An optimized stable clustering algorithm for mobile ad hoc networks

A mobile ad hoc network (MANET) is an autonomous infrastructure less network in which collection of mobile nodes (i.e., mobile, sensor, palmtop, laptop) dynamically communicates with each other through wireless medium within their own transmission range (using single hop or multiple hops) via intermediate nodes [1]. The infrastructure less nature of mobile ad hoc network causes frequent change in the topology of network due to dynamic mobility of mobile nodes, communication management, and creation of a stable network are the most challenging tasks in MANET. Clustering is a possible solution to address these existing challenges. With the help of clustering, nodes are organized into different groups that make the network more robust, durable, and scalable [2].

The mobile ad hoc network (MANET) can be deployed without requirement of any further extra cost and time. In a MANET, every mobile node plays a role of router along with its job as an ordinary host. MANET still has some challenges like limited bandwidth, limited battery power for each mobile node, and frequent topology changes because of node movement [3, 4]. To list such complex and dynamic environment challenges of mobile ad hoc network, various clustering algorithms have been proposed in the literate [5]. Several routing protocols have been proposed in the literature to handle the one hop and multihops, self-organizing network based on proactive, reactive, and hybrid protocols [1–9]. Routing protocols are categories in three categories which are Proactive, Reactive Active and Hybrid Protocol. Proactive protocol remains activate all the time in network even there is no data to transmit and keeps route information available all the time from source to destination.

Whereas, in reactive routing protocols, route information is available on request. This reduced power consumption in reactive protocols. Hybrid protocol works on the principal of proactive and reactive protocol.

For the large networks, flat routing structure requires excessive information. In order to overcome such problem, hierarchical structure (clustering) plays an important role in MANET reducing network overhead, increasing reusability of bandwidth, providing stability to the cluster structure, reducing battery power consumption of mobile node, and reducing clustering as well as in intracluster communication [10].

Clustering is about segregating a collection of mobile nodes into different virtual logical groups in a MANET. Each cluster is capable to connect with other cluster using cluster gateway to provide connectivity for a network. Each cluster consists of various mobile nodes such as cluster head, cluster member, and cluster gateway, and they perform different roles in cluster at the time of data communication in mobile ad hoc network.

Each cluster head is a special mobile node in a cluster which acts as local coordinator within the cluster. In a cluster, only one mobile node acts as cluster head at a time. Main responsibilities of cluster head include data forwarding from source node to destination mobile node, intracluster transmission, and managing all member nodes of a cluster. Cluster members in a cluster are treated as normal mobile nodes which can communicate with each other through cluster head. A cluster gateway is cluster member which is used to connect two or more clusters so that each cluster can access its neighbor cluster to send or receive data from other neighboring cluster.

It is evident that MANET is dynamic in nature; due to such nature, performance of network decreases as the size of network increases. Such problem can be reduced by introducing clustering, for such networks. Clustering increases scalability of wireless network and decreases network overhead [11]. Clustering also provides spatial reuse of resource to increase the network capacity. The same frequency code can be used if two clusters are not neighbors (not within the same radio range).

In [12], centralized metaheuristic based on “tabu search” and a distributed heuristic based on “ant colony” are applied to reduce computation overhead in wireless sensor network (WSNs). Energy-Efficient, Delay-Aware, and Lifetime-Balancing (EDAL) take care to minimize the system lifetime for individual node. It reduces the amount of traffic generated in the network by compressive sensing. The present taxonomy is of opportunistic routing protocols for disruption tolerant networks (DTNs). In [13] opportunistic routing solutions, number of features which are used to classified DTNs according to mobility, capability, and connectivity of nodes. It also describes opportunistic routing protocols for DTN and basic opportunistic routing building blocks (Fig. 1).

Energy consumption for mobile node is a very important issue in MANET because it reduces network connectivity as soon as mobile node dies due to battery drainage. In [14] the best next hop, mobile node is selected based on energy efficiency to provide network connectivity. Selecting the best nodes will reduce the path change from source to destination and decreased network overhead. In [11], optimal clustering and throughput are archived in hierarchical cooperation in mobile ad hoc network. The exact throughput for various numbers of stages can also be achieved [11]. The optimal cluster size is responsible to minimize throughput for all stages. The hierarchical scheme does not handle linear scaling.

In [9], a weight-based clustering algorithm is described to manage mobile nodes and maintaining the local topology within the network. Each mobile node calculates its weight by using weight function and compares its weight with other mobile nodes which are neighbor nodes within two hops to create the cluster head. The node with the highest weight will be elected as the cluster head and the remaining mobile nodes will be considered as cluster members. In this paper, neighboring mobile node has higher priority to group into the same cluster, which reduces number of cluster, improved cluster stability and reduced clustering overhead. In [15] various cluster heads, selection schemes have been proposed for WSNs and MANET. The main objectives of such schemes are to elect efficient cluster head for a cluster so that dynamic topology changes may be reduced as cluster head moves from the cluster or dies because of battery drainages. Hence, a solution is required which selects a reliable (stable) cluster head.

In clustering algorithms [16], routing information is shared with cluster head and cluster gateway. This reduces the total number of transmission for routing information and efficiently manages routing table in a network. A cluster structure increases scalability of network and energy consumption [17] and decreases network overhead.

In [18], present a routing algorithm that handles mobility of nodes and clustered wireless network, by a capable gateway selection that is responsible for load balancing capabilities. Virtual hierarchies of clusters are used to explore the contextual proximity of nodes. The protocol also creates application of a kernel-based link quality estimator which agrees to pick the most appropriate gateways with load balancing and disconnection predication capability in each cluster. In [19], N-layer discrete power control scheme is design to improve larger transmission capacity and spatial reuse factor in a cluster with N-layers. In [20], problem related to non-uniform load distribution in mobile ad hoc network is studded and proposed light-weighted dynamic channel allocation and cooperative load balancing algorithm based on cluster. Load balancing and dynamic channel allocation method reduced clustering overhead and message passing. In [21], particle swarm optimization (PSO) is used to reduce network overhead and message drop while maintaining message delivery ratio.

Disadvantages of clustering: In [9, 23], a large and flat network is managed in MANET using clustering topology. Clustering required construction and maintenance cost in comparison to other topology.

We propose “an optimized stable clustering algorithm for mobile ad hoc networks (OSCA)” which is an extension of a novel weight-based clustering algorithm [28]. The main aim of the proposed protocol is to form a stable (durable) cluster for mobile ad hoc network. In proposed algorithm, cluster head changes are reduced to make cluster head more stable and sustain for longer period which reduced clustering overhead and makes network more durable for such network. Each mobile node contains node_info packets which contains information related to mobile node. node_info is broadcasted to all its neighbors during clustering.

Cluster head = 1, cluster member = 0, and backup node = 2.

Working of proposed an optimized stable clustering algorithm for mobile ad hoc networks (OSCA) is as follows below:

Each cluster contains four types of mobile nodes which are cluster head, cluster member, backup node, and cluster gateway. In the proposed algorithm, initial cluster head formation is carried out using weighted clustering algorithm (WCA) and cluster maintenance will be done by our proposed algorithms.

Proposed here is an optimized stable clustering algorithm for mobile ad hoc networks (OSCA) that can describe as follows:

If new cluster head priority is less than old cluster head priority and new cluster head priority is greater than backup node priority, then new cluster head will act as backup node and backup node will act as cluster member and old cluster head remains as cluster head for cluster and else new cluster head will act as cluster member within cluster.

The proposed OSCA for mobile ad hoc networks is simulated using network simulator (NS) [29]. Node movement generator is used to generate various node movements following by random way point model. Number of nodes, pause time, maximum speed, and field configuration and simulation time are given to movement generator as input parameter.

It is very hard to develop software which contains various networking components like routers, network topology, and various network algorithms. Network simulator saves money and time to complete such task [30]. Network Simulator-3 (ns-3) is used to simulate platform for networking and educational purpose. It provides various models for how packet data network works. It comprises with set of inbuilt libraries which may combine together with some other external software.

Proposed OSCA algorithm is compared with existing WCA and ICMS in terms of number of cluster head changes, number of cluster member changes, and clustering overhead for proposed OSCA scheme.

Proposed OSCA is simulated for N number of nodes on a simulation area 500×500 m. All mobile nodes are free to move in all direction during simulation.

Figure 3 shows the performance of proposed OSCA in terms of number of cluster head changes as simulation time variation. As simulation time increases number of cluster head changes are reduced. From the Fig. 3, it is observed that our proposed OSCA is minimizing cluster head changes as compared with ICSM and WCA.

As mobile node moves faster within the cluster, possibility of changing in cluster head, cluster member, and backup will be higher. As mobile node moves from one cluster to another cluster, such condition increased reformation of cluster head or backup node for a cluster to make cluster functional. In Fig. 4, performance of proposed OSCA is shown; from the result, it is apparent that cluster head changes depend on the movement of mobile nodes and its movement speed. From the figure, it is clear that cluster head change depends on pause time for mobile node. At initial of pause time, cluster head changes are minimal. But as soon as pause time increased, possibility of cluster head changes also increased. This makes unstable network. Comparative study shows that our proposed algorithm performs better in terms of less cluster head changes as pause time increased.

Figure 5 shows the number of cluster member changes against simulation time. In Fig. 5, as simulation time increases, the total number of cluster member changes decreases. From the simulation result, it is apparent that our proposed OSCA algorithm performs better by reducing number of cluster member changes for a cluster against simulation time. As soon as simulation time increases, cluster member changes are decreased and performance of OSCA is better against ICMS and WCA.

From Fig. 6, it is apparent that proposed OSCA continuously decreases cluster member changes as pause time increases. Initially, in between 5 to 10 pause time, cluster member changes are decreased in ICMS, WCA, and our proposed OSCA. It is clear from the Fig. 6 that when pause time increases from 10 to 20 m/s, the rate of cluster member changes is increased in ICMS and WCA but in proposed OSCA, it continues to be in decreasing stage.

Clustering overhead is very important issue in mobile ad hoc network. The process of exchange of messages for selecting cluster head (backup node) from a cluster increases the clustering overhead. To reduce such clustering overhead, a new algorithm is proposed which minimizes clustering overhead, by making cluster head a more stable cluster. In Fig. 7, in OSCA, clustering overhead is continuously decreased but in ICMS and WCA, clustering overhead changes are very low.

Figure 8, for the proposed algorithm, shows the performance of clustering overhead over mobile node speed. As the result indicates that initially, for proposed algorithm, clustering overhead is higher but as mobile node speed increases, cluster overhead starts decreasing, as compared with ICMS and WCA for MANETs (Table 2).

Algorithms name	Method used for election of cluster head	Node mobility during cluster setup	Strength	Weaknesses
LID [20]	The node having the lowest IDs among neighbors elected as cluster head	Yes	Faster cluster head election process, simple method, unwanted message exchanges are not required.	If a mobile node that has higher priority and higher ID wants to become cluster head by using LID is not possible.
WCA [23]	Combined weight (W v) is calculated. W v is calculated based on degree of connectivity, mobility, the remaining battery power, and utilized transmission power for each node. The node having the lowest combined weight in network will be elected as cluster head.	Yes	Provide stable cluster, reelection of cluster head is on demand, it also recued clustering overhead.	WCA uses extra parameters for calculating combined weight, which required extra setup and unnecessary message exchanges during cluster head election process. Unnecessary message exchange consumes battery power of mobile nodes.
ICMS [21]	It handles cluster head changes if two or more than two cluster heads are within the same frequency range. Priority of old and new cluster head is calculated based on node degree and the remaining battery power of each cluster head, and cluster head changes will be delayed up to maximum limit. After waiting up to delay time, high priority cluster head elected as cluster head and other changes its status to cluster member within the same cluster.	Yes	Reelection of cluster head is on demand. It reduced cluster head changes to decrease clustering overhead and provide stable cluster head for networks.	If a cluster head acutely required to change within cluster, according to ICMS, cluster head changes will be delayed till it reaching up to maximum limit.
WBC [24]	WBC comprises of three phases. Preclustering phase, cluster formation phase, and cluster maintenance phase. In the first phase, node degree and bandwidth are calculated for each node to get node info. Weight entry is calculated for each node after receiving node info from local table. The node whose weight is higher among all neighbors’ node declares as cluster head.	Yes	Cluster head sustains for longer time by delaying cluster head change to increase stability of network and reduced unnecessary cluster head change.	Extra message exchanges are required for cluster head election which leads towards battery consumption of mobile node.
Prop OSCA	In proposed OSCA, a new node is introduced known as backup node. This node becomes cluster head in absence of cluster head for a cluster, and a new backup node is created later on. OSCA is the extension of WBC.	Yes	Because of backup node concept immediate cluster head election is not required if existing cluster head moves/died from the cluster because backup node upgrades to cluster head and new backup node is elected later on by newly cluster head.	In a cluster, cluster head is responsible for communication with members in a cluster which consumes extra battery power in comparison to cluster members. Election of backup node is extra work, which is done by cluster head, so more battery power will consumed in OSCA while electing backup node for a cluster.

To conclude, it can be prominently assumed that there is a valid action of relevant issue on MANET in the field of clustering algorithms. According to the literature review from the past articles, many algorithms and schemes are proposed for clustering and election of cluster head in mobile ad hoc networks. The present research pursuit analyzes a proposed algorithm named “an optimized stable clustering algorithm for mobile ad hoc networks (OSCA)” to minimize cluster head change and make cluster more stable and reduces clustering overhead. In the algorithm proposed, an extra mobile node (which is known as backup node) is introduced which will work as a second cluster head within the cluster, to make cluster a more reliable and consistent one.

According to the experimental results that proposed an optimized stable clustering algorithm for mobile ad hoc networks (OSCA) algorithm, it will not only be able to make a network more stable by reducing number of cluster head changes but also reduce the clustering overhead. In the proposed algorithm, if a cluster head moves from the cluster, the immediate cluster head election is not required because backup node will act as new cluster head (in absence of cluster head) to make network more stable and later on, a backup node is created by the new cluster head, using LID algorithm.