1. Introduction
The Internet of Things (IoT) is a dynamic wide-ranging network that includes varieties of mobile and static sensors, data gathering devices such as global positioning system (GPS) sensors, radio frequency identification (RFID) sensors, and laser and infrared sensing scanners that are connected through the Internet and exchange information with each other based on an agreement [
1,
2]. Wireless sensor network (WSN) is a group of static sensor nodes which gather information and deliver to the base station. The data gathered by some static sensor nodes can be imprecise and it may suffer network failure, which affects connectivity and reliability of the network. Hence, it has limited applications. Due to fast development in IoT and mobile Internet technology, mobile wireless sensor network (MWSN) has become a popular field of research and replacement of static WSN. MWSN is a collection of tiny mobile sensor motes that aim to sense data from the environment and effectively deliver the base station. These mobile motes maintain the communication links among the neighboring nodes while collecting and processing the data for efficient communication, enhancing the network performance [
3]. MWSN can be classified into three types [
4], which are as follows.
Mobile base station and mobile sensor nodes
Mobile base station and static sensor nodes
Static base station and mobile sensor nodes
The main difference between MWSN and WSN is sensor node mobility, which enhances the connectivity, network adaptability, and the reliability of the sensor networks. Simultaneously, MWSN stabilizes the energy utilization and improves the lifetime of the sensor network [
1]. However, due to frequent changes in topology and complex environment in different applications, MWSN faces challenges related to energy efficiency, data delivery and data aggregation [
5,
6]. MWSN has a wide range of applications that depend on environment and setup. Space mission sensor robots [
7], border monitoring mobile sensors [
8], flying bee robot-sensors (ROBOBEE) [
9] and undersea submarines movement detection by mobile sensor nodes [
10] are common examples of MWSNs.
Routing is the core aspect of all sorts of networks, which is used to send the (sensed) data between mobile sensor nodes and base stations. This requires an efficient and reliable communication. In the literature, several routing schemes have been studied for WSN, e.g., flooding, multipath, Quality of Service (QoS), hierarchical and geographical routing, etc. Recently, much work has been done on hierarchical based routing protocols for static and mobile networks. Since it chooses a comparatively short path for routing, it is resistant to failures and efficient in the use of remaining energy. Then, it has a low overhead and enhances network lifetime [
11,
12]. In a hierarchical architecture, sensor nodes are classified into optimal groups based on their homogeneous properties. Afterwards, these group members select a Cluster Leader (CL) that helps member nodes transfer sensed data to the base station.
Low-Energy Adaptive Clustering Hierarchy (LEACH) is a basic energy efficient routing protocol of hierarchical routing protocol’s family [
13], where the sensor nodes remain static. Hence, it achieves higher energy efficiency but has limited applications. However, mobile sensors have several applications but also bring many challenges such as energy depletion and squatter network lifetime. LEACH-Mobile is the first mobility-based LEACH routing protocol [
14], where CL has to wait for two successive time division multiple access (TDMA) failed cycles before declaring a mobile node as a non-member. Although such a mobile node may become a member of another CL, data for two TDMA slots is lost. Consequently, this approach increases overhead and reduces energy efficiency.
Figure 1 shows an example of a single hop mobile LEACH routing protocol, where a base station is static and sensor nodes are mobile.
The role of mobility models in the performance of MWSN is of enormous importance. Mobile sensor node distribution and selection of an appropriate mobility model not only improve the overall performance, but also improve the clustering process of a routing protocol [
15,
16]. To determine the best mobility model in a specific WSN application is still a complex issue. The mobility model includes movement order for MWSN, location, and acceleration showing maximum and minimum velocity over time. Therefore, it will help to evaluate the performance of MWSN routing protocols. In [
17], effect of mobility models on Distance Vector (DV)-hop based localization algorithm is discussed. In [
18], the performance of Routing Protocol for Low power and Lossy network (RPL) routing algorithm is analyzed on three different mobility models. There are various mobility models, but, in this study, we considered four mobility models: Reference Point Group Mobility model (RPGM), Random Waypoint Mobility Model (RWP), Gauss–Markov Mobility Model (G-M) and Manhattan Grid Mobility Model (MG) [
19]. In the areas such as battlegrounds, disaster relief mission and other hazardous missions, where mobile sensor nodes have got their applications, each member (mobile sensor) node has to follow a group leader to achieve successful completion of the mission [
1]. In RWP model [
20], the nodes move in random order within the simulation area without following any group leader. In GM model [
21], the values of direction and speed of any node at a particular time are updated depending upon immediate previous value. In MH model [
22], the simulation area is split into horizontal and vertical path lines. Each node has to change direction (it can turn left, turn right or go straight) at intersection point depending upon probabilistic value. In RPGM mobility model, which is quite similar to clustering process, each group of nodes has a logical center (group leader) along with member nodes [
17,
23]. RPGM is used to simulate real time applications such as battle field, border monitoring, and disaster relief mission where mobile sensor nodes need to move in the form of group while following a movement of the group leader [
24]. Consequently, RPGM shows higher connectivity than other mobility models [
25,
26].
Sensor node connectivity is specified as a ratio of number of sensor nodes that can actively communicate with the base station to the total number of sensor nodes. In MWSN, connectivity and coverage factors are interrelated. An ensured connectivity with a dynamic coverage is required for efficient sensing of any event. Connectivity depends upon several aspects such as sensor node distribution, communication energy, mobility, distance between sensor nodes, signal dissemination medium, signal dissemination loss, etc. [
27].
Due to mobility, sensor nodes alter their positions after initial distribution. Sensor node mobility and node failure affect the transmission path, which impacts connectivity in MWSN [
28]. Unplanned mobility can create coverage problem [
29], whereas planned mobility (mobility models) can be applied to improving connectivity and enhancing lifetime of network [
30]. To best of our knowledge [
31], only two connectivity based LEACH algorithms have been proposed thus far: LEACH based on Density of node distribution (LEACH-D) [
32] and Orphan- LEACH (O-LEACH) [
33]. However, both studies are proposed for static sensor nodes.
In this paper, a novel connectivity based LEACH-Mobile Energy Efficient and Connected (LEACH-MEEC) algorithm is proposed. The binary disk sensing model is used to calculate neighborhood density. We propose a probabilistic connectivity model to compute connectivity among neighboring nodes. The main contributions of this paper are given as below.
We propose LEACH-MEEC, where the connectivity and remaining energy of mobile sensor nodes are used as metrics for CL selection after the first round and onwards. This proposed metric significantly improves the performance as compared with the existing schemes.
The proposed LEACH-MEEC is analyzed under different mobility models, using eight datasets with two different speed levels.
The rest of paper is organized as follows.
Section 2 includes the related work.
Section 3 discusses the proposed framework of LEACH-MEEC.
Section 4 presents the simulation and results.
Section 5 concludes the paper.
2. Related Work
A heterogeneous mobile LEACH protocol is proposed in [
34]. It contains static sensor nodes with mobile base stations where CL is selected based on probability function and data are transferred to base station based on energy function. A mobility factor parameter is introduced for the CL election by initial mobile LEACH-Mobile-Enhanced (LEACH-ME) routing protocol [
16]. However, this protocol increases complexity and energy depletion but performs well at a high mobility. Another mobility-based clustering (MBC) protocol was proposed by Deng et al. [
35], who used two metrics for the selection of CL, i.e., remaining energy and node speed. It has applications for large-scale networks but there is a rapid change in distance between nodes due to the high mobility. CL node may select a member node that has maximum remaining energy and mobility factor, but it may have a maximum distance from the CL node as well. Consequently, it drains CL energy. A mobile LEACH algorithm for large-scale networks was proposed by Souid et al. [
36], where energy is considered as the main component, defining three levels of energies with round time length. However, it has applications for only small scale static sensor network. A energy efficient LEACH-1R was proposed by Khushbu and Khunteta [
37], where the CL selection is performed after the first round, only if the remaining energy is less than the threshold value. However, the author did not specify the mobility model. Similarly, LEACH-Centered Cluster-head (LEACH-CCH) was proposed by Corn and Bruce [
38], where the energy utilization is reduced by predicting the positions of mobile sensor nodes and reconstructing clusters accordingly. However, the nodes distribution is performed randomly and the mobility model is not mentioned. A LEACH-Mobile Average Energy (LEACH-MAE) based routing protocol is proposed in [
39], which selects CL based on remaining energy metric. Here, a CL can add member nodes in a cluster that may have maximum remaining energy, but their distance from the CL node may also be maximum. Hence, CL node may lose a lot of energy to aggregate data from member nodes. LEACHDistance-M [
40] is proposed for MWSN, where the selection of CL is based on remaining energy, lower-upper threshold distance and minimum mobility. However, it only assumes 30% of sensors nodes are mobile, while remaining sensor nodes are static.
Those above-mentioned approaches [
35,
36,
37,
38,
39] elect CL based on residual energy. The importance of energy metric in MWSN is vital, however, relative position, radio coverage and spatial density of mobile sensors are also important metrics for stability, consistency, and reliability of the CL, respectively. Connectivity in MWSN is a function of three important factors: transmission range, sensor speed and spatial density [
41,
42]. Therefore, connectivity among nodes achieves robust communication, energy efficiency, reduced communication overhead and network scalability. The concept of connectivity in WSN is used in different perspectives. In [
43], Abdel-Mageid et al. improved connectivity using potential field theory and local virtual force to calculate mobility and location among mobile neighbor nodes. The conditional connectivity based algorithms were discussed in survey article [
44], where the radio range were considered larger than twice of sensing range, consequently they achieved high connectivity at the cost of energy efficiency. To the best of our knowledge, connectivity as a metric for CL selection in mobile LEACH is unexplored, however, connectivity is discussed in MWSN in different perspectives. Hence, the motivation of this work is to combine connectivity with remaining energy in CL selection.
5. Conclusions
The mobility of nodes have many constraints in a WSN, energy efficiency being one of them. This study proposed an improvement in energy efficient LEACH-MAE and Mobile-LEACH. We proposed that the selection of CL is based on two parameters, i.e., remaining energy and probabilistic connectivity among neighboring nodes. We calculated neighborhood connectivity for mobile sensor nodes based on radio radial range. Hence, the improved selection of CL enhance remaining energy of mobile sensor nodes and it improved the life time of network. This study analyzed the proposed algorithm with LEACH-MAE, Mobile-LEACH and LEACHDistance-M using three performance parameters (NAN, RE and PDR). The proposed work outperformed other algorithms while using datasets from four mobility models with respect to different speeds.
Another contribution of this paper is selection of mobility model that is suitable for our proposed work. The results show that the performance of RPGM mobility model is better than that of other mobility models since it has higher connectivity and all the nodes move in the form of group.
To further strengthen our claim, we performed four statistical tests (difference of mean, one-way ANOVA, post hoc (Tukey’s test and LSD) and Heckman’s two-stage test. It was found that the difference of means of RPGM (considering ANAN, ARE, APDR and AC) is statistically significant in comparison with other mobility models. We verified significant difference within and between the groups of mobility models with respect to all performance parameters by applying one-way ANOVA and post hoc (Tukey’s test and LSD). It is proved through both tests that RPGM is more statistically significant within and between the groups as compared to other mobility models. Lastly, we verified through Heckman’s two-stage test that our proposed connectivity parameter is not selection biased. In addition, we found that there is no impact of another instrumental variable. Simulation results and statistical analyses suggest that RPGM mobility model is better for hierarchal clustering in MWSN. In the future, this research work can be extended by increasing the number of mobile sensor nodes and their speeds in a multi-hop environment. We are also working to make the full source code publicly available.