A hybrid clustering and evolutionary approach for wireless underground sensor network lifetime maximization

Abstract

Wireless Underground Sensor Networks (WUSNs) have attracted significant interest in recent years because of their applications in various fields. The major difference between WUSNs and terrestrial wireless sensor networks is that their signals travel through multiple layers: soil, air and a medium interface. As communications in heterogeneous channels result in more transmission loss, a solution is to deploy relay nodes to relay traffic from sensors to base stations/sinks. However, this poses several new challenges, including load balancing and transmission loss minimization in heterogeneous environments. This paper considers the problem of deploying relay nodes to prolong network lifetime under load balancing constraints. This problem can be formalized using a Mixed Integer Linear Programming model as a basis to achieve lower bound solutions. We show that the problem is NP-hard as it can be reduced to the Set Cover Problem. Two novel methods are proposed. The first is a hybridisation of a clustering heuristic and an exact algorithm using a maximum flow with min-max cost formulation. The second is an evolutionary approach to further improve our initial results. Experimental validation on a large set of benchmarks indicates that the proposed methods perform better than the existing methods.

Introduction

A wireless underground sensor network includes sensor nodes buried completely at a certain depth [3]. These nodes gather various information from the environment, which is transmitted to one or multiple aboveground data sinks via wireless signals like radio or infrared waves.

WUSNs have many real-life applications. Akyildiz et al. [3] proposed a taxonomy of WUSNs applications, which are divided into four classes. The first class of applications is environmental monitoring. This is often seen in agriculture to control irrigation and monitor underground soil conditions such as water and mineral content [3], [23], [31], [32]. The second class is monitoring infrastructure including pipes, electrical wiring, and liquid storage tanks as well as military applications such as minefield monitoring [3], [16]. The third class is object localization. Many civilian and military applications need to identify objects in a specific area, such as locating people in the event of a building collapse or patients in hospital [3], [9]. The last class is border patrol and security monitoring, where WUSNs can be used to monitor the above ground presence and movement of people or objects [3].

WUSNs have many characteristics that are different from other types of wireless sensor networks. Because sensors are buried underground, the underground wireless channel is influenced by soil properties such as volumetric water content, particle size, density and temperature. Moreover, signals are attenuated when transmitting through multiple environments, primarily caused by the absorption, reflection and refraction at the surface. There are three different communication channels in WUSNs including underground to underground channel (UG-UG), underground to aboveground channel (UG-AG) and aboveground to aboveground channel (AG-AG) [1], [2], [4]. The propagation characteristics of electromagnetic waves in soil are significantly different from those in air. The attenuation in soil depends on several environmental parameters, such as the soil composition in terms of sand, silt, and clay fractions, the bulk density, the particle density, and the volumetric water content. In addition, EM waves encounter much higher attenuation in soil, which severely hampers the communication distance. This is the most important challenge for WUSNs’ lifetime. Network lifetime is the time span from deployment to the instant when the network is considered non-functional. The concept of non-functional is, however, application-specific. In [22], there were three definitions of topological lifetime based on the criticality of a mission:

• $N-of-N$ lifetime: the mission fails when a node in N nodes run out of energy,

• $K-of-N$ lifetime: the mission fails when $N-K+1$ nodes in N nodes run out of energy,

• $m-in-K-of-N$ lifetime: the mission fails when m supporting nodes in N nodes are alive and $N-K+1$ nodes in N run out of energy.

In this work, we only consider problems in the first definition ($N-of-N$ lifetime), which seeks to maximize the lifetime for all network nodes. For example, for applications in earthquake prediction, underground sensor nodes (SNs) are deployed in a large area to track changes in soil properties. If a single sensor shuts down, the collected information is no longer accurate and the whole network becomes non-functional.

In WUSNs, sensor nodes send their data towards base stations or sinks, whose locations are usually fixed. However, sensors are often located far away from their base stations/sinks, and can rapidly exhaust energy. Moreover, it is often impossible to recharge or replace the sensors’ power sources after deployment. Consequently, power-conservation is one of the most prioritized tasks in WUSNs. A viable method is to deploy a layer of relay nodes, positioned close to underground sensors as to reduce the transmission loss from the sensors to the sink. Relay nodes play an important role in the effective and successful deployment of WUSNs. In wireless sensor networks, relay nodes are used for maximizing the network lifetime, energy-efficient data gathering, load-balanced data gathering as well as making the network fault-tolerant [28].

In this paper, we study the problem of relay node placement optimization to prolong the lifetime of wireless sensor networks. The contributions of this paper can be summarized as follow:

• Firstly, we present the network model, energy model, and mathematical model of the problem of minimizing the maximum transmission loss between a sensor-relay pair;

• Secondly, we prove that the problem is NP-hard as it can be reduced to the Set Cover Problem [8];

• Thirdly, we formulate the problem into a mixed integer programming model as a basis to achieve bound solutions. However, the exact Mixed Integer Linear Programming (MILP) solvers have exponential time complexity and do not scale well to large instances;

• Finally, we propose two methods to solve the Network Lifetime Optimization (NLO) problem. In the first method, we propose a new clustering-based heuristic for the Load Unrestricted Relay Node Selection (LURNS) problem. We then reformulate the Load Balanced Sensor Node Assignment (LBSNA) problem as a Maximum Flow with Min-max cost (MXF-MC) problem, which can be solved in polynomial time. The output for MXF-MC will satisfy three requirements: uniquely connected sensor-relay pairs, load-balanced relays and maximum network lifetime (for the sensor nodes). Based on the first method, we propose a evolutionary algorithm-based method for further improvement.

The rest of the paper is organized as follows. In Section 2 we review related works. Section 3 is for problem definition and models. Our main contributions will be shown in Section 4 and Section 5. Section 6 presents our evaluations of the proposed methods on benchmark datasets. Finally, we conclude our work and present future works in Section 7.