^{1}

^{1}

^{2}

^{3}

^{3}

^{1}

^{1}

^{2}

^{3}

Maximizing the network lifetime and data collection are two major functions in WSN. For this aim, mobility is proposed as a solution to improve the data collection process and promote energy efficiency. In this paper, we focus on Sink mobility which has the role of data collection. The problem is how to find an optimal data collection trajectory for the Mobile Sink using approximate optimization techniques. To address this challenge, we propose an optimization model for the Mobile Sink to improve the data collection process and thus to extend the network lifetime of WSN. Our proposition is based on a multiobjective function using a Weighted Sum Method (WSM) by adapting two metaheuristics methods, Tabu Search (TS) and Simulated Annealing (SA), to this problem. To test our proposal by experiment, we designed and developed an Integrated Environment of Optimization and Simulation based on metaheuristics tool (IEOSM). The environment IEOSM helps us to determine the best optimization method in terms of optimal trajectory, execution time, and quality of data collection. The IEOSM also integrates a powerful simulation tool to evaluate the methods in terms of energy consumption, data collection, and latency.

A Wireless Sensor Network (WSN) is a particular

Recently, advances in miniature mobile robotics have been made seriously in practice and therefore have allowed the emergence of new innovative applications of WSN based on a Mobile Sink (WSN-MS) and mobile relay nodes, which are able to move inside the region of the WSN, are deployed [

In this paper, we focus on Sink mobility which has the role of collecting data sensed by sensor nodes in the WSN. We also consider the problem of data collection process which consists of finding the optimal path to follow by the Mobile Sink in the WSN-MS field in terms of highest data collection and lowest energy consumption. To do so, several constraints must take into consideration to calculate the optimal trajectory of the Mobile Sink.

Some research works [

Authors in [

In [

In some cases, the moving plan of the Mobile Sink or collector depends on the topology or characteristics of the WSN. In addition, some pervious works try to improve the process of gathering data, especially in applications where the sensing data is important. In their proposed mechanism, the mobile collector stops at each sensor and moves again to another sensor after receiving the data from this sensor.

In [

For the optimization issues in the field of WSN, some works have presented some optimization methods to formally solve the problems related to WSN such as in [

Other research works have proposed to use metaheuristics [

As mentioned above, the aim of this work is to find the moving plan of the Mobile Sink in the WSN-MS. Considering that the sensors represent the cities and the Mobile Sink or data collector acts as the salesman, this problem is very similar to the famous Traveling Salesman Problem (TSP) which finds the shortest path visiting all the cities exactly once.

Some research has addressed the problem of finding the moving plan of the Mobile Sink as in [

In [

Since the problem of finding the path of the Mobile Sink in the WSN is similar to TSP problem, some research works carried out a comparative study of the metaheuristics methods which were used in solving the problem of TSP [

For the perspective of improving the data collection and maximizing network lifetime, we propose in this paper a mathematical model and we define a multiobjective function by using Weighted Sum Method (WSM) to minimize the tour length of the Mobile Sink, minimize the displacements of the Mobile Sink towards sensors that have the lowest error rate (ER), and maximize the signal quality. We use this multiobjective function in adapting the chosen metaheuristics TS and SA to our problem. For the experiments, we have designed and developed a fully Integrated Environment of Optimization and Simulation based on metaheuristics approach (IEOSM) where we have successfully integrated the simulator Omnet++/Castalia [

The reminder of this paper is organized as follows. In the next Section, we give an overview of the problem. Section

In this study, we assume that sensor nodes are deployed in a predefined grid topology with 25, 49, 100, and 144 sensors (Figure

Example of deploying 25 sensor nodes in a sensing field using a 5x5 grid topology used for optimizing the Sink movement.

The Mobile Sink moves in the sensing field according to the moving plan or the trajectory calculated by itself to gather data from sensor nodes. In order to show the efficiency of the moving plan found by the proposed optimization system on the network performance, we set the communication protocol in this study to one single hop routing protocol;

Figure

To optimize the displacement of the Mobile Sink, we propose a multiobjective function

To aggregate these three objectives, we use Weighted Sum Method (WSM) which allows to put all these objectives in one multiobjective function. Each objective is multiplied by a Weighting Factor (WF). All involved factors are normalized to 1;

where

In (

In (

where

In (

Constraint makes the Mobile Sink visits only one sensor

We use the energy consumption model proposed in [

Energy consumption model for WSN.

To transmit a packet of

The energy consumed

Figure

Maximizing the network lifetime and data collection are two major functions in WSN. We aim to find the optimal path for Mobile Sink in the WSN-MS sensing field using metaheuristics optimization approach to calculate the Sink’s trajectory. The addressed problem is similar to the well-known TSP problem. TSP is considered as NP-hard; thus our issue of finding the optimal trajectory of the Mobile Sink in the sensing field is also an NP-hard problem. So, we have adapted the metaheuristics algorithms to our problem. According to the studies in [

We have introduced in the previous section the optimization model of Sink mobility. In this section, we present two optimization approaches on which our contribution is based: Tabu search (TS) and Simulated Annealing (SA) by integrating and adapting the optimization model of Sink mobility into these optimization methods.

The first method that we adapted to achieve our goals is the TS method. TS is based on determining the number of iterations to reach a satisfactory result. The choice of the best solution consists in exploring the neighborhood of the current solution and retaining the best trajectory. In this study, the neighborhood exploration of a solution is based on the permutation in the order of the sensors in the trajectory (Algorithm

Algorithm

Algorithms

The second method we adapted is the SA method. This method is also based on the temperature parameter which is involved in the probability

where

The algorithm of the neighborhood exploration of a solution is presented in Algorithm

Algorithm

Algorithms

We have presented our optimization model and the adapted metaheuristics methods used for the Sink mobility in Sections

Figure

Architecture of the IEOSM tool.

Example of results generated by IEOSM tool.

The following points represent the different modules implemented and constituted our developed tool named IEOSM.

In Figure

We implemented TS and SA in the Optimization Module, which is responsible for calculating the moving plan of the Mobile Sink by using the parameters and settings entered in the Setting and Data Input Module. We can also display in this module the optimization results such as the execution time and the best solution for the sake of comparing the results of both methods TS and SA.

By using the resulting trajectory file in the Optimization Module and the topology file created in the Setting and Data Input Module, the Simulation Module executes a simulation process of the optimization results to evaluate the network performances of the TS and SA.

The Animation Module allows us to visualize the moving plan as illustrated in Figure

CastaliaVIZ tool used for animation purpose [

In this part, we present the numerous assumptions and considerations in order to simulate the different proposed scenarios using our proposed tool (IEOSM).

In this work, we consider a wireless sensor network with the following characteristics:

The Sink is the only mobile node with an additional optimization and mobility system.

The mobility module of the Sink allows it to move freely in the space (aerial displacement or on a surface without obstacles, that is why we eliminate the part of the signal quality in the objective function by fixing its WF to the value 0,

The movement speed of the Mobile Sink is 15 m/s.

Sensor error rates are calculated from a statistical field study for each sensor. The results obtained by this study are used to fill in the error rate parameter in the data file.

All sensors have the same amount of energy. Initially, the default value of energy in Castalia is 18720 Joules, which corresponds to two AA batteries.

The transmission range for each sensor is 50 meters.

Each sensor generates 200 packets to be sent towards the Sink.

The deployment area is represented as a two-dimensional space (2D-zone) of 25, 49, 100, 144 sensor nodes in 600x600m^{2}, 800x800m^{2}, 1000x1000m^{2}, and 1200x1200m^{2} zones, respectively.

In all scenarios, we have used a grid topology (Figure

Presentation of the three studied scenarios.

| | | |
---|---|---|---|

| | ||

| We favor the minimization of trajectory | ||

(100%) | |||

| We favor the minimization of Error Rate | ||

| We balance between the | ||

(50%) | (50%) |

In this section, we present the results obtained by optimization and simulation phases, then we analyze the results in terms of optimization and in terms of network performance.

Previous works as in [

We analyze the results in terms of optimization according to three performance metrics:

Execution time of methods

Traveling distance of the Sink

The cost of the objective function

Regarding the execution time of the two methods showed in Figure

Execution time of methods in all scenarios.

Table

Comparison of the two methods results according to traveling distance and objective function in each scenario.

| | ||||
---|---|---|---|---|---|

| | | | ||

| | | | ||

| | 2988.63m | | 2988.63 | |

| 5601.6m | | 5601.6 | | |

| 13062.7m | | 13062.7 | | |

| 18257.9m | | 18257.9 | | |

| |||||

| | 3763.17m | | 35796.3 | |

| 7144.44m | | 81275.8 | | |

| 17294.5m | | 179599 | | |

| | 28627.2m | | 267472 | |

| |||||

| | 3465.03m | | 19399 | |

| 7830.25m | | 46360.8 | | |

| 17684.2m | | 101497 | | |

| | 28427.1m | 152345 | |

The results obtained in all scenarios in terms of traveling distance and the cost of objective function are depicted in Figures

Graphical illustration of traveling distance.

Graphical illustration of objective function costs.

Tables

Energy consumed

Packets reception rate

Latency

Energy consumed in each scenario.

| |||
---|---|---|---|

| | | |

| | ||

| | 33.54J | |

| 45.4J | | |

| | 59.02J | |

| | 88.09J | |

| |||

| | 33.81J | |

| 46.14J | | |

| 87.07J | | |

| | 171.70J | |

| |||

| | 33.9J | |

| 45.83J | | |

| | 87.28J | |

| | 171.46J |

Packets reception rate in each scenario.

| |||
---|---|---|---|

| | | |

| | ||

| | 97.82% | |

| | 99.54% | |

| | 99.44% | |

| | 98.90% | |

| |||

| | 99.62% | |

| 99.8% | | |

| 96.3% | | |

| | 96.15% | |

| |||

| | 93% | |

| 96.12% | | |

| | 96.98% | |

| | 95.57% |

Average latency calculated in all scenarios.

| |||
---|---|---|---|

| | | |

| | ||

| | 89.23s | |

| | 169.68s | |

| | 383.27s | |

| | 551.22s | |

| |||

| | 103.24s | |

| | 194.13s | |

| 497.5s | | |

| | 829.46s | |

| |||

| | 92.81s | |

| 211.39s | | |

| | 526.63s | |

| 838.35s | |

Average energy consumption for each node versus number of nodes for the two methods.

Packets reception rate for the two methods by varying the number of nodes in each scenario.

Average latency calculated in all scenarios.

Table

Table

The packets loss is a very common issue in WSN due mainly to collisions and sensor node failure in the classical WSN when sensor nodes and the Sink are considered stationary. The packets loss in this study is due to the mobility of the Sink;

In Scenario 2, when we promote the error rate in the objective function, we are interested in the quality of data reception. So, it is clearly showed that the results obtained in most cases are high for both methods, and the reception rate decreases with the increase of number nodes in different topologies.

The decrease of the reception rate in Scenario 3 compared to the other scenarios is due to the movements of the Sink. While the Sink moves in the deployments field towards a specific or destined sensor node in its calculated trajectory, the Sink passes through radio ranges of other sensor nodes nondestined, and these sensor nodes start a communication with the Sink. So, the communication established between the Sink and these sensors is not completely done.

Usually, the latency refers to the time required for a data packet to be transmitted from the source node to the destination node through a network. Latency in mobile WSN is the time taken for a communication to send a data packet from a sensor to the Mobile Sink,

The latency of a data packet

Table

Each sensor node generates data packets at the application layer and waits for the Mobile Sink to approach its radio range to send the stored data. So, the latency in this case depends heavily on the travel time of the Mobile Sink to reach the space area where sensor nodes are deployed. Thereby, the latency increases with the increase of the number of sensor nodes which the Sink must pass through to finally reach the destination. We note that, in all the scenarios, both TA and SA methods give satisfactory results in terms of optimization and network performance.

The data collection is an important function in WSN and packet losses are a common problem in this kind of networks. These losses are the consequence of several collisions and nodes failure as they exhausted their batteries when acting as relay nodes in the network. Moreover, in this study, we find that in all scenarios the reception rates are higher than 93%, which is a very important and a satisfactory reception rate in WSN.

In this paper, we are interested in optimizing data collection in WSN with Mobile Sink (WSN-MS). The main goal consists of using metaheuristics to optimize the movements of the Mobile Sink in the network. For this sake, we developed an Integrated Environment of Optimization and Simulation to this matter, named IEOSM. In IEOSM, we have implemented the proposed multiobjective function and the two state-of-the-art metaheuristic methods, TS and SA. We have adapted the algorithms of these methods to the addressed problem of finding the optimal trajectory for the Sink in the WSN-MS network. More precisely, our main objectives in this work are twofold: minimize the movement of the Mobile Sink in terms of distance and minimize the movement to sensor nodes that have the lowest error rate. The desired gain here is the reliability of data collection and energy efficiency in WSN-MS.

In defined scenarios, both methods offer significant results in terms of optimization and network performance. Due to the great exploration of the search space that can lead to a global optimum, the SA gives better solutions in terms of optimization in most cases compared to those obtained by the TS. In terms of network performance, we observed that both methods used for determining the trajectory of the Mobile Sink give satisfactory results. The choice of the optimization method and the Weighting Factors depend on the use case and the objective set.

The developed environment IEOSM is extensible to include other metaheuristics and other optimization tools. Also, it would be more interesting to really validate the results of this work to experience them in practice using test-beds.

The data used to support the findings of this study are available from the corresponding author upon request.

The authors declare that they have no conflicts of interest.

This work was supported by Shenzhen Science and Technology Innovation Committee with Grant No. JCYJ20170306170559215 in Shenzhen China and Wuhan Science and Technology Bureau with Grant No. 2017010201010105 in China. This work was supported also scientifically by the laboratory of Research in Industrial Computing and Networks (RIIR), University of Oran 1 Ahmed Benbella, Oran, Algeria.