Grid based Energy-Efficient Cross-Layer Optimization Model in WSN Using Dual Mobile Sinks


 In recent days, wireless sensor network (WSN) gained more attention among researchers as well as industries. It is composed with massive number of sensors which are independently organized cooperate with one another for collecting, processing and transmitting data to the base station (BS) or sink. Since sensors undergo random deployment in harsh environment, it is difficult or not even possible to replace the batteries. So, energy efficient clustering and routing techniques are preferable to reduce the dissipation of energy and improve the network lifetime. This paper introduces a new Grid based Energy-Efficient Cross-Layer Optimization Model in WSN Using Dual Mobile Sink (GEECLO). The proposed method involves three main processes namely grid partitioning, clustering and routing. Initially, the entire network is partitioned into different zones and then sub zones. Then, type II FL process gets executed to select the CHs and construct the clusters. Finally, dolphin swarm optimization algorithm (DSOA) based routing process takes place to select the optimal path for inter-cluster communication. A detailed simulation analysis takes place to ensure the betterment of the GEECLO algorithm. The obtained experimentation outcome depicted that the GEECLO model offers maximum energy efficiency and network lifetime.


Introduction
The rapid development in wireless communication and micro-electronic models enable a fast growth in tiny, minimum-cost, multi-operational sensors and so on. Some of the sensors are generally applied in target region by random monitoring of external features present in ecological factors like moisture, humidity, temperature, pressure etc. Then, the observed information is transmitted to data collector or sink under the application of cooperative model which is termed as multi-hop where the sink forwards the data to remote server in order to perform data examining process. On the other hand, a sensor has the potential of self-organizing the local collaboration which tends to build a Wireless Sensor Networks (WSN) [1]. Some of the most useful features of WSNs are fast deployment, maximum fault tolerance, self-organized, realistic data transfer and so on. This forms an adaptive environment to be fixed in an unmanned platform, specifically in armed forces or disaster monitoring. In addition, WSNs have been used in observing commercial product line, farming as well as wildlife analysis, healthcare, modern homes, and so on. [2,12].
In general, Sensors are recharged by using energy-filled batteries which is impossible to replace battery often due to massive number of sensors are expensive. To resolve the limitation, the sensors must be equipped with higher-constrained batteries. Unfortunately, if the battery attains its choke point then, it is complex to enhance the battery [3]. Thus, energy limitation issue in WSN is reported by obtaining energy effective protocols [4]. Also, WSN has the shortcoming of uneven energy of sensors. Every sensor is comprised with a monitoring limit in case of a node expiry, and then fade zones would arise, which results in decreased network performance. The mechanism of irregular energy in WSN is same as energy holes where it is caused due to "hot spots" issue. Generally, "hot spots" problems occur in WSN by static sink as well as permanent network topology. Some of the nodes in the sink are loaded with engaged data as it has to forward the data packages from external layer to sink which leads to drain out the energy in a rapid manner. As it is comprised with static sink as well as with fixed network topology, energy becomes more heterogeneous in further process. Here, the network lifespan is a significant evaluation procedure to calculate the function of network which is generally described as the duration of first node expiry.
A major challenge in this work is to deal with problem of energy efficiency as well as energy maintenance where optimal solutions are attained. The clustering model reduces the power utilization of WSN by classifying the sensors as clusters on the basis of specified patterns. For every cluster, more than one cluster heads (CHs) has been selected and fed as relay nodes for the cluster members. Clustering helps to simplify the network topology architecture as well as eliminates the direct communication among sensors and sink. In addition, data fusion is applied in CHs for extracting the unnecessary information to reduce the overhead of CHs. Conventional routing protocols use clustering with Low-Energy Adaptive Clustering Algorithm (LEACH); but, the model is used to select the CH for unreasonable as well as more additional function has to be performed which is relied on LEACH protocol respectively.
Here, Sink mobility method is evolved which is assumed to be an effective model to solve the uneven energy constraint in WSN. For mobile sink-supported WSN, sink is adopted by smart robots; also it has the capability of freely moving over sensing field. There are few benefits which are emerged by establishing the sink mobility approach . Initially, "hot spots" issues could be reduced by motion of sink. Usually, area of sink is traffic hubs and if sink is in motion, then traffic hub is transferred. Subsequently, sensors act as "hot spots" for balanced application of energy. Secondly, overall power application is alleviated by minimizing the transmission distance from communication pairs by considering that sink mobility criteria is properly developed. Followed by, delay of network could be decreased and throughput of network is maximized by applying the sink mobility. Consequently, the network link has to be assured, especially at the case of disconnected sensors. Though it has mobile sink existence which has several measures, it meets various challenging issues [5]. Hence, position of mobile sink must be frequently broadcasted or detected by sensor networks that tend to improve the workload of network. In addition, sink mobility has to be developed more accurately to comply with local nodes for transmitting data.
Here, [6] presented a technique named Load Balanced Clustering and Dual Data Uploading (LBC-DDU). For LBC-DDU, the entire network is segmented to 3 layers: sensor layer, CH layer, and SenCar layer. Initially, SenCar is able to measure an optimal path and use the path to collect data by applying single-hop transmission. Once the selected point has been visited, SenCar would come back to base station (BS) and prepare for the upcoming process. Hence, it is constrained with 2 antennas where it is capable of interchanging data using 2 CH at same time by consuming the Multi-User Multiple-Input and Multiple-Output (MU-MIMO) that reduces the delay as well as to improve the efficiency. [7] introduced a Tree-Cluster-Based Data-Gathering Algorithm (TCBDGA). In TCBDGA, the node weight is a measure of various aspects like remaining energy, number of neighbors as well as distance to BS. Every node selects the corresponding neighbor along with higher weight as parent node. Similarly, a tree-construction is applied and all trees are degraded to various sub-trees which are based on the depth and data traffic. Hence, the final outcome reveals that, it is optimal with respect to power application. This paper introduces a new Grid based Energy-Efficient Cross-Layer Optimization Model in WSN Using Dual Mobile Sink (GEECLO) [8 -10]. The proposed method involves three main processes namely grid partitioning, clustering and routing. Initially, the entire network is partitioned into different zones and then sub zones. Then, type II FL process gets executed to select the CHs and construct the clusters [11]. Atlast, DSOA based routing process takes place to select the optimal path for inter-cluster communication. A detailed simulation analysis takes place to ensure the betterment of the GEECLO algorithm. The obtained experimentation outcome depicted that the GEECLO model offers maximum energy efficiency and network lifetime.

Network model
A sensor is constrained with N number of sensor nodes that is deployed in a random manner in which has to be observed and some considerations are developed.
• Sensor nodes and BS are immobile • Every nodes has same quantity of energy after node deployment • All nodes are homogeneous • The distance among nodes and BS could be estimated by Received Signal Strength

Indicator (RSSI)
• Node death is due to exhaustion of energy • Sensor nodes are capable of changing the power of transmission by applying power control based on its the distance to receiving node

Energy model
A simple first order radio method is used as energy model of network. The energy drained for transmission and reception of l bit packet across distance d is given in Eq. (1) and Eq. (2).
where denotes dissipated energy in transmitter or receiver unit, d0 indicates threshold distance that is measured by = � ⁄ . According to the transmission distance d, free space ( ) or multipath fading ( ) is employed in transmitter amplifier.

Mobility model
The mobility model defines the moving principle of the nodes in a network and estimates velocity, location, as well as node's acceleration of network region. The purpose of introducing mobility model is to examine the operation of routing protocol. Assume 2 two nodes as and that is placed in ( , ) and ( , ) so that ∈ ( , ); ∈ ( , ). Thus, and migrate to a specific dimension using variable velocity by forming the angle φ 1 andφ 2 .
Generally, nodes and occupy distance 1 and 2 , then, the nodes attain novel location ( new , new ) and( new , new ) , correspondingly. Hence, Euclidean distance at primary duration for nodes that are placed at ( , ) and ( , ) which is provided in the following, Let nodes and is iterated by a velocity ϑ as well as ϑ that forms an angle φ 1 as well as φ 2 to enclose a distance 1 and 2 that is given as, In time , the node is transferred to a new position which is accomplished by node that is given as follows, If ( , ) is moved at 2 by creating an angleφ 2 , it concentrates on new position which is depicted as, If nodes attain a novel position, then the distance from nodes and are formulated as,

The proposed GEECLO model
The proposed GEECLO model incorporates different sub processes namely grid partitioning, clustering and routing. Initially, the entire network is partitioned into different zones and sub zones. Then, type II FL process gets executed to select the CHs and construct the clusters.
Finally, DSOA based routing process takes place to select the optimal path for inter-cluster communication.

Grid partitioning
Consider a network area of 1000x1000m 2 which undergo the deployment of sensor nodes in a random manner. Then, the network area is partitioned into different zones of 200x200m 2 .
Afterwards, clustering process takes place at every zone. The cluster process will construct different clusters or sub-zones under every zone and separate CHs will be elected in each subzone.

Type II FL based clustering process
After the zones are constructed, clustering process is carried out at every zone and a number of clusters are organized along with CHs selection mechanism. A set of three parameters namely residual energy, distance to BS and node mobility are used to elect the probability of becoming CHs. A node with maximum residual energy, minimum distance to BS and low mobility has the higher chance of becoming CHs.
The T2FL generates an optimal computation and performs far better than T1FL method. Some of the inference methods as well as fuzzy system applied to presented technique as provided Fig. 1.
There are 3 fuzzy input parameters which are adopted to choose the tentative CH.  Fig. 1. set. Assume FOU is implied as , when º [0, 1], and f → 0, then MF is named as T1FL whereas → 0 to 1, then T2FL has a wider range of FOU which is from 0 to 1. However, the rule formation is similar to T1FL. It is represented as: T2FL method has 4 elements as given below:

1) Fuzzifier: Converts inputs values to fuzzy values.
2) Fuzzification Module: Inference engine integrates rules and provides a mapping of input type-2 fuzzy sets to output type-2.
3) Defuzzifier: The type-reducer produces a T1FL result which is translated to numeric outcome by implementing the defuzzifier.

4) Knowledge base:
Consist a group of fuzzy rules, and a MF set is termed as data base.
Hence, the rules are obtained from a formula that is shown in Eq. (12).

DSOA based routing process
Once the CHs were chosen and clusters are effectively constructed, the CM starts to sense the environment and transmits the data to its respective CHs. Then, it is necessitating forwarding the data to sink by the optimal path. At this point, effective routing mechanism plays a vital role which offers a set of optimal paths between two nodes. The dolphin technique is emerged from the extension of dolphin population's hunting model. In this method, dolphins achieve predation by using 4 significant levels, namely as given below: • Searching stage where the optimal solution is attained by frequent processing.

Initialization
In an optimization problem, every dolphin denotes a possible solution. Dolphin in this work is where implies constant acceleration which creates rapid sound. In this point, , is renamed as major contact timeT 2 . If ( ) > � �, is substituted using ; else, does not change. The multi fitness function is relied on the aspects of energy, distance to BS and mobility.

Fitness based on energy
The energy constraint has been determined by applying the Eq. (15). The overall energy of a cluster is the combination of residual energy in th CH as well as power unique nodes. Hence, the residual energy existed in nodes must be improved where the network functions are retained with prolonged lifespan of the network.
where and implies total number of CH and nodes present in network, signify the energy present in th node from th cluster whereas indicate the power of th CH.

Fitness based on Distance to BS
Inter-cluster distance is defined as distance measured from CH and nodes present in the corresponding cluster which must be lower to obtain optimal network. Therefore, fitness applying the inter-cluster distance can be written as, where and represent the position of th CH and place of th node. � , � 2 shows the distance among position of th CH as well as th node in respective cluster. Thus, the variable resembles normalizing factor.

Fitness based on mobility
Mobility of the th cluster from past and present location is obtained to estimate the mobility of node which is measured by,

Predation stage
In predation stage, every dolphin requires to estimate surround radius R and compute distance from dolphin neighborhood optimal solution and location after predation stage which is relied on the predefined data. This process tends to obtain a novel position.

Performance Validation
A brief explanation of the performance evaluation of the presented GEECLO model is provided here. The presented model has been simulated using OMNET++ 4.  Followed by, the FUCHAR method offers slightly better lower energy dissipation over other methods. But, it is revealed that it does not outperform the two methods namely GEECLO and ECDRA methods. At the same time, it is apparent that the ECDRA method exhibits energy efficient characteristic over the other methods except GEECLO. The presented GEECLO is found to be highly energy efficient and consumes only limited amount of energy due to the proper election of CHs and optimal route selection.   shown that the LEACH model offers least network lifetime over the compared methods. But, the TEEN protocol outperforms LEACH by offering a higher network lifetime. But, it does not outperform the rest of the methods. Next to that, it is depicted that the FUCHAR model reported moderate and manageable network lifetime outcome. But, it is revealed that it does not outperform the two methods namely GEECLO and ECDRA methods. In the same way, it is evident that the ECDRA method provides high network lifetime over the compared models.

Throughput Analysis
Interestingly, the presented GEECLO exhibits maximum network lifetime under varying hop count in a significant way. From the detailed experimental analysis, it is found that the presented GEECLO is found to be highly energy efficient and consumes only limited amount of energy due to the proper election of CHs and optimal route selection.

Conclusion
This paper has introduced an energy efficient clustering and routing technique called GEECLO model to achieve energy efficiency and maximize network lifetime in WSN. The GEECLO model involves three major processes such as grid partitioning, clustering and routing. To begin with, the entire network is partitioned into different zones and then sub zones. Then, type II FL process gets executed to select the CHs and construct the clusters. Next, DSOA based routing process takes place to select the optimal path for inter-cluster communication. A detailed simulation analysis takes place to ensure the betterment of the GEECLO algorithm. From the detailed experimental analysis, it is found that the presented GEECLO is found to be highly energy efficient and consumes only limited amount of energy due to the proper election of CHs and optimal route selection. In future, the presented GEECLO model can be further improved by the use of hybrid data transmission schemes andpredicting the effective position of mobile sinks.