Abstract—In spite of various constraints present in Wireless Sensor Networks (WSN), it has become popular in various application domains which need useful information. Among these constraints, energy conservation is the most important aspect. One of the known strategies to save energy and to prolong the lifetime of the WSNs is topology control. In this paper, a reliable energy-efficient topology control algorithm in wireless sensor networks is proposed. The proposed algorithm considers the residual energy and number of neighbors of each node for cluster formation, which is critical for well-balanced energy dissipation of the network. Knowledge-based inference approach is employed to select cluster head. Finally, reliable cluster heads are selected that maintain connectivity and coverage with respect to time. The algorithm not only balances the energy load of each node but also provides global reliability for the whole network. Simulation results demonstrate that the proposed algorithm effectively prolong the network lifetime and reduces the energy consumption. Keywords-- Topology Control, Wireless Sensor Networks, Clustering.

I.� INTRODUCTION A wireless sensor network (WSN) consists of a number of

tiny, low-powered, energy-constrained sensor nodes with sensing, data processing and wireless communication components. Sensor nodes in WSNs are small battery powered devices with limited energy resources, and their batteries cannot be recharged once the sensor nodes are deployed. Therefore, minimizing energy consumption is an important issue in the design of WSNs protocols. Clustering is an effective solution in reducing energy consumption, prolonging the lifetime of the networks and providing network scalability [1]. Another similar development is the wireless ad hoc network, which is characterized by independent mobile hosts without the support of fixed infrastructure [2]. Both sensor network and ad-hoc network share the feature of multi-hop communication.

Extensive research has been conducted in topology control of wireless sensor networks. Power control technique is addressed in [3, 4] that reduce interference and improve throughput. Topology control by tuning transmission powers is discussed in [5, 6, 7]. Both IEEE 802.11 and Bluetooth support low-power modes [2, 8]. Designing low-power modes in 802.11 which is based on multi-hop network is addressed in [9].

Among several challenging issues, in our paper, we discuss the problem of topology construction in a hierarchical wireless sensor network. In most current designs, nodes are deployed using random and uniform distributions as they are the popular schemes due to their simplicity. However, node deployment schemes have a greater impact on system functionality as well as on lifetime. The traditional random and uniform distributions are responsible for the infamous sink routing-hole phenomenon. In this phenomenon, nodes nearer to the sink tend to get depleted sooner in terms of their batter power than the nodes that are farther from the sink.

WSNs are also inherently different from existing system and network. Unlike traditional mobile ad-hoc networks (MANETs), WSN consists of one or more nodes (sink) that are designated as gateways between the networked system and users. These sinks play a critical role as loss in connection to the sink directly leads to the failure of the entire WSN. Furthermore, not only the sink but the direct neighbors of the sink are also essential for network functionality. Direct neighbors are established to maintain connectivity between the sink and distant data nodes. When the set of direct neighbors fails, data cannot be delivered to the sink due to the lack of forwarders and thus the whole system fails. Often the sinks are more powerful with permanent power sources. Their failure may not be frequent. However, direct neighbor nodes are common sensor nodes with limited resources. These nodes are vulnerable to failure (e.g., running out of energy). An interesting observation is that for most data gathering applications, the forwarding workload of sensors increases inversely with their distances to the sink. A node closer to the sink usually has a higher relay workload than those of distant nodes. Accordingly, direct neighbors of the sink have the greater transmission workload. They are likely to deplete their

A Reliable Two-Tier Energy-Efficient Topology Building Algorithm for Wireless Sensor

Networks

Chiranjib Patra1, Matangini Chattopadhyay2, Parama Bhaumik3, Munmun Bhattacharya3, Saswati Mukherjee2

1Department of Information Technology, Calcutta Institute of Engineering and Management, Kolkata 2School of Education Technology, Jadavpur University, Kolkata

3Department of Information Technology, Jadavpur University (Salt Lake Campus), Kolkata 1chiranjibpatra@gmail.com, 2chttpdhy@yahoo.com

3bhaumikparama@rediffmail.com, 3munmun@it.jusl.ac.in, 2saswatimuk@gmail.com

���������� �������������� ���������������� ����������

��� � ���� !��� !"��"#!�$���%������&&&

��'

energy at the very beginning. When they fail, no matter how many remaining nodes are still active, none of them can communicate with the sink. From the sink and user’s perspective, the whole system fails if the sink is isolated from the rest of the sensor nodes. Note that multiple sinks or mobile sinks cannot eliminate the problem as many sink routing-holes will be generated accordingly. Data funneling and aggregation [15] techniques may alleviate the problem to some extent, but cannot eliminate the problem either. The main objective of our work is to provide a long-term continuous connectivity. We attempt to address the problem by designing a power-aware topology management scheme. Major attention is given to the connectivity of the network as it is a prerequisite for other purposes such as sensing coverage. Note that without a valid data path, an active node has the same role as a dead one. The sensing area covered by unconnected nodes is still inaccessible. However, many traditional topology management approaches do not consider differential loss of energy in different sub-areas into consideration. Consequently, they may work well for some snapshots of the network (e.g., in the initial stage), but may not guarantee the quality of service in the whole system lifecycle.

Our main contributions include production rules for managing network topology. Cluster-head selection mechanism is developed that considers residual energy, number of neighbors, and centrality of each node and uses production rules for cluster-head selection. Reasoning mechanism is used to develop reliable multi-hop routing algorithm among cluster heads. The resulting reliable energy-efficient two tier routing protocol is implemented and evaluated through simulations.

We have organized the paper as follows. In Section II, we describe the A3 algorithm. Section III presents the production rules. In Section IV assumption, proposed algorithm and analysis are described. In Section V some experimental results are given to characterize the performance evaluation. Finally we draw our conclusions.

II.� OVERVIEW OF A3 PROTOCOL In this section a brief description of A3 protocol [14] is presented.

A3 algorithm assumes no prior knowledge about the position or orientation of the wireless sensor nodes. Therefore, nodes do not have an exact geometric view of the topology. However, nodes can determine how distant a node is from another node, based on the received signal strength. This information is enough to select a close-to-optimal Connected Dominating Set (CDS) tree, based on the belief that farther nodes will offer more area of communication coverage.

The A3 algorithm is executed in three phases: Neighborhood Discovery, Children Selection and Second Opportunity.

1) Neighborhood Discovery: The CDS building process is started by a pre-defined node that might be the sink node just after the nodes are deployed. The sink node starts the protocol by sending an initial Hello Message. This message allows the neighbors of the starting node to know its ‘parent’ node. If the node that has not been covered by another node receives the Hello message, it sets its state as covered, adopts the sender

as its ‘parent’ node and replies with a Parent Recognition Message. This message also includes a selection metric that is calculated based on the signal strength of the received Hello Message. The parent node sorts the candidates using the metric. If another node has already covered the receiver, it ignores the Hello Message. The A3 algorithm uses four types of messages: Hello Message, Parent Recognition Message, Children Recognition Message, and Sleeping Message. If a parent node does not receive any Parent Recognition Message from its neighbors, it goes to sleep state.

2) Children Selection: The parent node sets a timeout to receive answers from its neighbors. Each metric is stored in a list of candidates. Once this timeout expires, the parent node sorts the list in decreasing order according to the selection metric. The parent node then broadcasts a Children Recognition Message that includes the complete sorted list to all its candidates. During the timeout, nodes wait for Sleeping Messages from their brothers. If a node receives a Sleeping Message during the timeout period, it turns off. This is because the recipient of the Sleeping Message understands that one of its brothers is more appropriate to become a part of the CDS tree. Based on this scheme, the best node according to the chosen metric sends a Sleeping Message thereby blocking any other node in its range. Therefore, only other candidate nodes outside its area of coverage have the opportunity to start their own generation process.

3) Second Opportunity: Although this methodology works very well, there are some cases where a node is sent to sleep to avoid bottleneck. In order to avoid this situation, every node sets a timer once it receives the Sleeping Message to send a Hello Message and starts its own building process.

III.� BACKGROUND OF PRODUCTION RULES The production rules are used for its simplicity to achieve

the desired goal for the system. The production rules used here are based on the parameters of WSN. These parameters are: the number of neighbors and the residual energy of the node. The justification is as follows. The Residual Energy (Er ) is the remaining energy of each node. More Er implies more data can be processed and transmitted and the node will have a longer life time. The number of neighbors, N, means The number of neighbors of a cluster head. It is reasonable to select a node as a cluster head so that the node has more neighbors.

Before firing a rule, let W1 and W2 are two weight values, associated truth degrees be a1 and a2 respectively of Er and N. After firing, physically and logically permissible combination rules are formed.

Rule 1: A = a1*W1+a2*W2 Rule 2: a1= (A *W), a2 = (A *W) Rule 3: A = Max (a1*W1, a2*W2)

���

Rule 1

Rule 2 Rule 3

Fig-1: State diagram for Rules

To illustrate Rule 1 we consider the situation before firing as depicted in the Figure 1.

The parameters a1 and a2, for the truth values of Er and N it may be defined as

ai = Present value / Max value So a1(truth value) may be defined as

a1 = Present battery residual energy at time t of the node under consideration/Unused battery energy of the node under consideration.

Similarly a2 truth value can be obtained as a2 = Present number of neighbors at time t of the node under consideration / 6. The choice of 6 in the denominator is taken from [10].

Lastly A is defined as the chance function which maps from places to real values in the range from zero to one. Larger value of A indicates greater opportunity for the node to become a cluster head.

IV.� ASSUMPTIONS, PROPOSED ALGORITHM AND ANALYSIS

A. Assumptions The consideration of the cluster head selection algorithm is as follows:

1.� The nodes do not know about their positions or orientation hence prior idea of topology is not known.

2.� All nodes are located in a two dimensional space and have a perfect communication coverage disk.

3.� Every node starts in an unvisited state. 4.� The sink is the initiator of the process and it has a large

amount of energy e.g., a base station. 5.� The time ti of the ith node broadcasting the CH (Cluster

Head) message is ti=Ai*T [T = predefined max time allowed for CH competition].

6.� As the cluster heads change after time T+Δ, to connect between the cluster heads we use an approximate connected dominating set algorithm to form a reliable multihop connection among the cluster

heads.[ Δ= the interval between two runs of CH selection algorithm].

7.� There is no packet loss at the Data Link Layer.

B. Algorithm According to cluster head election chance Ai, we can get the time ti of the ith node broadcasting the cluster declaring message CH as shown below. ti = Ai × T, where T is predefined as the maximum time cluster head is competing.

If a sensor node does not receive the message CH before time t expires, it will broadcast message CH to its neighbor nodes. If the jth node receives some CH message before the time tj expires, the jth node will not compete for cluster head selection and will construct a cluster head candidate table containing the sender of the CH messages. Then, the jth node selects the node with the maximum chance as its cluster head. If there are multiple nodes having the same maximum chance, the node having more energy is selected as the cluster head. Finally, the node transmits the JOIN message to the cluster head.

C. 1. Cluster head selection algorithm INPUT: the cluster head election chance Ai of each node Ni. OUTPUT: the set of cluster heads. Step 1: Set the timer ti of the node Ni for competing the CH start. Step 2: while (ti is not expired) if (Ni does not receive CH msg) Ni broadcast CH msg to neighbors Else if (Ni receive 1 CH msg from Nj ) Ni selects Nj as CH Ni transmits the Msg JOIN to Nj

else if (Ni receives m CH msgs from other m nodes) from m nodes Ni selects Nk with highest cluster head election chance as CH Ni transmits the msg JOIN to Nk End if End while 2. Cluster heads connecting algorithm After the clusters are formed there is a need to connect the cluster heads to aggregate data and forward it to the base station through a multihop path. As the cluster heads change after time T, we use an approximate connected dominating set algorithm to form a reliable multihop connection among the cluster heads.

Step 1: Calculate , xi(α)( maximum degree of

neighbors in distance 2, a value due to primal-dual algorithm [11]) Step 2: Become a dominator (i.e. go to the dominating set) with probability

Step 3: Send status (dominator or not) to all neighbors Step 4: If no neighbor is a dominator, become a dominator yourself

���

D. Analysis Proposition 1: The time complexity of cluster head selection is O(N). Proof: At the beginning of each round, every node sends a HELLO message to calculate the number of neighbors . The maximum number of HELLO messages is N. As soon as the cluster head selection starts, for n cluster heads, it will broadcast n CH messages. In the meantime, N-n number of cluster members will broadcast JOIN messages. Hence, the total number of messages is N+n+N-n=2N and thus the time complexity is O(N)

V.� PERFORMANCE EVALUATION Above algorithms described above are tested with Public Atarraya version 1.2.3 [13] .The simulations were carried with the following parameters and constants.

Parameter Value Initial energy per node

1J

Deployment area 600 m X 600 m Number of sinks 1 Node distribution Uniform (100,) Communication radius

100 m

Sensing radius 20 m Energy Consumption

Eelec = 50nJ/bit; Eamp = 10pJ/bit/m2 Short Messages = 25Bytes Long Messages = 100Bytes Idle state energy consumption assumed negligible

Table 1: Simulation Parameters

The process of topology building involves 2 tier processes. Firstly, cluster heads are selected and then inter cluster connecting algorithm is executed so that multiple hops can be created and data transfer to the sink can take place. The comparison is done with the existing A3 algorithm which is an established protocol.

Simulation diagrams

Figure 2 Figure 3

Figures 2 and 3 show the cluster head selection due to A3

protocol [14] and the proposed protocol for one problem instance. From Figure 3, it can be seen that the number of cluster heads is 34 while the number of cluster heads is 31 in case of A3 protocol as can be seen in Figure 2. This is really significant in terms of better coverage and efficiency in terms of data handling.

In order to shoe the superiority of the proposed approach, we present the following performance graphs.

Figure 4. Energy Spent ratio vs Time

Figure 5. Sensing Area Coverage vs Time

Figure 6. Number of Dead nodes vs time

Figure 4 depicts that in the proposed scheme, energy is

spent judiciously to build the topology compared to the A3 algorithm. Similarly in Figure 5, the sensing area coverage due to the proposed algorithm is initially less but after some time it reaches a steady state which is at the same level as A3 protocol. Combining the interpretation of the graphs obtained in Figures 4 and 5, it can be said that A3 protocol is time efficient whereas the new proposed protocol is energy efficient. Figure 6 shows that the new protocol prolongs network life because same number of dead nodes as A3 algorithm is obtained at a later time point.

VI.� CONCLUSIONS AND FUTURE WORK The proposed approach shows that knowledge based reasoning can be successfully used in building topology management algorithms so that it can be competitive with other established algorithms like A3. In future, the results obtained can be analyzed and interpreted by varying the number density of nodes (i.e. sparse networks and dense networks.) and also for heterogeneous environments.

REFERENCES [1]� Heinzelman, W.B.; Chandrakasan, A.P.; Balakrishnan, H. An

application specific protocol architecture for wireless microsensor networks. IEEE Trans. Wirel. Commun. 2002, 1, 660-670.

dead_nodes vs time

�� �

�� �� '� ��

��� ���

� �� �� !� �� (� time

��) �!

sensing_area_coverage Vs time

� �$�� �$�� �$�' �$�� �$�

�$��

� �� �� !� �� (� time

��) �!

energy_spent_ratio VS time

� �$�����$�����$���!�$�����$���(�$���'�$�����$�����$�����$���

� �� �� !� �� (� time

��) �!

���

[2]� C. E. Perkins. Ad Hoc Networking. Addison-Wesley, 2000. [3]� C. F. Huang, Y. C. Tseng, S. L. Wu, and J. P.Sheu.”Increasing the

Throughput of Multihop Packet Radio Networks with Power Adjustment” .International Conference on Computer Communications and Networks, 2001.

[4]� Cheng, M.X.,Xuan Gong, Lin Cai, Xiaohua Jia, "Cross-Layer Throughput Optimization With Power Control in Sensor Networks". IEEE Transactions on Vehicular Technology, Sept. 2011,Page(s): 3300- 3308.

[5]� Kyung-Joon Park, LaeYoung Kim, and Jennifer C. Hou"Adaptive Physical Carrier Sense in Topology-Controlled Wireless Network" .IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 9, NO. 1, JANUARY 2009.

[6]� R. Ramanathan and R. R. Hain. Topology Control of Multihop Wireless Networks using Transmit Power Adjustment. In IEEE INFOCOM,pp. 404–413, 2000.

[7]� R. Wattenhofer, L. Li, P. Bahl, and Y.-M. Wang.Distributed Topology Control for Power Efficient Operation in Multihop Wireless Ad Hoc Net-works. In IEEE INFOCOM, pages 1388–1397, 2001

[8]� J. C. Haartsen and S. Mattisson. Bluetooth -A New Low-Power Radio Interface Providing Short-Range Connectivity. Proceedings of the IEEE, October 2000.

[9]� Y. C. Tseng, C.-S. Hsu, and T. Y. Hsieh. Power-saving protocols for ieee 802.11-based multi-hop ad hoc networks. In IEEE INFOCOM, 2002.

[10]�L. Kleinrock and J. Silvester, "Optimum Transmission Radii for Packet Radio Networks or Why Six is a Magic Number," in Conference Record, National Telecommunications Conference, Birmingham, Alabama, December 1978, p. 4.3.2–4.3.5

[11]�Carsten Moldenhauer “Primal-Dual Approximation Algorithms for Node Weighted Steiner Forest on Planar Graphs” ICALP 2011, Part I, LNCS 6755, pp. 748– 759, 2011. Springer-Verlag Berlin Heidelberg 2011

[12]�Stefan Schmid and Roger Wattenhofer. “Algorithmic Models for Sensor Networks“14th International Workshop on Parallel and Distributed Real-Time Systems (WPDRTS), Island of Rhodes, Greece, April 2006.

[13]�Migue lA. Labrado , (10th July 2012) , http://www.csee.usf.edu/~labrador/Atarraya/download.htm

[14]�Pedro Wightman and Miguel A. Labrador, “A3: A Topology Control Algorithm for Wireless Sensor Networks,” IEEE Globecom, November 2008.

[15]�D. Petrovic, R. C. Shah, K. Ramchandran, and J. Rabaey, “Data Funneling: Routing with aggregation and compression for wireless sensor networks,” in Proceedings of IEEE Sensor Network Protocols and Applications (SNPA), 2003.

�(�