Optimizing MANETs Network Lifetime Using a Proactive Clustering Algorithm

In wireless sensor networks that consist of a number of power constrained sensor nodes, the foremost challenges are the limited energy and system lifetime. Therefore, designing efficient routing protocols, which prolong the network lifetime, is one of the most critical issues. This paper evaluated several clustering algorithms, namely: Highest Degree Clustering Algorithm (HDCA), and Lowest Identifier Clustering Algorithm (LIDCA) under three metrics: throughput, Packets Delivered Ratio Factor (PDR) and network lifetime. One of the most important challenges facing Mobile Ad hoc Networks is saving energy that led to a longer network lifetime, which is why we proposed a new clustering algorithm that is considered to be more efficient under network lifetime, and it compared to the clustering algorithms mentioned above. Our experiment occurrences showed that the proposed clustering algorithm supplied a relatively better network lifetime and a more efficient energy distribution for the nodes.


(LIDCA):
This algorithm searches for the oldest and lowest ID from the current cluster. The node has been selected as (CH) for having the lowest ID. Since it is an identifier-based clustering algorithm, it will first assign a unique id to the assigned node or the nodes which are within one cluster [12].

(HDCA):
As this algorithm utilizes area data for cluster arrangement, it selects the cluster head from the highest degree node in an area. It is a connectivity-based clustering algorithm, and the degree of a node depends on its distance from others, taking the node with the highest degree. The degree of a node is based on the number of nodes associated with or connected to that node. Whenever progressively many local nodes are associated with the cluster node, the highest degree node increments, after which that particular node turns into the cluster head of that cluster [11], [12].

Literature review
Throughout the last decade, several studies have been conducted for the selection of cluster heads (CH) in ad-hoc mobile networks using certain the clustering algorithms such as LIDCA [13], [14], WCA [15], and HDCA [16].
The experimental work presented in [2], [17] mainly dealt with (WCA), and has proven that the battery power is consumed by the cluster head more than ordinary nodes due to its additional responsibilities. Similarly, in WCA the weight of each node is determined by its coverage, mobility, transmission power, and battery power, where the node with the lowest weight among its neighbors is chosen as the cluster head.
As for the research on LIDCA conducted in [18], it showed how this algorithm generates a lot of unused clusters, bottlenecks,and causing network traffic. The clustering here is fast and inexpensive because of the random allocation of ID to the nodes. Besides, in this algorithm; the node with the minimum ID among its neighbors will be selected as the cluster head. To avoid weakness and tiredness; the cluster head nodes are changed at certain intervals along with the changes in the ID of the nodes.
The work on HDCA presented in [19], [12] pointed out that this algorithm creates a group mobility pattern, which is a group of nodes transferring in a similar direction and at the same speed. Moreover, it leads to the reduction of the number of clusters, and thereby supplies stability to network. However, it has been noted that it has e relatively very low throughput value.
A position-based clustering algorithm is proposed in[8], [18] for large-hop vehicular ad-hoc networks, which is based on the geographic position of vehicles and traffic information regarding the cluster formation for this protocol. Additionally, a maximum distance between members and cluster head is used to manage the cluster size as much as possible. Furthermore, the overhead is high for V2I and V2V communications in this protocol.
As for [20], a packet delivery ratio (PDRSC) is proposed between the candidate set calculation model and source node, which leads to the consideration of the network interface. The ERTO is also produced, which optimizes PDRSC, the degree of relay nodes, and the expected energy consumption.
A static clustering algorithm was proposed in [21] to expand the network lifetime. At first, they examined the energy balancing approach and inferred the guideline for the clustering algorithm, after which they proposed an organization system to disperse CHs and MNs in prior known areas, in contrast to the previous arrangements.
In the work of [22], a Mobile Data Gathering-based, Asynchronous Clustering method is proposed, depending on (ACMDGTM). It enhances the lifetime of the network through diminishing the hot spot problem area in the sensor node. The cluster head was selected in view of the remaining energy as well as the sensor's area. Moreover, the sink node considers the moving time from clusters to sink node, and makes use of the data overflow time. Finally, the network lifetime is drawn out and the energy is used productively.
In [23], the CHs is first selected via PSO, followed by the sensor nodes to the CHs which are assigned based on a Weight Sum Approach (WSA) approach. WSA depends on the distance from the CH to the sink, the CH node degree, the residual energy of the CH, and the distance between the sensor node and the CH.
As for [24], the clustering trouble in WSNs was solved by means of the Firefly Algorithm which is a metaheuristic algorithm. Numerous parameters have been used, such as energy and distance to discover the most suitable set of CHs among the normal sensor node. To decrease the delay in the network, the distance between the nodes and the sink is regarded as a significant parameter.
In the work of [25], the network lifetime is enhancement according to the recommended novel algorithm. This algorithm is based on the solid design recommendations installed in [26]. The central processing entity or the sink node dynamically improve the communication activity levels of the cellular sensor nodes via the residual energy information mentioned through the sensor nodes, in order to save energy without sacrificing the sensory information data throughput.

Research method
A virtual work environment was created to evaluate the performance of each selected clustering algorithm: Highest Degree Clustering Algorithm (HDCA), and Lowest Identifier Clustering Algorithm (LIDCA) , as well as the proposed clustering algorithm, all regarding three metrics, namely throughput, Packets Delivered Ratio Factor (PDR) and network lifetime. The three scenarios will be clarified in detail afterwards. For each scenario, the simulation program for any of the clustering algorithms was implemented one hundred times. Table (1) shows the parameters of environment for each scenario.

Results and discussion
In this section, three scenarios will be explained so as to obtain the simulation results when implementing each of HDCA, LIDCA, and our proposed algorithm.

First scenario
The first scenario is realized by creating a virtual work environment including the parameters mentioned in Table 1, for the proper stimulation of the performance of Highest Degree Clustering Algorithm (HDCA) for three metrics: throughput, PDR, and network lifetime. The node with maximum number of neighbors is selected to be a cluster head. A node's degree means the number of neighbor nodes. Figure 1 shows how the clustering algorithm works.  Table (2) shows the results of implementing the simulation program of HDCA a hundred times.

Second scenario
The second scenario is realized by creating a virtual work environment in light of the parameters referred to in Table 1, so as to simulate the performance of Lowest Identifier Clustering Algorithm (LIDCA) in three metrics: throughput, PDR, and network lifetime. The oldest nodes in the network is selected to be the cluster head. Figure 2 shows how the clustering algorithm works.

Third Scenario
The third scenario is creating a virtual work environment with parameters mentioned in Table 1, so as to simulate the performance of the proposed clustering algorithm under the metrics of throughput, PDR, and network lifetime. The main goal of the proposed clustering algorithm is to save energy and thereby enabling the network to function as long as possible. This algorithm depends on two factors in selecting the cluster head node, namely the node battery energy and the node degree. The standard of the node is represented by a simple equation called weight equation, which has two weights, each equaling 0. 5, the sum of which equals 1. The weighted equation can be explained as shown below: W = 0.5 × Battery Energy + 0.5 × Node's Degrees The node with maximum weight is selected to be a cluster head. Figure (3) explains the proposed clustering algorithm work.

Algorithms Performance Comparison
A comparison has been drawn among the proposed clustering algorithm and each of the clustering algorithms HDCA, and LIDCA in light of three metrics: throughput, PDR, and network lifetime.   Figures (4-6) to illustrate the differences in algorithm performance. These figures show the average values for the throughput, PDR, and network lifetime of the clustering algorithms, respectively.

Conclusion
Clustering algorithms make the mobile ad-hoc network work efficiently, as specific nodes called a cluster head node are assigned to be responsible for transmission operations like gathering data and sending them to the goal node (Base Station Node). Therefore, each clustering algorithm has different rules for the selection of a cluster head node, which in its turn leads to different performances, as has been pointed out in this paper. Moreover, the most important challenge for Mobile Ad-hoc Networks is the saving of energy, which prolongs the network lifetime as much as possible, thereby leading to an increase in PDR value. Higher PDR values are associated with lower rates of packets loss.
When evaluating the performance of the selected algorithms, it has been noticed that the performance for High Degree Clustering Algorithm (HDCA) were better than that of the Lowest Identifier Clustering Algorithm (LIDCA) in light of three metrics: throughput, PDR, network lifetime. Therefore, the proposed algorithm included particular criteria for both algorithms in selecting the cluster head. Overall, the merging of criteria for both algorithms is represented by a simple weights equation, where the higher weight node is the most appropriate one to be a cluster head. Therefore, the conclusion can be drawn that the proposed algorithm has offered a lower performance under the metric of throughput, as this algorithm appeared to be rather timeconsuming in forming clusters, as compared to the time required for the selected algorithms.