Sink placement on a 3D terrain for border surveillance in wireless sensor networks

Abstract

With their distributed nature and redundant operation capability, wireless sensor networks are very suitable for border surveillance scenarios that track intruders trying to breach to a safe side. In such scenarios, keeping the operation going on for as long as possible is the most important aspect of the network. We propose that by placing sink at a carefully selected coordinate will results in a longer living network. We also place restrictions on the candidate locations so that the sensing quality of the network is above a useful predetermined value and the sink is placed in a relatively safe location to avoid destruction. In order to find the suitable coordinates we propose a modified lifetime metric which takes quality and safety measures into account. We also propose a genetic algorithm which uses a discrete event simulator-in-the-loop over a three dimensional terrain to find locations for the sink that fits the given quality and safety restrictions. Using a three dimensional underlying terrain makes the proposed approach more realistic. The results obtained for various sensor network scenarios indicate that the proposed algorithm can find locations that increase the lifetime by also considering the sensing quality and safety.

Introduction

Due to their distributed operation ability and scalable coverage ranges, wireless sensor networks (WSNs) are very suitable for sensing different phenomena over large areas (Akyildiz et al., 2002). Composed of many sensing/communication units and mostly one data gathering unit called a sink, those networks are dispersed over the area to perform the given sensing tasks for as long as possible (Yick et al., 2008). The key steps of the whole life line of a sensor network can be listed as:

• Deployment of the sensors over the area.

• Initial discovery and communication between nodes, network formation.

• Start of environment sensing.

• In case of an event detection, related data is sent to the sink.

• (If needed) Routing information update.

• Network announced dead when sensing or communication quality falls below a certain threshold.

The distributed and automated nature of a WSN makes it tolerant against partial losses of sensing units. This demonstrates WSNs an increasingly suitable and robust candidate for security operations such as border surveillance. In a border surveillance application, it is of the highest importance to detect any intruder trying to reach the secure site by passing through the border, as shown in Fig. 1. Nodes of a WSN are deployed in order to properly cover the whole border area, as better coverage will help detecting any kind of intrusions that occur within that border.

WSNs should be designed to operate with as little human intervention as possible due to the deployment characteristics. Sensors are deployed to unreachable areas such as high mountain terrains, underwater fields or not easy to access areas such areas flooded with land mines or poisonous sites like volcanic craters (Stoianov et al., 2007, Werner-Allen et al., 2006). Sensors can be deployed to such sites by means of planes or submarines. Both the cost and the characteristics of the deployment necessitate mass deployments and prohibits fine tuning operations.

Being mostly battery powered and operationally autonomous WSNs suit to the requirements of such deployments, yet those facts again limit the lifetime of the network (Sadler, 2005). The sensor nodes in time use up their battery capacities one by one, causing outages within the network and the overall network finally dies. Moreover, the sensing quality for especially surveillance schemes must be kept over a certain threshold to meet the required operational quality at any time.

There are various challenges associated with the WSNs. Sensor nodes are battery powered and mostly impossible to recharge causing energy limitations. Due to the node architecture and battery usage, the communication ranges of the sensors are limited (Raghunathan et al., 2002). Most nodes cannot reach the sink node directly and routing over other nodes is inevitable. This fact increases the energy burden on the intermediate routing nodes, because the message receiving and submitting operations are very energy consuming and for every operation there is unavoidable energy loss which cannot be lowered without changing the electronic architecture of the nodes. The energy terms for a node is a total of the energy to sense, to receive and to transmit (Bhardwaj and Chandrakasan, 2001). The transmission energy term depends on the distance and the medium, whereas other terms have fixed values. Transmission energy depends on the path loss value, a physical property of the medium and is proportional to the path loss exponent and the distance. Very long communication distances cause high energy consumptions, yet increasing the hop count will increase the receiving costs in the network as the number of sensors to receive the traffic will also increase. For three dimensional terrains, obstructions can also cause communication and sensing disruptions.

Given those facts, it is very crucial to optimize the network parameters as much as possible to increase the sensing quality and the network lifetime. Many approaches have been proposed to optimize the information routing from sensors to the sink, optimize the radio communication between sensors, decrease the battery expenditure by sleep schedules (Al-Karaki and Kamal, 2004, Kahn, 1999). One other approach is to adjust the coordinates of the sink nodes such that the overall energy loss due to the routing is minimal. Such a problem is called the “Sink Location Problem”. It can be defined shortly as “Given a group of sensors, whose locations, sensing and communication ranges and capacities are known, finding the coordinates of the sink that maximize the lifetime of the network during which a certain performance criteria is always met.” It can be formulated as follows with the parameters defined in Table 1:

Problem P:

$$\max t$$

such that

$$C(t)>C_R$$

where

$$C(t)=\bigcup _{i,y_i(t)>0,E_i(t)>0}{C}_i$$

$$C_i=\bigcup _{(x,y)\in (W,H),\mathit{dist}((x,y),i)\le {\mathit{RS}}_i}(x,y),\forall i$$

$$y_i(t)=\left\{\begin{array}{cc}1\hfill & \mathrm{if}\mathrm{node}i\mathrm{is}\mathrm{connected}\mathrm{to}M(x,y)\mathrm{at}\mathrm{time}t\hfill \\ 0\hfill & \mathrm{otherwise}\hfill \end{array}\right.$$

$$E_i(t=0)=B_i,\forall i$$

$$E_i(t+1)=E_i(t)-{\mathit{ES}}_i(t),\forall i$$

$${\mathit{ES}}_i(t)={\varepsilon }_i^S+\sum _{i\in A_j}({\varepsilon }_{\mathit{ij}}^R⁎f_{\mathit{ji}}(t))+\sum _{l\in A_i}({\varepsilon }_{\mathit{il}}^T⁎f_{\mathit{il}}(t)),\forall i$$

$$\sum _{j\in A_i}{f}_{\mathit{ij}}(t)=\sum _{k\in A_i}{f}_{\mathit{ki}}(t)+p(t),\forall i$$

$$A_i=\bigcup _{j\in N,\mathit{dist}(i,j)\le {\mathit{RC}}_i,y_j(t)>0,E_j(t)>0}j,\forall i$$

Here we maximize t, the lifetime of the network. Eq. (1.2) shows that the coverage of the network stays above a threshold value for all times. Eq. (1.3) explains the coverage as the union of areas covered by sensors which are connected to the sink node and still have energy. Eq. (1.4) defines the neighbours of a node as the nodes that are inside the communication range of the network and are connected to the sink node. Eq. (1.5) defines a Boolean value which implies whether a node at a given time is connected to the sink or not. Eq. (1.6) shows that the initial battery capacities are the same for all sensors. Eqs. (1.7), (1.8) show the energy expenditure at each time step. In each step, total energy spent is the sum of sensing operations, receive operations from nodes that this node is one of the neighbours and sensing operations to the nodes that are neighbours. Eq. (1.9) shows the flow of information from and to a node. Eq. (1.10) defines the coverage area of sensor as the field points within the sensing range of a node.

This problem is somewhat similar to the “Facility Location Problem”. Given a group of demand nodes, a place for the source node is found such that the overall communication (or transportation) cost is minimized (Balinksi, 1996, Drezner and Hamacher, 2002). Facility location problem specifies that the demand nodes can directly communicate with the source node (e.g., a customer can go directly to a supermarket). Because of this fact, the sink location problem differs significantly from the facility location problem. In sensor networks, intermediate nodes lose energy due to data routing operations. This requirement adds one more level of complexity to the problem. The coverage of the sensor networks should be kept above a threshold to perform the required duty. Also, the sink location should be chosen among safe candidates to decrease the risks associated with the nature of surveillance tasks.

The rest of the paper is organized as follows. In Section 2, we analyse other works that are related to our approach showing similarities and differences. In Section 3, we present our modified genetic algorithm by explaining the steps and the decisions made for each step. Section 4 present the results by comparing the algorithm to other heuristics for different parameters on both flat and three dimensional deployment sites. Section 6 summarizes the presented approach and draws conclusions.