Data Aggregation Techniques to Remove Redundancy in Wireless Sensor Networks: Brief Overview. Samarth Anavatti, Sumedha Sirsikar Research Scholar, Department of Information Technology Assistant Professor Department of Information Technology Maharashtra Institute of Technology, Pune University, Pune 411038 anavattisamarth@gmail.com, sumedha.sirsikar@mitpune.edu.in A B S T R A C T In Wireless Sensor Networks (WSN), some sensor nodes are mobile in nature. Due to mobility of nodes, there is no guarantee of reliable delivery of information. To ensure reliability, many sensor nodes are deployed in the monitoring environment. These sensor nodes sense the same kind of data and forward it to the sink node. This redundant information sustains the reliability; but at the same time, sink node wastes its energy in processing the redundant data. So there is a need to eliminate the redundancy in sensed data up to adequate level in order to maintain the tradeoff between energy conservation and reliability. There exist many data aggregation techniques that perform data redundancy removal in order to improve life time of sensor nodes. Data aggregation is a technique in which each intermediate node in the routing path receives multiple input packets, process them and transmits a single packet. In this paper we have studied different data aggregation strategies and focused on some data aggregation techniques based on these strategies. Further we havediscussed advantages and limitations of these techniques. Index Term- Wireless Sensor Network, Data Aggregation, Energy conservation, Traffic load, Reliability I. INTRODUCTION Wireless Sensor Network (WSN) is collection of large number of tiny self-powered sensor nodes connected to each other wirelessly within their radio range. These nodes sense the data and pass it to the base station using hop by hop communication. As these sensor nodes have less power it is necessary to reduce energy consumption at sensor node to enhance overall life time of WSN. We can achieve this by removing redundancy from WSN. Because nodes waste their power in processing redundant data. Therefore removing redundancy will become solution for improving WSN life time. Data Aggregation is the technique which gather and aggregate data in an energy efficient manner and reduces redundancy so that network lifetime is enhanced. This paper presents study of various data aggregation techniques which includes, Dynamically Balanced Spanning Tree (DBST) [2] which provides dynamic structure of tree to solve hotspot problem and improve energy conservation. SVM based Data Redundancy Elimination (SDRE) [3] algorithm makes use of SVM for redundancy elimination. Bandwidth efficient Heterogeneity aware Cluster based Data Aggregation (BHCDA) [4] which performs both intra cluster and inter cluster aggregation to eliminate redundancy. Redundancy Eliminated Data Dissemination (REDD) [5] algorithm makes use of context aware 10 2014, IJAFRC All Rights Reserved www.ijafrc.org
system for validation and correlation coefficient is used to eliminate redundancy from valid data. Adaptive Energy Efficient Reliable Data Aggregation Technique (AEERDAT)[6]this algorithm changes the size of cluster and allows only valid nodes to be in the cluster in order to maintain reliability. The rest of the paper is organized as follows. The section 2 gives related work. Section 3 contains overview of data aggregation. Section 4 focuses on Data aggregation techniques. Section 5 shows advantages and limitations of data aggregation techniques and section 6 contains conclusion followed by references. II. RELATED WORKS Nandini. S. Patil, Prof. P. R. Patil [1] performed comparison of performance of WSN with data aggregation and performance without data aggregation. Avid Avokh, GhasemMirjalilyPatil [2] proposed an approach called Dynamic Balanced Spanning Tree Approach which is improvement over fixed spanning tree approach. In fixed spanning tree there was problem of hotspots which is eliminated here.this work results in minimum energy consumption and also it balances the traffic load. Prakash goudpatil, Umakant Kulkarni [3] developed a new technique which makes use of Support Vector Machine (SVM) to eliminate redundant data and then used LSH algorithm to eliminate outliers who may send false data. In this way, this technique not only reduces energy consumption but also eliminate false data. DnyaneshwarMantri, NeeliRashmi Prasad, Ramjee Prasad [4]developed cluster based approach which works efficiently in heterogeneous environment. This approach reduces redundant transmission by performing intra cluster and inter cluster data aggregation which utilizes bandwidth efficiently and reduces energy consumption. Sumalatha Ramachandran, Aswin Kumar Gopi, GiridaraVarmaElumalai, MurugesanChellapa[5]proposed a novice cluster based approach which uses context aware system to validate data and then eliminate redundancy of validated data. This redundancy is removed using correlation coefficient technique. Basavaraj S. Mathapati, Siddarama. R. Patil [6] have developed a technique which performs reliable and energy efficient data aggregation and they compared its performance erdc [12]. III. DATA AGGREGATION The deployment of large number of sensors over sensing area increases the data accuracy. The sensors deployed in the nearby region sense the same phenomena which leads to produce lot of duplicate data. This duplication of data causes redundancy and leads to more bandwidth and energy consumption. Data De- duplication means removing duplicates to reduce redundancy. The data aggregation techniques are used to perform De-duplication [9]. In data aggregation sensor data is collected by sensor nodes are aggregated by using some data aggregation algorithms and then aggregated data is forwarded towards base station. There are different strategies for data aggregation. A. Data Aggregation Strategies There are many strategies for aggregation some of them are listed below: Centralized Approach, In- Network Aggregation, Tree-Based Approach and cluster-based Approach 1. Centralized Approach: It is an address centric approach. In this each node sends data to a central node via the shortest possible route using a multichip wireless protocol. The central node also called as header node aggregates the data which can be queried, to reduce the redundancy. 2. In-Network Aggregation: The aggregation is the global process of gathering and routing 11 2014, IJAFRC All Rights Reserved www.ijafrc.org
information through a multi-hop network and processing data at intermediate nodes with the objective of reducing power consumption. There are two approaches for in-network aggregation: 3. With size reduction: It refers to the process of combining & compressing the data packets received by anode from its neighbors in order to reduce the packet length that is to be transmitted or forwarded towards sink. 4. Without size reduction: In-network aggregation without size reduction refers to the process merging data packets received from different neighbors into a single data packet but without processing the value of data. 5. Tree-Based Approach: In the tree-based approach aggregation is done by constructing an aggregation tree, which could be a minimum spanning tree, rooted at sink and source nodes are considered as leaves. Each node has a parent node to forward its data. Flow of data starts from leaves nodes up to the sink and the aggregation done by parent nodes. 6. Cluster-Based Approach: In cluster-based approach, whole network is divided in to several clusters. Each cluster has a cluster-head which is selected among cluster members. Cluster-heads do the role of aggregator which aggregate data received from cluster members locally and then transmit the result to sink. [1] IV. DATA AGGREGATION TECHNIQUES Based on the nature of the network, data aggregation can be done via data aggregation tree (DAT for flat networks) or by a clustering strategy for hierarchical networks. In flat network architecture, all nodes are equal and connections are set up between nodes that are within each other s radio range although constrained by connectivity conditions and available resources. In a hierarchical network, all nodes typically function both as switches/routers, with one node in each cluster being designated as the cluster head (CH).The number of tiers within a hierarchical network can vary according to the number of sensor nodes. Traffic between nodes of different clusters must always be routed through their respective CHs or via gateway nodes. Gateway nodes are responsible for maintaining connectivity among neighboring CHs. Although the hierarchical network architecture is energy efficient for collecting and aggregating data from the entire WSN or all nodes within a larger target region, using knowledge of their relative locations, flat network architecture is suitable for transferring data between source-destination pairs separated by a large number of hops. Following figure shows taxonomy of data aggregation techniques we considered in our study. 12 2014, IJAFRC All Rights Reserved www.ijafrc.org
A. Dynamic balanced spanning tree (DBST) Figure 1: Data aggregation techniques This algorithm considers distance, residual energy and node weight as parameters to improve lifetime of network. In previous researches, a fixed routing tree is used for all rounds. A round is time needed for collection of one data unit from every node in the network and delivering the aggregated data to sink. But as nodes are fixed, the hotspot problem occurs. Hotspot drains the battery quickly. DBST [2] solves this hotspot problem. This algorithm not only minimizes the maximum energy consumption among the sensor nodes but also balance the traffic load. DBST is a tree based approach, in this smallest possible weight spanning tree can be formed by using kruskals algorithm [7]. DBST uses residual energy as a parameter for the root node selection in tree formation process. The node with the highest residual energy after each round will be selected as a root. Due to this, responsibility of root is fairly get distributed among all nodes and problem of hotspot get solved. Though DBST solves hotspot problem there is need to form spanning tree for each round. To form spanning tree they considered node weight and link weight as parameters. Now, to find node weight they considered energy required for communication, residual energy and heterogeneity of network as main criterions. Link weight is determined by using node weight of all nodes as: F i,j= F j,i=wiwj, ( ) i,j=1,2 n After tree formation and root node selection DBST performs data gathering and aggregation. Aggregated data is sent towards root androot forwards it to sink (Destination). DBST has a dynamic routing tree which minimize the energy consumption and also balance the traffic load. So bandwidth overhead is less. But this algorithm needs to create new tree for each round resulting, little bit increase in delay. B. SVM based Data Redundancy Elimination (SDRE) This algorithm considers an input tree which has one parent node and one gateway node. All parent nodes are connected to gateway node. Whenever a sensor node receives data packet first time, the node from which it gets the packet will acts as parent node. If the same node receives same packet from some other node that node will acts as backup parent node and all the similar messages are ignored. In this way Data Aggregation Tree (DAT) is formed [8]. The Support Vector Machine (SVM) is applied on DAT to eliminate redundancy. SVM Performs two functions one is classification and other is correlated data elimination. It uses linear classifier method to represent the classification of redundant data. This method divides the hyper plane in two classes, redundant (no) and not redundant (yes). 13 2014, IJAFRC All Rights Reserved www.ijafrc.org
To check the similarity Locality Sensitive Hashing (LSH) is used. This method generates the hash code which is small in size than data. These LSH codes are sent to supervisor node. Aggregation supervisor maintains redundancy count for similar LSH codes. The similarity count below or equal threshold is accepted and those who have more than threshold are rejected. In this way supervisor node eliminates outliers [3]. This protocol keeps the redundancy at adequate level so as to provide reliability and reduces the average energy consumption. C. Bandwidth efficient heterogeneity aware cluster based Data Aggregation (BHCDA) BHCDA [4] reduces number of transmissions of data packets from sensor nodes to mobile sink. The network model of BHCDA is a connecting graph of different clusters of Wireless Sensor Network (WSN).Each cluster is having heterogeneous nodes and mobile sink. Some nodes from the cluster have high energy are called as Super nodes, some have moderate energy are called as advanced nodes and some are with normal energy are normal nodes. Cluster head (CH) is selected from each cluster. Data Aggregation is done at cluster head (CH) for the packets inside the cluster. These cluster heads (CH) will further acts as a nodes and one of the cluster head (CH) from these CH s, will become aggregator node. This aggregator node will forward aggregated data to the sink. The aggregation function used in BHCDA is based on correlation of data packets generated by high energy nodes and low energy nodes. The aggregation function is as follows: F(C a)= (Xi)+ (Yj) Where Xi and Yj are variables that represent the correlation of the number of data packets generated by the u and h. BHCDA performs both intra cluster and inter cluster aggregation which results not only in improving life time of sensor network but also in improving bandwidth utilization. D. Redundancy Eliminated Data Dissemination (REDD) In this approach total geographical area is divided into grid based clusters. In each cluster one representative node called header node is elected. This header node is elected based upon battery power. The nodes of WSN are moving so there might be a chance that node may go in other cluster. This is handled by dynamic topology management module of REDD [5]. Whenever sink node queries for data of interest from source nodes, that query is forwarded by header to header forwarding upto source [9]. This forwarding is done through shortest path which is found by sink. REDD uses context aware system for validation. Context aware system checks the sensor data against the rules of its rule engine. If provided context satisfies the rules then the data is transmitted else it is rejected. Then redundancy only from valid data is removed in two ways: intra-cluster redundancy removal and inter cluster redundancy removal. To eliminate redundancy correlation coefficient is used. 14 2014, IJAFRC All Rights Reserved www.ijafrc.org
International Journal of Advance Foundation and Research in Computer (IJAFRC) Figure 2: Intra cluster redundancy elimination Figure 3: Inter cluster redundancy removals REDD optimizes data transmission therefore bandwidth overhead is less and reduces amount of power consumption in order to improve network lifetime. E. Adaptive Energy Efficient Reliable Data Aggregation Technique (AEERDAT) AEERDAT [6] addresses reliability issue of Wireless Sensor Network (WSN). One disadvantage of data aggregation is that there might be a chance of malicious attack on node whichh performs aggregation called aggregator node. EERDAT forms a cluster and for each cluster coordinator node (CN) is randomly selected. This CN monitors the function of clusters. Each node inside the cluster maintains Neighbor Information Table (NIT). Each node sends this NIT information to CH. Coordinator Node (CN) selects CH The CH is selected based up on cost value. Highest cost value node in the cluster is selected as CH. Cost value is calculated based on residual energy and distance of a node from CN. Minimum distance and maximum residual energy node has highest cost value. The CH performs aggregation and sends this aggregated data to CN. The CN calculates Loss Ratio (LR) which is amount of data packets forwarded to the data packets received. This value of LR is forwarded to CH. Based upon this, the size of cluster is changed that is the number of forwarding nodes added or removed. As the size of cluster is changing adaptively outliers are get removed. 15 2014, IJAFRC All Rights Reserved www.ijafrc.org
Reliability is increased due to altering the cluster size before data transmission and also LR is measured at CN itself therefore, energy consumption is effectively reduced. F. Energy Efficient and Balanced Cluster-Based Data Aggregation Algorithm for Wireless Sensor Networks (EEBCDA) This protocol addresses the problem of cluster-based and homogeneous WSNs in which cluster heads transmit data to base station by one-hop communication. The operation of EEBCDA [10] is also divided into rounds and every round consists of a set-up phase and a steady-state phase, especially, there is a network-division phase before the first round. The network is divided into rectangular regions firstly, called swim lanes, then, each swim lane is further partitioned into smaller rectangular regions, called grids. The node with the maximal residual energy of each grid is selected as CH. The grids further away from BS are bigger and have more nodes to participate in CHs rotation. It divides the network into rectangular grids with unequal size and makes cluster heads rotate among the nodes in each grid respectively, the grid whose cluster head consumes more energy has more sensor nodes to take part in cluster head rotation and share energy load, by this way, it is able to balance energy dissipation. Besides, it adopts some measures to save energy. EEBCDA remarkably enhance energy efficiency, balance energy dissipation and prolong the network lifetime. G. Delay Efficient Distributed Data Aggregation (DEDA) This scheduling algorithm is proposed to handle the delay and energy tradeoff in the process of aggregation using the timeout concept. This approach first builds aggregation tree, and then distributed aggregation scheduling algorithm is applied to achieve the optimized energy efficiency and delay aware data aggregation. This algorithm makes use of Decision Making Unit (DMU) to handle energy and delay tradeoff. One of the advantage of DEAD is to achieve the ideal energy consumption by limiting a number of redundant and unnecessary responses from the sensor nodes. V. ADVANTAGES & LIMITATION OF DATA AGGREGATION TECHNIQUES Following table shows the comparison between the aggregations techniques discussed earlier. All techniques are compared based on strategy which they used either tree based or cluster based, delay that is time required for data aggregation, redundancy, accuracy, energy consumption and bandwidth overhead. Table I: Advantages and Limitations Of Data Aggregation Techniques Techniques Advantages Limitations DBST Balances traffic load, provides energy conservation Need to form dynamic minimum spanning tree for each round SDRE Handles tradeoff between energy Traffic load on supervisor node conservation and accuracy BHCDA Provides more bandwidth efficiency Energy saving is not that much up to the mark REDD Increases life time and reliability of WSN Algorithm may produce delay if the cluster head itself moves into another grid, again 16 2014, IJAFRC All Rights Reserved www.ijafrc.org
AEERDAT International Journal of Advance Foundation and Research in Computer (IJAFRC) Provides energy efficiency and reliability by Adaptively changing the size of cluster DEDA Handles energy and delay tradeoff EEBCDA header election algorithm needs to be executed Delay is increases if the number of nodes are increased Does not remove redundancy completely Balance energy dissipation To enhance Used for homogenous network only energy efficiency VI. CONCLUSION Data aggregation techniques presented in this paper focuses on to remove the redundancy in order to enhance the life time of wireless sensor network.to improve network life time it is necessary to eliminate redundancy but at the same time redundancy is necessary to maintain accuracy. This paper presents survey of data aggregation techniques which are helpful in balancing the tradeoff between energy efficiency and accuracy. SDRE is the protocol which is best in maintaining tradeoff between energy efficiency and accuracy. VII. REFERENCES [1] Nandini. S. Patil, Prof. P. R. Patil, Data Aggregation in Wireless Sensor Network, IEEE International Conference on Computational Intelligence and Computing Research, ISBN: 97881 8371 3627 [2] Avid Avokh, GhasemMirjalilyPatil, Dynamic Balanced Spanning Tree (DBST) for Data Aggregation in Wireless Sensor Networks, 5th International Symposium on Telecommunications (IST2010), 978-1-4244-8185-9/10 2010 IEEE [3] PrakashgoudPatil, Umakant Kulkarni, SVM based Data Redundancy Elimination for Data Aggregation in Wireless Sensor Networks, 978-1-4673-6217-7/13 2013 IEEE [4] DnyaneshwarMantri, NeeliRashmi Prasad, RamjeePrasad, BHCDA:Bandwidth Efficient Heterogeneity aware Cluster based Data Aggregation for Wireless Sensor Network, ICRTIT 2011, 978-1-4673-6217-7/13 2013 IEEE [5] Sumalatha Ramachandran, Aswin Kumar Gopi, GiridaraVarmaElumalai,Murugesan Chellapa, REDD: Redundancy Eliminated Data Dissemination in Cluster Based Mobile Sinks, ICRTIT 2011, 978-1-4577-0590-8/11 2011 IEEE [6] Basavaraj S. Mathapati, Siddarama. R. Patil, Energy Efficient Reliable Data Aggregation Technique for Wireless Sensor Networks, International Conference on Computing Sciences, 978-0-7695-4817-3/12 2012 IEEE [7] A. Gagarin, S. Hussain, and L.T. Yang, Distributed hierarchical search for balanced energy consumption routing spanning trees in Wireless Sensor Networks, J. Parallel Distrib. Comput. Vol. 70, no. 9, pp. 975-982, 2010 [8] PrakashgoudPatil, UmakantKulkarni Delay Efficient Data Aggregation Algorithm in Wireless Sensor Network, International Journal of Computer Application (0975-8887) vol.69 No.1,May 17 2014, IJAFRC All Rights Reserved www.ijafrc.org
2013 [9] A.C. Viana, A.Ziviani and R.Friedman, Decoupling Data Dissemination from Mobile Sink s Trajectory in Wireless Sensor Networks, IEEECommunication Letters, vol. 13, no. 3, Mar. 2009 [10] Yuea Jun Weiming Zhang, Weidong Xiao, Daquan Tang, and Jiuyang Tang, Energy efficient and balanced cluster-based dataaggregation algorithm for wireless sensor networks, ProcediaEngineering 29 (2012): 2009-2015 [11] Ren P. Liu, John Zic, Iain B. Collings, Alex Y. Dong, and Sanjay Jha, Efficient Reliable Data Collection in Wireless Sensor Networks, inproceedings of IEEE 68th Vehicular Technology Conference, VTC2008, 2008 18 2014, IJAFRC All Rights Reserved www.ijafrc.org