You are on page 1of 5

Energy Efficient Information fusion in Wireless Sensor Networks

Ajai H K 1
Abstract: Sensor networks are event-based systems that dier from
traditional communication networks in several ways: sensor networks have severe energy constraints, redundant low-rate data, and many-to-one ows. The problem of energy-efficient reliable wireless communication in the presence of unreliable or lossy wireless link layers in multi-hop wireless networks has been a serious issue. In this paper, A novel approach for data gathering in wireless sensor networks is proposed which involves data fusion with broken Links. The number of transmissions by the sensor node is limited by approximating the sensor data at the sink using prediction Approach with buffered datasets. By adaptively using redundant transmission on fusion routes without acknowledgments, packets with more information are delivered with higher reliability. The proposed method is effective and greatly reduces the number of transmissions to the sink node which considered the compromised nodes data forwarding. That optimally solves the minimum energy reliable communication problem in presence of unreliable links.
2

Keywords-- WSN, data fusion, energy efficiency, Energy


Efficiency, Data aggregation I. INTRODUCTION

ireless communication networks have been deployed at an increasingly fast rate, and are expected to reshape the way we live in this physical world. The wireless sensor networks of the near future are envisioned to consist of hundreds to thousands of in-expensive wireless nodes, each with some computational power and sensing capability, operating in an unattended mode. Sensor networks are quintessentially event-based systems. A sensor network consists of one or more sinks which subscribe to specic data streams by expressing interests or queries. The sensors in the network act as sources which detect environmental events and push relevant data to the appropriate subscriber sinks. For ex-ample, there may be a sink that is interested in a particular spatiotemporal phenomenon (does the temperature ever exceed 70 degrees in area A between 10am and 11am?). During the given time interval all sensors in the corresponding spatial portion of the network act as event-based publishers. They publish information toward the subscribing sink if and when they detect the indicated phenomenon.
1

Ajai H K. Author is currently pursuing M.E (Communication Systems) in PSN College of Engineering.. e-mail:hkajaime@gmail.com, Ph: +91-9578832032

Because of the requirement of unattended operation in remote or even potentially hostile locations, sensor net-works are extremely energy-limited. However since various sensor nodes often detect common phenomena, there is likely to be some redundancy in the data the various sources communicate to a particular sink Data aggregation has been put forward as an essential paradigm for wireless routing in sensor networks [1],[6]. The idea is to combine the data coming from different sources enroute eliminating redundancy, minimizing the number of transmissions and thus saving energy. This paradigm shifts the focus from the traditional address-centric approaches for networking (nding short routes between pairs of addressable end-nodes) to a more data-centric approach (nding routes from multiple sources to a single destination that allows in-network consolidation of redundant data). There exist two broad approaches to achieve desired reliability of data delivery in unreliable wireless sensor networks: 1) redundant deployment and 2) redundant transmission. However, most k-coverage schemes provide only sensing reliability without considering the transmission reliability. While multipath strategy can offer reliable transmission for single data with controllable reliability, it requires complicated duplicate filter when it works with duplicate-sensitive aggregation functions (such as sum, average, and median.) To increase the information reliability, the main idea of our scheme is to repeatedly transmit data on fusing routes without acknowledgments (ACKs). There are two benefits of this scheme. The first advantage is that it can work together with any data fusion routing algorithm and also any fusion function. The second advantage is that it does not need any ACK control mechanism, and hence, reduces the latency of data delivery. In this paper we study the energy savings and the delay tradeoffs involved in data aggregation and how they are impacted by factors such as source-sink placements and the density of the network. Our proposed solutions assign larger number of transmissions to the nodes with more information while fewer number of transmissions to the odes with less information. The proposed approach uses a simple yet powerful tool of pre buffering strategy to reduce the data losses that exist during the node failure. This paper is organized as follows: Section II proposes data aggregation. Section III proposes energy efficient data fusion while the experimental result is shown in section IV and conclusion is defined in section V

II. DATA AGGREGATION

Data aggregation is the combination of data from different sources, and can be implemented in a number of ways. The simplest data aggregation function is duplicate suppression in the example of gure 1, if sources 1 and 2 both send the same data, node B will send only one of these forward. Other aggregation functions could be max, min, or any other function with multiple inputs. For our modeling purposes in this paper we make a simplifying assumption - the aggregation function is such that each intermediate node in the routing transmits a single aggregate packet even if it receives multiple input packets. We will refer to the information received by the sink when it has obtained the messages transmitted by all sources in a given ows (whether or not these messages are aggregated) as a datum.

Corollary: Assuming an arbitrary placement of sources, and a general network graph G, the task of doing DC rout-ing with optimal data aggregation is NP-hard. The latter follows from the NPcompleteness of the min-imum Steiner problem on Graphs [10].

B. Suboptimal Aggregation The following are three generally suboptimal schemes for generating data aggregation trees that we examine in this paper. 1. Center at Nearest Source (CNS): In this data aggregation scheme, the source which is nearest the sink acts as the aggregation point. All other sources send their data directly to this source which then sends the aggregated in-formation on to the sink. Shortest Paths Tree (SPT) : In this data aggregationscheme, each source sends its information to the sink along the shortest path between the two. Where these paths over-lap for different sources, they are combined to form the aggregation tree. Greedy Incremental Tree (GIT) : In this scheme theaggregation tree is built sequentially. At the rst step the tree consists of only the shortest path between the sink and the nearest source. At each step after that the next source\ closest to the current tree is connected to the tree.

2.

3.

Fig 1: Simple example of data fusion

A. Optimal Aggregation: Say there are k sources, labelled S 1 through S k , and a sink, labelled D . Let the network graph G = (V , E ) consist of all the nodes V , with E consisting of edges between nodes that can communicate with each other directly. With the assumption that the number of transmissions from any node in the data aggregation tree is exactly one, the data aggregation tree can be thought of as the reverse of a mul-ticast tree: instead of a single source sending a packet to all receivers, all the sources are sending a single packet to the same receiver. It is well-known that the multicast tree with a minimum number of edges is a minimum Steiner tree on the network graph. The following can therefore be readily obtained: Result 1: The optimum number of transmissions re-quired per datum for the DC protocol is equal to the num-ber of edges in the minimum Steiner tree in the network which contains the node set (S 1 , ...S k , D ).

This is by no means an exhaustive list, but is representative of some of the data aggregation tree heuristics that can be implemented. In exploring the gains and tradeoffs involved in data-centric protocols, we need to specify performance mea-sures of interest. Energy Savings: By aggregating the information com-ing from the sources, the number of transmissions is re-duced, translating to a savings in energy. Data Fusion Structure: It define the fusion format for message that can be traverser in the network. In a fusion tree, each node fuses all incoming packets with its own to a single packet and sends it out to its parent. Different applications have d ifferent fusion ratio. For example, in camera sensor networks, two spatially over-lapped images can be merged into one picture. In this case, the packet size of the fused picture is smaller than the total size of the two original images but may be larger than the size of any original image. As a result, all packets sent out from different nodes have the same packe t size. We call it as full fus ion . M os t commonly

used fusion functions, such as max/min, sum, average, count, and median, belong to this category.

III. ENERGY EFFICIENT DATA FUSION

Fig 2: Tree fashion Fusion structure

A. Tree Structure Based Fusion: In a tree fusion structure, the reporting nodes send their data to parents, while a parent node aggregates data from children with its own to produce a new data, and then, sends this new data to its parent. This procedure is repeated on the routes from leaves to the root (the sink). Finally, the sink receives the overall data of the network. The whole network is divided into several fusion clusters. In each cluster, nodes send their data to the cluster head through multiple hops, and hence, perform data fusion along routes. Data from different clusters are not fused because of low data redundancy or high fusion cost [3]. For this kind of networks, each cluster is considered as a fusion tree rooted at the cluster head. The optimal tree fusion structure is defined [1] by the node i and j ,denoted by T i and Tj which can be given as

Information Weight:The information weight sent out by a fusion node will be defined as the sum of all information weights it received (including its own). Since all sensor nodes have the same sensing range, let I0 denote the primitive information weight of raw data sensed by each node; let Ii denote the information weight of packet sent by node i; and let Ri denote the successful transmission probability of the packet sent by node i to its parent. Then, as an example in Fig. 2, the information weight sent out by node 1 is given by I1= I2 R2 +I3R3 I4R4 +I0 1

Where IW defines the information weight at the sink end of reception. This can be proven by Hong Lou. [1] which can be simply specified as follow s: Let Sinfo defined the sink information which can be given by total number of information arriving to the sink that can be given by the following equ.

From this example, it is easy to see that the information weight of each node is different even though all their data weights are equal to D. Delay: There is latency associated with aggregation.Data from nearer sources may have to be held back at in-termediate nodes in order to combine them with data from sources that are farther away. Energy Consumption: It defines the amount of energy is being used for the aggregation of the data. Infor mation Reliability: In a sensor network, the sink needs abundant information to get the accurate result and make a decision. In reality, however, not all data packets can be successfully delivered to their parents. we define the desired information reliability, denoted by R as 2

4 By using the Lagrange function to solve the above equation we get the following: 5 Thus we can simply get 3 equation

Similarly the reliability of the Information transmission can also be expressed as

B. Node Failure Bufferings: The Binary Tree is constructed with the incoming elements and the codes framed by traversing the tree as in Adaptive Huffman algorithm. Elements are grouped as nodes and codes are specifically assigned as in Static Huffman algorithm. At every update the weights at every level is checked and updated such that the higher weights will be occupying the initial stages of the tree. This enhances the performance in two ways. It reduces the number of levels in the tree. It brings the maximum possible elements to the top level of the tree. // Creation of root in the receiver. createroot( ) // Traversal of tree in search of node. traverse ( btree ) { if( code[ i ] = 0 ) curr = curr -> lchild if( code[ i ] = 1 ) curr = curr -> rchild } if( curr = node ) read suffix data = data [ suffix ] increment node->wt // Encoding of data and insertion of node in the tree. if ( curr = nyt ) read suffix if( suffix = 5bits) data = suffix|4 if(suffix = 4 bits) data = suffix nyt->lchild = createnyt() nyt->rchild = createnode(diff) nyt = nyt->lchild update(btree) balance(btree) endmodule. The concept of Grouping the data into sets is taken from the Static Huffman algorithm and the concept of Tree construction for dynamic assignment of probabilities is taken from the Adaptive Huffman algorithm The above module set is being define for the identification of the failure

node and then define joining strategy with the existing tree so as to assume the communication as usual.
III. EXPERIMENTAL RESULT We consider 10 sensors uniformly distributed in a square region of size 50m 50m. We assume that each node produces one unit data (I0) and sends it to the sink located at the bottom-right corner. All sensors act as both sources and routers. We also performed a set of experiments with different numbers of sensors and different sizes of elds, the results are similar and omitted here. In order to distinguish the correlation between data originated from two nodes and that among aggregated data, we use a forgetting factor on the correlation coefcient among aggregated data. For example, the correlation between aggregated information at two parent nodes is only a fraction of their own data correlation calculated according to their distance. Throughout the simulation, we use a factor of 0.8. A set of other values are also studied which lead to similar results and are omitted here. For the fusion cost, in the simulation, we assume that q (x, y ) = (x + y ), where denotes fusion cost of unit data. In other words, fusion cost is linear with the total amount of data to be fused. Table I summarizes the parameters and values used in our simulation.

module buffer_encoding( code[ ], btree )

Fig 3 : Data Fusion will be carried among the node for processing.

Fig 4 : wireless communication involves data fusion through different Node .

[7]

[8]

[9]

[10]

[11]

Fig 5 : Data traversing through the nodes that involves data fusion.
[12]

V. CONCLUSION Wireless sensor networks are an important type of resourceconstrained distributed event-based systems. We have modelled and analyzed the performance of data ag-gregation in such networks. We identied and investigated some of the factors af-fecting performance, such as the number of placement of sources, and the communication network topology. We presented some suboptimal data aggregation tree generation heuristics and showed the existence of polynomial special cases. The modelling tells us that whether the sources are clustered near each other or located randomly, signicant energy gains are possible with data aggregation. These gains are greatest when the number of sources is large, and when the sources are located relatively close to each other and far from the sink. Data-centric architectures such as directed diffusion should support a Type of Service (TOS) facility that would permit applications to effect desired tradeoffs between latency and energy. Future work will be to implement spatial correlation based clustering which can be used to schedule the sensor nodes in each cluster to transfer the data. This way the sensor nodes which are not transmitting can be put into sleep mode which can further increase the lifetime of sensor network.
[13]

[14]

[15]

[16]

[17]

[18]

[19]

REFRENCES
[1] Hong Luo, Huixiang Tao, Huadong Ma, Data Fusion with Desired Reliability in Wireless Sensor Networks IEEE Trans. On parallel and distributed system-2011 Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci, A Survey on Sensor Networks, IEEE Comm. Magazine, vol. 40, no. 8, pp. 102114, Aug. 2002. B. Krishnamachari, D. Estrin, and S. Wicker, Impact of Data Aggregation in Wireless Sensor Networks, Proc. 22nd Intl Conf. Distributed Computing Systems, pp. 575-578, July 2002. Y. Liu and S.K. Das, Information-Intensive Wireless Sensor Networks: Potential and Challenges, IEEE Comm. Magazine, vol. 44, no. 11, pp. 142-147, Nov. 2006. H. Luo, Y. Liu, and S.K. Das, Routing Correlated Data in Wireless Sensor Networks: A Survey, IEEE Network, vol. 21, no. 6, pp. 40-47, Nov. 2007. W. R. He inzelman, A. Chandra kasan, and H. B alakrishnan, EnergyEfficient Communication Protocol for Wireless Micro-sensor Networks, Proc. 33rd Ann. Hawaii Intl Conf. System Sciences, Jan. 2000. W. Heinzelman, A. Chandrakasan, and H. Balakrishnan, An Application-Specific Protocol Architecture for Wireless Microsen-sor Networks, IEEE Trans. Wireless Comm., vol. 1, no. 4, pp. 660-670, Nov. 2002.

S. Li nd se y an d C .S . R ag ha ve nd r a, P eg as is : P ow er -E f fi c ie nt Gathering in Sensor Information Systems, Proc. IEEE Aerospace Conf., Mar. 2002. H. Su and X. Zhang, Power-Efficiency Data Gathering Schemes for Wireless Sensor Networks, Proc. 15th Intl Conf. Computer Comm. and Networks, pp. 328-333, Oct. 2006. C. Intanagonwiwat, R. Govindan, D. Estrin, J. Heidemann, and F. Silva, Directed Diffusion for Wireless Sensor Networking, IEEE ACM Trans. Networking, vol. 11, no. 1, pp. 2-16, Feb. 2003. C. Intanagonwiwat, D. Estrin, R. Govindan, and J. Heidemann, Impact of Network Density on Data Aggregation in Wireless Sensor Networks, Proc. 22nd Intl Conf. Distributed Computing Systems, July 2002. H. Luo, J. Luo, Y. Liu, and S.K. Das, Adaptive Data Fusion for Energy Efficient Routing in Wireless Sensor Networks, IEEE Trans. Computers, vol. 55, no. 10, pp. 1286-1299, Oct. 2006. H. Luo, Y. Liu, and S.K. Das, Distributed Algorithm for En Route Aggregation Decision in Wireless Sensor Networks, IEEE Trans Mobile Computing, vol. 8, no. 1, pp. 1-13, Jan. 2009. T. Le, Y. Dong, R. Liu, S. Jha, and Z. Rosberg, Implementation A sp ec ts o f R e li ab le T r a ns p or t P r o t oc o ls in W ir e le ss Se ns or Networks, Proc. Third Intl Conf. Comm. Systems Software and Middleware and Workshops (COMSWARE 08), pp. 574-580, Jan. 2008. Z. Abrams, A. Goel, and S. Plotkin, Set K-Cover Algorithms for Energy Efficient Monitoring in Wireless Sensor Networks, Proc. ACM Intl Symp. Information Processing in Sensor Networks (IPSN 04), pp. 424-432, Apr. 2004. S. Yang, F. Dai, M. Cardei, and J. Wu, On Multiple Point Coverage in Wireless Sensor Networks, Proc. IEEE Intl Conf. Mobile Adhoc and Sensor Systems, Nov. 2005. P.-J. Wan and C.-W. Yi, Coverage by Randomly Deployed W ir e le s s S en s or Ne t w o r k s , IEEE Tr an s. Info r mat io n T he or y, vol. 52, no. 6, pp. 2658-2669, June 2006. A. Shrestha, L. Xing, and H. Liu, Modeling and Evaluating the Reliability of Wireless Sensor Networks, Proc. Reliability and Maintainability Symp. (RAMS 07), pp. 186-191, Jan. 2007. O.B. Akan and I.F. Akyidiz, Event-to-Sink Reliable Transport in Wireless Sensor Networks, IEEE/ACM Trans. Networking, vol. 13, no. 5, pp. 1003-1016, Oct. 2005. J. Lian, Y. Liu, K. Naik, and L. Chen, Virtual Surrounding Face Geocasting with Guaranteed Message Delivery for Ad Hoc and Sensor Networks, IEEE/ACM Trans. Networking, vol. 17, no. 1, pp. 200211, Feb. 2009.

[1]

[2]

[3]

[4]

[5]

[6]

You might also like