8, energy aware MAC protocol9
and low power radio communication hardware10. The sensor nodes consume energy mainly due to sensing,
processing and receiving/ transmitting data. The radio subsystem usually consumes considerably more energy as compared to sensing and processing units. However, during idle state, the transceiver consumes almost equivalent power11-13. On the contrary, they consume considerably less energy during the sleep mode (i.e., in low-power mode).
One of the most effective approaches for saving this energy is duty cycle, where the radio subsystem of a node is switched between sleep and wake up mode. Designing a proficient duty cycling scheme is not simple, as it introduces extra delays in the message delivery. Sometimes sensor nodes miss some event due to inefficient schemes of duty cycling. Both message latency and data loss are very critical parameters for some applications. For example, the detection of a chemical leakage should be done as early as possible because delay can be hazardous and hence high latency may not be tolerated. Most of the techniques that employ duty cycling use wake up and sleep periods which are specified prior to the deployment of the WSN which cannot be modified and hence perform non-optimally. Therefore, an adaptive scheme of sleep/wake up schedules needs to be essentially addressed.
In this paper, we propose an efficient scheme for duty cycling, which is adaptive in nature. The scheme can conserve energy of the sensor nodes significantly by scheduling their sleep timing in an efficient manner so that the lifetime of the network can be increased further with the constraint of preserving the functionality of the network. The proposed technique delivers two main advantages. First, it is not associated with any particular medium access control (MAC) protocol and hence it can be used with various sensor platforms. Second, it is capable to quickly adapt the sleep/wake up periods of every node to the working conditions (such as network congestion, traffic demand and so on). This results in better utilization of energy. We perform extensive experiments on the proposed scheme and evaluate it with respect to the various performance metrics. Now onward, nodes and sensor nodes, sleep and deactivate, wake up and activate will be used interchangeably in the respective manner.
The organization of the paper is as follows. Section 2 describes some related work. Section 3 describes the duty cycle and energy models. In Section 4, we present the proposed algorithm. The experimental results and its analysis of the proposed algorithm are described in section 5. Section 6 concludes the paper.
2. Related Works
Researchers have proposed several techniques for energy conservation of the sensor nodes. However, we mainly focus on duty-cycling as it is related to our present work. There are two basic approaches for duty cycling, namely Topology Control (TC) and Power Management (PM) in which the former deals with node redundancy and uses minimum number of nodes to work to guarantee the network connectivity. The remaining nodes keep their radio inactive and thus save energy, which in turn increases the network lifetime. The authors in3,14 have given
comprehensive studies on TC protocols. Power Management (PM) protocols are intended to synchronize the sleep periods of neighboring sensor nodes such that the active nodes can cover the whole area of interest and can communicate with other nodes.PM protocols are so flexible that it can be implemented at different layers of the WSN. It can be implemented by embedding duty cycling with the MAC layer protocols. It can also be used as an autonomous sleep/wake up protocol at the application or network layers. The duty-cycle MAC protocols15-23were
proposed with the basic idea of optimizing the channel access and energy conservation. However, most of them are not practically feasible on the platform of WSN. On the contrary, independent sleep/wake up technique is more flexible and can be personalized according to the requirement of any application. This technique has also the property that it can be incorporated with any MAC protocol. The scheme proposed in this paper can be categorized as the independent sleep/wake up protocol. This duty cycle (independent sleep/wake up) scheme can be broadly
classified into three different categories: on demand, asynchronous and scheduled rendezvous. The on-demand schemes propose to activate any node just before the time it is required. Many different techniques24,25 have also
employed two different types of radios. The first radio referred as data radio and is used during the normal data transfer, whereas the second one is a wake up radio, which sends a low power signal to intimate the sleeping node to get ready for communication. Moreover the use of two radio subsystem is bit costly and is practically not feasible but provides very high energy efficiency with the assurance of low message latency. Several techniques26,27 have
also been developed by using asynchronous scheme. In such techniques, nodes can just wake up whenever it wants and it can still communicate with its neighbors. The asynchronous schemes are very simple to implement but generally have high latency in message forwarding hence is not desirable. The scheduled rendezvous scheme is the last category where the nodes are required to be synchronized. Here TDMA schedule are created at the application layer, and the actual data transfer is carried out by using the MAC protocols. Flexible Power Scheduling (FPS) 28-29uses this technique and utilizes an on-demand mechanism for channel reservation and dynamically adapts to the
traffic demands. However, like TDMA schemes15,16,19,23 FPS do have some drawbacks. It has a limited scalability
and also lacks in flexibility to adapt the changes in traffic and/or topology. Most of the solutions proposed on the basis of scheduled rendezvous approach basically consist of simple duty-cycling techniques. Our approach is a topology control duty-cycling technique where, duty cycle is implemented simply by switching a node between on and off state depending upon the current scenario of the network.
3. Model Assumptions
3.1. Duty cycle/Network Model
It is observed that the idle periods play a crucial role in saving energy for WSNs32. Most of the existing radios
used in Sensor nodes support various modes of operation, such as receive/ transmit mode, sleep mode and idle mode. In idle mode, the radio is not communicating, but its circuitry remains turned on, resulting in energy consumption. This energy is slightly less than that the energy required during the transmitting or receiving states. Therefore, a better way is to shut down the radio during the idle mode30. In our proposed scheme, we assume a densely deployed network in which the sensor nodes are deployed arbitrarily and are stationary after deployment. We also assume that the location of all sensor nodes is known by using inbuilt GPS or some other techniques such as described in31-33.We also assume that the target region is a square/rectangular grid. However, in a real scenario,
any target region will be a closed polygon. In that case, we subscribe it with a square grid as shown in Fig. 1. The whole region of the WSN is divided into square cells of equal size.
Fig. 1. Subscribing a polygon target region
All the communications are over the wireless links established between the nodes. We use the following terminologies in the proposed algorithm.
1. S – {s1, s2,… Sn} Set of sensor nodes.
2. n - Number of sensor nodes.
3. r - Communication range of sensor nodes.
Actual Target region
Subscribed grid Square cell
4. rs - Sensing range of sensor nodes. 5. Cs - Length of each side of each cell. 6. Cn - Number of cells in the region. 7. R - Number of rows in the grid. 8. C - Number of columns in the grid. 9. CR - Coverage Ratio.
10. Active cell- Cell with an active node.
3.2. Energy Model
We assume a similar radio model for energy as in11,13. However, we assume only the free space channels for
communication as the distance between any sender and receiver is less than the threshold in our case. The energy required by the radio to transmit a k-bit message over a distance d is given as follows,
) 1 ( 0 for 4 0 for 2 ) , ( ° ¯ ° ® t d d d mp k elec kE d d d fs k elec kE d k T E ] ]
Here, Eelec is the energy required by the electronic circuit. ζfs and ζmp are the energy required by the amplifier in free space and multi-path respectively. In our work we are only concerned with the free space energy. The radio expends energy to transmit and to receive message34. Eelec depends on several factors such as modulation, digital
coding, filtering and spreading of signal, whereas the amplifier energy depends on the distance between the transmitter and the receiver. In our proposal, we consider the factor of distance and attempt to minimize the energy loss.
4. Our Scheme
The goal of the proposed scheme is to develop an adaptive scheme for duty cycling, which automatically adjusts the on/off activity of the sensor nodes. The basic objective is to achieve low power consumption without missing any event. Firstly, we discuss how to create the cell with their optimal size as follow.
4.1. Optimizing Cell Size
As stated earlier, in our proposed work we divide the whole region of interest into square cells of equal size. If the sensor node is activated at the center of the cell (see Fig. 2), it can cover the whole area and other node can remain in sleep mode. To obtain the optimum size of the cells, we proceed as follows. Let, rs be the sensing range of each sensor node and Cs be the length of each side of the cell. Then, by Pythagoras theorem, we obtain,
Cs2+ Cs2=(2rs)2
Hence, Cs= 2 rs
Fig. 2. Calculating the Cell Size
The idea behind optimizing the cell size is that, by using such sized cell and activating only a single node in it, can remove any redundant data generated, leading to better utilization of communication channel. This also reduces the total data reaching the destination, leading to the reduction of its computational overhead.
2rs
Cs Cs
4.2. Proposed Algorithm
Like other duty cycle techniques, we also activate a single sensor node in a specific cell. However, we perform it with different thoughts stated as follows. Firstly, we optimize the cell size so that a single active node of the cell can cover it completely and secondly, we consider the case of the partial coverage of the sensing area of neighboring cells by an active cell. Generally, the nodes in a cell do not lie at the exact center and thus an activated node of a cell can cover some part of its neighboring cell. A cell is surrounded by eight neighboring cells and at any instance of the operation of the network; it is possible that a cell gets totally covered by the activated neighboring cellsFig. 3. For
activating a sensor node in a cell we proceed as follows.
We first scan the odd rows followed by the even rows of the grid. While scanning each row, we attend the columns in the following order: 1st, last, 2nd, second last… and so on. Note that scanning to activate the cells in this order can result into the covering of the neighboring cells efficiently. Fig. 3 demonstrates how 6 cells (marked by grey color) are covered by their neighboring cell at some intermediate period of network operation. We use the term Coverage Ratio (CR) to refer to the ratio between the total covered area of the deactivated cell by all its neighbors and the total area of the cell. If the CR of any inactive cell satisfies the application requirement, we keep it deactivated; otherwise activate one node of the cell with maximum remaining energy. The specified CR can change on the basis of application scenario.
Since it is an adaptive technique, at any instance of time the decision to activate a cell is taken after the consideration of the effect of its neighboring cells. In a case of densely deployed network, the proposed algorithm has the possibility of many such deactivated cells, which can be activated later and thus simultaneously resulting in an increase in overall network lifetime. The simulation results of the algorithm prove its effectiveness. The proposed algorithm works in two phases: (i) setup phase and (ii) steady phase. In the setup phase, the algorithm selects the exact node to activate. Once the cell is visited and necessary decision regarding its activation is taken, the cell enters into the steady phase. In steady phase, the cell maintains its states for a few rounds and after that both the phases are again carried out throughout the lifetime of the network.
Fig. 3. Coverage Scenario at any instance of time
The pseudo code of the proposed algorithm is given in Fig. 4
.
In a fault free environment, the centralized system runs the algorithm for the given WSN scenario prior to exact implementation and calculates the state of each sensor node for the complete network lifetime. Then, they pass the necessary information to each sensor node about, when they have to be in sleep mode and when they have to be working during the complete network lifetime. Since the centralized system communicates with the sensor node for only once at the beginning, the communication bandwidth afterward will be dedicated to other functionality of the network. The scenario with the faulty environment is a bit different as there is some probability of failure of each node we cannot assume it prior to its occurrence, hence the algorithm have to run throughout the lifetime of the network to decide the appropriate state of each sensor node at any specific time period.- -Active nodes - sleeping nodes
Total number of sensor nodes=68 Total number of cells=30
Total number of cells with nodes=27 Total number of cells with no nodes active nodes=21
Fig. 4. Pseudo code of proposed algorithm
5. Simulation Result
The simulation of the proposed algorithm was carried out on Matlab-R2013a. The input to our algorithm is the position (coordinates) of the sensor nodes after deployment. The experiments were performed in a 220 × 220 m2
area with variable number of sensor nodes and appropriate number of gateways to cover them. Each sensor node initial has energy of 2 joules. A node is considered to be dead when energy level reaches to 0 joules and the network is considered to be dead if half of its sensor node dies. In our simulations, the typical energy model and parameters are set same as35. For the experiments, the parameters used for communication energy were set as: Eelec=50nJ/bit,
ζfs=10pJ/bit/m2 and ζmp= 0.0013pJ/bit/m4. The energy for data aggregation was set as 5 nJ/bit/signal. We also
assumed that the message packet size to be of 4000 bits and that of CR to be 80%. Hence, the total energy required by any sensor node for a single round is,
P×(Eelec + ζfs × r2)+Eda (2)
where, P is the packet size of data transmitted in each round and Eda is energy for data aggregation. We assumed that once the node enters into the steady phase, it would remain in it for 60 rounds. After several simulations of the proposed algorithm, the recorded result was compared with the two scenariosFig. 5, first a basic sensor network with
no duty cycle where, all the sensor nodes are at their active state throughout the lifetime and the second is an asynchronous duty cycling, where a sensor node from each cell of the grid is activated, independent of the current scenario. The result of the proposed algorithm is comparatively better and acceptable.
The reason for the increase in the network lifetime is simply because of the reduction of the total number of active cells at a time. Since this is an adaptive algorithm and takes decisions on the basis of the current scenario, it does not activate cells blindly. The algorithm is greedy about power conservation and emphasizes to save the power by switching off the unnecessary or redundant nodes. This saves a lot of energy for future use, resulting to the prolonging of the network lifetime. Table 1 shows the number of active cells required for 1st eight rounds. It also
Input: Coordinates of all sensor nodes Output: Activated nodes
Step 1: Divide the whole area of interest into square grids of sides Cs= 2 rs Step 2: Repeat step 3 to 5 until the total number of alive nodes > n/2; Step 3: For i=1 to Cn
Find the sensor nodes within the ith cell of the grid.
Endfor
Step 4: For i=1 to R incremented by 2 each time (i.e. attending odd rows) For j=1 to C
Visit each cell in the order (1st,last,2nd ,second last…so on) and check if it is already covered (by
nodes activated in neighbouring cells) by more than CR. If so, do not activate any node in the cell otherwise, activate the node with maximum power left.
Endfor Endfor
Step 5: For i=2 to R incremented by 2 each time (i.e. attending even rows) For j=1 to C
Visit each cell in the order (1st, last, 2nd , second last…so on) and check if it is already covered
(by nodes activated in neighbouring cells) by more than CR. If so do not activate any node in the cell otherwise activate the node with maximum power left.
Endfor Endfor
clearly justifies that for any scenario, the total number of active cells required is far less than the total number of cells with at least one node.
Fig. 5. Increase in network lifetime with proposed Adaptive Duty Cycle algorithm Table 1: Active Cells in First Eight Successive Rounds
The total numbers of active cells in successive rounds are very similar to these and we cannot present it due to space limitation. It is also observed that for a particular scenario the result of the algorithm varies within a range because of the chance in the positions of sensor nodes, but it has been observed that for a particular scenario the result of the proposed algorithm is always better than the asynchronous duty cycling.
6. Conclusion
In this paper, we have presented an adaptive algorithm to implement sleep/wake up duty cycle scheme in the network layer of a WSN. We have shown that the proposed algorithm is adaptive in nature and can take any decision regarding the state of the radio subsystem on the basis of current scenarios of the network. The algorithm has been simulated rigorously and simulation results show that the proposed scheme performs better than the asynchronous duty cycle in terms of network lifetime.
References
1. I.F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci: Wireless sensor networks: a survey, Computer Networks, 2002; 38:393– 422.
2. G. Platt, M. Blyde, S. Curtin, and J. Ward: Distributed wireless sensor networks and industrial control systems - a new partnership, in Proceedings of the 2nd IEEE workshop on Embedded Networked Sensors (EmNets05), Washington, DC, USA, 2005; 157–158. 3. A. Mainwaring, D. Culler, J. Polastre, R. Szewczyk, and J. Anderson: Wireless sensor networks for habitat monitoring, in Proceedings of
the 1st ACM international workshop on Wireless sensor networks and applications (WSNA02),2002; 88–97.
4. T. He, P. Vicaire, T. Yan, L. Luo, L. Gu, G. Zhou, R. Stoleru, Q. Cao, J. A. Stankovic, and T. Abdelzaher: Achieving real-time target tracking using wireless sensor networks, in Proceedings of the 12th IEEE Real-Time and Embedded Technology and Applications Symposium (RTAS06), 2006; 37–48.
n rs r CR Non empty cells Cells with active nodes in 8 successive rounds
150 18 25 .8 73 66 59 60 58 60 58 59 58 250 18 25 .8 77 71 63 63 64 63 61 63 64 350 18 25 .8 79 74 64 61 62 59 60 62 61 500 18 25 .8 81 73 63 70 63 64 61 64 62 500 16 25 .8 99 88 79 78 78 80 77 79 76 500 14 25 .8 123 115 105 95 91 96 90 93 92
5. J. A. Stankovic, Q. Cao, T. Doan, L. Fang, Z. He, R. Kiran, S. Lin, S. Son, R. Stoleru, and A. Wood: Wireless sensor networks for in-home healthcare: in Proceeding of High Confidence Medical Devices, Software, and Systems (HCMDSS05), 2005; 2–3.
6. Hamed Abbasi, Mohamad Younis: A Survey on clustering algorithms for wireless sensor networks. Computer Communications, 2007; 30:2826-284.
7. Md Azharuddin, Pratyay Kuila, Prasanta K. Jana, Energy efficient fault tolerant clustering and routing algorithms for wireless sensor networks, Computers & Electrical Engineering, http://dx.doi.org/10.1016/j.compeleceng.2014.07.019.
8. Kemal, A., Mohamed, Y., 2005. A survey on routing protocols for wireless sensor networks. Ad Hoc Networks 3:325–349. 9. Ahmad, A., et al., 2012. MAC Layer Overview for Wireless Sensor Networks. In: CNCS 2012;16-19.
10. Calhoun, Benton H., et al. "Design considerations for ultra-low energy wireless microsensor nodes." Computers, IEEE Transactions on 54.6; 2005; 727-740.
11. V. Raghunathan, C. Schurgers, S. Park and M. B. Srivastava: Energy Aware Wireless Microsensor Networks, IEEE Signal Processing Magazine, March 2002;19:40-50.
12. W. Heinzelman, A. Chandrakasan and H. Balakrishnan: Application specific protocol architecture for wireless microsensor networks, IEEE Transactions on wireless communications, 2002; 1:4:660-670.
13. D. Ganesan, A. Cerpa, W. Ye, Y. Yu, J. Zhao, D. Estrin: Networking Issues in Wireless Sensor Networks, Journal of Parallel and Distributed Computing, 2004;64:799-814.
14. H. Karl and A. Willig: Protocols and Architectures for Wireless Sensor Networks, Chapter 10 (Topology Control), Wiley, 2005. 15. P. Santi, Topology Control in Wireless Ad Hoc and Sensor Networks, ACM Computing Survey, June 2005; 37:2:164-194.
16. W. Ye, J. Heidemann, and D. Estrin: An Energy-efficient MAC Protocol for Wireless Sensor Networks, Proc. IEEE INFOCOM 2002, New York, USA, 2002; 23-27.
17. K. Arisha, M. Youssef, M. Younis: Energy-aware TDMA-based MAC for Sensor Networks, Proc. IEEE IMPACCT ‘02, New York City (USA), May 2002.
18. T. van Dam and K. Langendoen, An Adaptive Energy-Efficient MAC Protocol for Wireless Sensor Networks, Proc. ACM SenSys 2003, Los Angeles, USA, Nov. 2003.
19. IEEE 802.15.4, Part 15.4: Wireless Medium Access Control (MAC) and Physical Layer (PHY) Specifications for Low-Rate Wireless Personal Area Networks (LR-WPANs),2006.
20. V. Rajendran, K. Obracza, J. J. Garcia-Luna Aceves: Energy-efficient, Collision-free Medium Access Control for Wireless Sensor Networks, Proc. ACM SenSys 2003, Los Angeles, USA, November 2003.
21. J. Polastre, J. Hill and D. Culler: Versatile Low Power Media Access for Sensor Networks, Proc. ACM SenSys 2004, November 2004. 22. W. Ye, J. Heidemann and D. Estrin: Medium Access Control with Coordinated Adaptive Sleeping for Wireless Sensor Networks,
IEEE/ACM Transactions Networking,Jun. 2004; 12:3:493-506.
23. G. Lu, B. Krishnamachari and C.S. Raghavendra: An Adaptive Energy-efficient and Low-latency MAC for Data Gathering in Wireless Sensor Networks, Proc. PDSP 2004, April 2004.
24. J. Li, G. Lazarou: A Bit-map-assisted energy-efficient MAC Scheme for Wireless Sensor Networks, Proc. Int’l Symp. on Information Processing in Sensor Networks (IPSN 2004), Berkeley, USA, 2004.
25. C. Schurgers, V. Tsiatsis, M. B. Srivastava: STEM: Topology Management for Energy Efficient Sensor Networks, Proc. IEEE Aerospace Conference 2002, Big Sky, USA, March 2002; 10-15.
26. X. Yang, N. Vaidya: A Wake up Scheme for Sensor Networks: Achieving Balance between Energy Saving and Endto- end Delay, Proc. IEEE Real-Time and Embedded Technology and Applications Symposium (RTAS 2004), 2004.
27. V. Paruchuri, S. Basavaraju, R. Kannan, S. Iyengar: Random Asynchronous Wake up Protocol for Sensor Networks, Proc. of BROADNETS ‘04, 2004.
28. R. Jurdak, P. Baldi and C. V. Lopes: Adaptive Low Power Listening for Wireless Sensor Networks, Transactions on Mobile Computing, Aug. 2007; 6:8: 988-1004.
29. B. Hohlt, L. Doherty and E. Brewer: Flexible Power Scheduling for Sensor Networks, IEEE and ACM International Symposium on Information Processing in Sensor Networks, April 2004.
30. B. Hohlt and E. Brewer: Network Power Scheduling for TinyOS Applications, Proc. IEEE Intl Conf. on Distributed Computing in Sensor Systems (DCOSS 2006), San Francisco (USA), 2006.
31. L. M. Feeney and M. Nilsson: Investigating the energy consumption of a wireless network interface in an ad hoc networking environment, in IEEE Conference on. Computer Communications (INFOCOM), 2001;1548–1557.
32. Chang-Soo Ok et al., Distributed Energy balanced routing for wireless sensor networks, Computer and Industrial Engineering 57 2009; 125-135.
33. IEEE, Standard for Information Technology-Telecommunications and Information Exchange Between Systems - Local and Metropolitan Area Networks - Specific requirements Part 15.4: Wireless Medium Access Control (MAC) and Physical Layer (PHY) Specifications for Low-Rate Wireless Personal Area Networks (WPANs) 2006.
34. P. Baronti et al., Wireless sensor networks: A survey on the state of the art and the 802.15.4 and ZigBee standards, Computer Communications 2007; 1655-1695.
35. YIN Rong-rong , LIU Bin, LI Ya-qian, HAO Xiao-chen: Adaptively fault-tolerant topology control algorithm for wireless sensor networks, science direct, 2012; 19:13–18.
