1. INTRODUCTION:

The dynamics of routing is being seen as an essential component in the clustering and employment of new protocols to handle the constantly altering topologies of wireless sensor networks. In awe of the rapidly changing paths through which data units of the mobile networks travel along with the limitation of resources put mobile ad-hoc networks in a compromising situation. Additionally, if a specific threshold is crossed by the size of the network concerned, then the efficiency of the protocols do not stand up to the mark. Battery consumption along with bandwidth limitations requires the appropriate management of the network. The routing problems present themselves in a variety, including scalability (inability to scale up according to huge size) issues and fault tolerance issues (inappropriate or no measures taken for debugging the network topology problems).

In such circumstances, one of the prime objectives is to reduce the overheads faced due to communication. With the help of appropriate cluster architecture, it is possible to solve these problems of routing. In the Figure below, one of the various pats is traced by the routing packet, ultimately reaching the user via the gateway sensor node. This paper revolves around technical concepts including k-means clustering and employment of AODV protocol (Ad hoc On Demand Distance Vector Routing) and NS2, an open-source network simulation software and other concepts as well.

Fig. 1.1: A figure demonstrating the routing mechanism.

Let us try to bring to attention about the specific details of the technical jargon, as we are to deal with in the future sections:

· K-means clustering, in which a set of n items are organized into a set of k-clusters, thus giving an easy and efficient management of the cluster.

· The AODV (Ad-hoc On Demand Distance Vector routing) protocol, which was designed specifically for mobile ad-hoc networks, also used for performance analysis.

· Ns2, which is a widely known open-source networks simulation software.

Working of AODV Protocol:

The prime task of the AODV protocol is to discover the feasible routes a packet can take and choosing the optimal one via a set of control messages; an expanding ring search technique is employed by the node sending out the packets for transmitting to the destination nodes. The control messages are the “route request” and “route reply” messages. In order for routing to take place, a routing table is set up which is always updated with the information.

2.EXISTING METHODS:

We comprehend from the previous works based on the use of AODV routing protocols about how to solve the variety of problems, which take place in routing mechanisms. In the paper-An AODV based clustering approach for efficient routing in MANET, a clustering architecture is proposed which helps solving the scaling of network problem as well as reducing the overhead caused by communication, the reason being AODV, though having great performance with mobile nodes, involves a great deal of overheads. This paper tends to solve these problems by inculcating and constructing a cluster architecture to refine and maintain the data. It also helps in easy discovery of routes for efficient packet transfer. In another journal where the implementation of AODV routing protocol is made in network simulator (ns2) for multi-hop wireless networks. In this paper, the problems for throughput and packet-loss ratio reduction are solved. Here, there are two major components- the source, which produces the packets and the sink for accepting the packets at the destination. In the Performance evaluation of AODV, LEACH and TORA protocols through simulation, AODV is considered to be the base protocol as it is actually a functionally enhanced version of the DSDV protocol. The routes prepared by the AODV are as per the demands. Be it multicast routing tables or unicast routing tables, the links used between nodes are symmetric. Abbasi et al. ^[1] in his research paper that was related to the algorithms required for clustering has made classification for clustering and has discussed the taxonomy related to it. Then compressed distinctive grouping calculations for WSNs in light of characterization of variable merging time conventions and consistent union time calculations, and has underlined their various functionalities, goal and their intricate structures. At last, these algorithms have been compared on the basis of stability of the cluster, rate at which they converge, node interaction within a cluster, overlapping of the cluster and the awareness of the location of the cluster. Arboleda et al^.[2] in his paper has displayed a correlation review between various protocols that were related to clustering. The creators of the overview talked about some fundamental ideas identified with the process of clustering, for example, structure of cluster, different types of cluster, cluster points of interest, and further in his work has discussed about protocol that was based on LEACH and other protocols used in WSN, which are categorized as reactive and proactive protocol. The fundamental attributes of these conventions were thought about and the proofs where they can be utilized at present were sketched out. Kumarawadu et al^.[3] overviewed the algorithms for clustering that are most prominent in WSNs and categorized them on the basis of parameters that are related to formation of the cluster and guidelines required for electing the head of the cluster. The creators of the overview additionally considered the main outline challenges and talked about the execution issues related to protocols of clustering in light of the arrangement of character based on identification of clustering, vicinity data based and naturally enlivened clustering algorithms. Jiang et al. ^[4] examined a sum of three noticeable favorable circumstances of methods for clustering required in WSNs, for example, more versatility, less overheads, and simple support, and after that present an arrangement of WSN clustering plans in light of a sum of eight clustering traits. The creators likewise examined inside and out six famous WSN algorithms of clustering, for example, LEACH, EEUC, PEGASIS and so forth. Then he has analyzed these WSN algorithms based on different properties.

3. PROPOSED WORK:

3.1. Problem Statement:

Wireless sensor systems are battery worked. Sensor nodes gather the information and pass them on to the system for further utilize. This passing and getting of information uses a large portion of the vitality of the system. Therefore, for better operation and expansion the lifetime of the system, vitality utilization must be the central point of concern. In this anticipate, bunch proficiency is kept an eye on AODV protocol. Firstly, performance of nodes in AODV protocol has been examined and after that through the continuous sensor information, it has been outwardly seen that CLUSTERING utilizing k-means is viable. When this strategy was proposed couple of objectives were set, as takes after

1. Vitality dispersal of the system should be minimized.

2. Life to the system should be increased.

3. Reliability of the clusters should be properly adjusted.

4. Within the system, the head of the cluster should be equally distributed.

3.2 System Design:

The given class diagram shows that collection of large number of nodes and few sinks as a whole make a network. There can be connection between or more networks. These nodes might be attached to other nodes and all are associated to the sinks. A sink can be altered or versatile (e.g. cell phone). Nodes can be Sensor hubs (general data collection hubs) or Super nodes (that gather information from different hubs). Nodes can be composed into groups, and every bunch has a Cluster Head that controls the information gathering inside the group. Generally, three variant of routing protocols is used by MANETS. There are protocols, which are transmit data only when there is a need of data. They are called reactive protocols and themselves decide their path for routing, some of them are Ad hoc On-interest Distance Vector (AODV), Dynamic Source Routing (DSR) and Temporally Ordered Routing Algorithm (TORA). The proactive conventions, for example, Destination Sequenced Distance Vector (DSDV), Optimized Link State Routing (OLSR) and Fisheye State Routing (FSR) manages that steering tables be kept up at every hub.[7] Crossbreed directing conventions, for example, Zone Routing Protocol (ZRP) are likewise utilized, which incorporates the qualities of proactive and receptive conventions, additionally has faults i.e. cannot be assessed for unidirectional connections and it can be connected just for expansive systems. AODV convention supports the slightest congested course rather than the most limited course and it additionally bolsters both unicast and multicast bundle transmissions notwithstanding for hubs in consistent development. It likewise reacts rapidly to the topological changes that influence the dynamic courses. AODV does not put any extra overhead on information bundles, as it does not make utilization of source directing. Though, DSR convention is not adaptable to extensive systems and even requires fundamentally all the more handling assets. Essentially, so as to get the steering data, every hub must invest parcel of energy to handle any control bundle it gets, regardless of the possibility that it is not the expected beneficiary. Indeed, even DSDV acquaints a lot of overhead with the system because of the prerequisite of the occasional upgrade messages.”

3.3 k-Means clustering:

K-means is the most frequently used clustering technique, which ordinarily based upon the partitioning method .This tries to discover a client determined number of exact cluster (k), which has their own centroids. ^[5]The function that has been created for data of low dimensions like square error may not be suitable for data of high dimension and thus will not be able to give same yield and there will be variations in the results because of the presence of various anomalies. There are two basic ways to deal with the computation of the center of the cluster i.e. either to choose the underlying qualities arbitrarily, or to pick the primary k tests of the information focuses. As an option, diverse arrangements of starting qualities are picked (out of the information focuses) and the set, which is nearest to ideal, is picked. Sensor network comprises of large number of sensors, which collect huge amount of data. Now this information aggregation scheme helps to eradicate repeated data and thus helps in reducing the load of data within a network. The computational multifaceted nature of unique K-implies calculation is high, particularly for substantial information sets. Additionally, the quantity of separation estimations increments exponentially with the expansion of the dimensionality of the information.

3.3.1 PROPOSED ALGORITHM:

K-implies calculation is constructed essentially with respect to the Euclidian separations and head of the cluster choice is proportional to the node's energy. Therefore, it is the responsibility of the head to gathers the data about the id of the other nodes, position of the other nodes with respect to the head of the cluster as well their energy and all this data needs to be stored in a form of a list in the head. Once the cluster receives this data, head then it start working upon it using k-means algorithm.

3.3.2 PROCEDURE FOR THE ALGORITHM:

1. Within a network which has been created using 20 nodes can be clustered into k groups or clusters and this k will represent the total centroid number, which might be placed at arbitrary places. ^[6]

2. Euclidian distance will be measured from the centroid to each of the node and the node that is found to be the nearest to the centroid will be assigned to that particular cluster and that is how k cluster will be formed for 20 nodes.

3. Once these k cluster are formed, the centroid position will be calculated again so as to test whether the position of the centroid has changed or not from the past one.

4. Step 2 will be followed if the centroid position is found to be changed otherwise the k clusters will be declared as fixed for the network and this will end the process of clustering.

4. METHODOLOGY:

4.1 Working:

In this project, firstly, the network topology containing 20 nodes is created to check how the network looks like which has AODV protocol and how this protocol works has been simulated using NS2.Then the performance is checked using the real time data set, which is based on detection of intrusion in wireless network. Then the clustering is performed on the bases of packet delivered which is considered as a cluster parameter.

4.1.2. Fact Finding Setup:

Network simulator is used to for this purpose and simulation is carried out in 1000m x 1000m area within 20 nodes.

TABLE 4.1: PARAMETER FOR SIMULATION

PARAMETER	VALUE
ROUTING PROTOCOL	AODV
NUMBER OF NODES	20
PACKET SIZE	1500
TRAFFIC MODEL	FTP

4.1.3 INFORMATION FOR DATA SET USED:

TABLE 4.2: DATA SET PARAMETER

PARAMETER	VALUE
Protocol for Routing	AODV
MAC/Physical Layer Standard	IEEE 802.15.4
Network for Simulation	Wireless Sensor Network
Number of Nodes used	20
Estimated Time for simulation	4 hours
IP addresses	10.0.0.1 - 10.0.0.19
Base Station IP address	10.0.0.1

5. RESULTS AND DISCUSSION:

5.1 Simulation Output:

Fig 5.1 topology creation

This fig is displaying a network topology of 20 nodes in which n0, n2, and n8 are sending data to node n7.tcp0, tcp1, tcp2 are connected to the sender node and a sink is attached to destination.

Fig 5.2 transmission range of few nodes

The above fig 5.2 demonstrate the packet transfer between nodes so as to reach the final destination as we know in AODV protocol, the route is determined while transferring the packets to reach the destination

5.2 Performance evaluation of AODV

This figure show the packet size and within a packet range the min and max which will be helpful in discussing the cluster

5.3 CLUSTERING RESULTS:

This section will depicts the clustering of the nodes, preprocess graphs is shown and then after applying the k-means clustering algorithm, the result has been displayed .the dataset has been extracted from Wire shark tool and is converted to csv format

6. CONCLUSION:

Remote Sensor Networks, which might be spread over immeasurable geological range, are finding applications in numerous ranges. In this connection, there is a need of methodologies, which can deal with these WSNs in way that is more efficient. In such manner, this work, introduced requirement for clustering to conquer a few constraints of WSNs. Elaborated explanation of the work has been given in this thesis. Importance of clustering, the protocol that has been used for the purpose has been briefly explained .K-means clustering that is used for clustering the real time data is demonstrated with weka software. Before that, using NS2 software the creation of wireless network topology, which comprises of 20 nodes, is shown using tcl .then taking the real time data set in which AODV protocol was used and has been analyzed through the graphs in Wire shark tool and then clustering is applied. So through this work it can be concluded that if use clustering in wireless network with AODV protocol, network efficiency will improve.

7. FUTURE WORK:

In this paper, clustering has been done with AODV protocol, lot more researches has been done with other protocols as well. For increasing the performance of sensor and increasing their lifetime ,advanced form of protocols and new clustering algorithms can be applied on it so that it can data security can be handled and large area can be covered.

8. REFERENCES:

1. Abbasi, A.A.; Younis, M. A survey on clustering algorithms for wireless sensor networks. Computer. Communication. 2007, 30, 2826–2841.

2. Arboleda, L.M. C.; Nasser, N. Comparison of Clustering Algorithms and Protocols for Wireless Sensor Networks. In Proceedings of IEEE CCECE/CCGEI, Ottawa, ON, Canada, May 2006; pp. 1787–1792.

3. Kumarawadu, P.; Dechene, D.J.; Luccini, M.; Sauer, A. Algorithms for Node Clustering in Wireless Sensor Networks: A Survey. In Proceedings of 4th International Conference on Information and Automation for Sustainability, Colombo, Sri Lanka, December 2008; pp. 295–300.

4. Jiang, C.; Yuan, D.; Zhao, Y. Towards Clustering Algorithms in Wireless Sensor Networks— A Survey. In Proceedings of IEEE Wireless Communications and Networking Conference, udapest, Hungary, April 2009; pp. 1–6

5. Arthur Ndlovu. Improved Energy Efficient AODV Routing using K-means Algorithm for Cluster Head Selection. In Proceedings of International Journal of Computer Science and Mobile Computing, Vol.4 Issue.8, August- 2015, pg. 177-187

6. P. Sasikumar, Sibaram Khara,k-MEANS Clustering in Wireless Sensor Networks. Fourth International Conference on Computational Intelligence and Communication Networks, 2013

7. G. Vijaya Kumar, Y. Vasudeva Reddyr and Dr. M. Nagendra, “Current Research Work on Routing Protocols for MANET: A Literature Survey”, (IJCSE) International Journal on Computer Science and Engineering 2010.

Received on 26.09.2016 Modified on 10.11.2016

Research J. Pharm. and Tech. 2017; 10(1): 73-82.

DOI: 10.5958/0974-360X.2017.00019.1