Localization in wireless sensor networks: A dimension based pruning approach in 3D environments

Highlights

• DPSO and HDPSO consider each dimension separately for particle deployment.

• High accuracy is attained.

• Mature convergence in proposed models reduces the execution time.

• The target fitness (f_r) in evaluation function is met in minimum number of iterations.

Abstract

Accurate localization is a key agenda for location-based services in wireless sensor networks. In this paper, two dimensionality based 3D node localization algorithms using the particle swarm optimization framework is proposed. The anchor nodes are randomly distributed in 3D search space and the proposed positioning methodology estimates the location of target node amidst of noise factors. It is also reliable in highly complex node deployments, i.e., anisotropic environments. Instead of deploying particles in terms of co-ordinate positions as followed in existing particle swarm optimization (PSO) based techniques, dimensionality based particle swarm optimization (DPSO) and hybrid dimensionality based particle swarm optimization (HDPSO) considers each dimension of the participating co-ordinate for particle deployment to attain the optimized values in individual dimension. The node deployment environment faces more on-field complexities such as radio signal irregularities, interference and path loss to calculate received signal strength (RSS). This results in estimation of inappropriate distance estimates. To overcome the disjunctive between received signal strength and distance estimates, the proposed algorithms include pre-processing factors for manipulating RSS and follows a sphere-based deployment region for deployment of particles. Both algorithms utilize less computational power and minimize the error rate in location estimation. The mature and fast convergence behavior of both DPSO and HDPSO directly reduces the computational run time. The results of two proposed techniques are compared with the existing algorithms such as PSO and HPSO in terms of time, average localization error and scalability. The models outperform well on random test cases and behave well in heterogeneous environments with increased network lifetime.

Introduction

Wireless sensor networks (WSN) have drawn major attention of all researchers due to its wide impact on various fields of engineering and technology. In recent years, it is considered as one of the vital component for a variety of applications that includes automation spectrum, health monitoring and military services. These applications serve the society by solving practical challenges using intelligent electronic components. WSN has gained the interest of all sectors that act as solution providers with proven technologies. It is composed of several sensor nodes connected either as static or dynamic links, which rely on its mode of establishment. In addition to the core functionality of gathering data by any sensor node, there also exists a set of situations that needs to be addressed with the locations of sensor nodes on consistent intervals with desired accuracy. The characteristics of WSN may vary depending upon the environment and constraints imposed over it. Although the application and environment differs, the fundamental aim of any application is to provide solution based on its current location and surrounding circumstances. Hence, location-based solutions are introduced in WSN. This caters geographical knowledge to the legitimate authority for administration of entire network.

Global positioning system (GPS) provides precise location information within accuracy range of 5–10 m. However the level of accuracy is appreciable, the process of triggering the position of sensor nodes using GPS lone is typical in all network scenarios. On the other hand, the cost of overall network with more GPS establishments (i.e., anchors) become more costlier. It also affects the goal of providing budget-optimized solutions. The primary objective of the WSN becomes meaningless when a considerable amount of energy and cost are being spent with GPS enabled nodes. The characteristics of WSN are restricted transmission range, limited energy levels and dynamic links in network. These characteristics should be taken into consideration for modeling an energy efficient network. These characteristics have laid down its way in hunt of location based units with partial GPS nodes. The need for an accurate location-based system led localization problem as one of the major concerns in the field of wireless communications. The nodes that are deployed with inbuilt GPS units are called as anchor nodes while those nodes deployed without inbuilt GPS units are termed as target nodes.

The goal of localization algorithm is to identify the exact location of any node at an instant time. Several localization techniques specified in the literature [1], differ in terms of accuracy level, error rate and computation cost. It is also categorized [1] on the basis of “sparse vs Dense”, “Anchor based vs Anchor free”, “Indoor vs Outdoor” and “Static vs Mobile”. In addition, most of the localization techniques are classified as range-based [[2], [3], [4]] and range-free [[5], [6]]. The range-based localization techniques are encouraged in all disciplines compared to the range-free techniques because the results of range-free techniques are not accurate than range-based techniques [7]. The recent study in [8] states that the nodes in range-based scenarios with inbuilt GPS units require consistent communication medium with global navigation satellite systems (GNSS) to acquire location estimates. They also emphasize on establishment of network with fault-tolerant mechanism.

The distance estimates between nodes are used as input for localization process. The process of attaining distance estimates are done by measurement techniques [9] such as angle of arrival (AOA), time of arrival (TOA), time difference of arrival (TDOA) and received signal strength (RSS). RSS is considered as a dominating parameter in majority of applications to reveal the radio connectivity information among adjacent nodes. It is calculated based on the strength of the received signal between two nodes in slotted time intervals. RSS is highly dependable on the prescribed communication channel, installed hardware units and noise parameters. Hence, these factors must be taken into consideration for evaluating the RSS values between any nodes.

The RSS values are considered as input parameters for most of the existing solutions [1] to evaluate the performance in application-specific scenario. The target nodes estimate the RSS values with their neighboring anchor nodes. The manipulation of those RSS values into distance estimates are carried out by Euclidean distance formations. Although these steps are followed in literature [10], there exists a gap in the accuracy level of distance estimate. This accuracy level is met by incorporating various soft computing techniques in the localization process. This results in optimized outputs. The co-ordinates of target nodes are optimized to the desired level of accuracy with acceptable error range that may vary based on network. The drawbacks of localization process are stated below:

• High computational cost is spent to develop an accurate localization model.

• Immature convergence is shown by evaluation models which results in faulty target node estimation.

• More execution time is required to achieve the desired target fitness (f_r).

Similarly, the existing techniques [7] have concentrated on achieving high accuracy which indirectly increases the computational cost of overall system. On the other hand, another set of optimization techniques [11] have concentrated on minimizing the computational run time that results in imprecise estimation of target node location. In order to overcome these drawbacks, an optimized algorithm that focuses on dimension based individual optimality to find the exact location of target nodes is needed in all WSNs. Therefore, two localization algorithms namely dimensionality based particle swarm optimization (DPSO) and hybrid dimensionality based particle swarm optimization (HDPSO) are presented in this paper. The contributions of this paper are as follows:

• DPSO, follows a dimension based optimization technique that takes the essence of PSO by considering each dimension to identify location of target node.

• HDPSO, follows a grouping technique with dimensionality based estimation technique to achieve fast convergence.

• The noise parameters are modeled using uniformly distributed standard values with respect to distance and environment variations. This is used for computing distance estimates between nodes.

The effectiveness of DPSO and HDPSO algorithms are tested in various deployment scenarios, where both target nodes and anchor nodes are deployed randomly in a 3D space. The performance of the proposed models does not depend upon the landscape of the problem because any wireless sensor network can be modeled into a three dimensional network scenario irrespective of its application domain. The two models work well in limited transmission range and boundless network establishments in different outdoor environments such as terrain, forests and deserts. The presented models consume minimum energy for localization process due to its fast convergence behavior. Therefore, the network has increased lifetime.

The paper is organized as follows. In Section 2, the spectrum of various existing solutions is discussed. Section 3 discusses about the prerequisites for efficient parameter refinement and the models DPSO and HDPSO in the process of localization. Section 4 portrays the discussions of proposed models with existing solutions and shows the simulation results for various test cases with efficient performance factors. In Section 5, conclusions and the possible future directions of proposed work are presented.