A hybrid wireless sensor network framework for range-free event localization

Abstract

In event localization, wireless sensors try to locate the source of an event from its emitted power. This is more challenging than sensor localization as the power level at the source of an event is neither predictable with precision nor can be controlled. Considering the emerging trend of long sensing range for cost-effective sensor deployment, locating events within a region much smaller than the sensing area of a single sensor has gained research interest. This paper proposes the first range-free event localization framework, which avoids expensive hardware needed by the range-based counterparts. Our approach first develops a sensing range model from the statistical information on the emitted power of a type of events so that user-defined event-detection quality can be provisioned using a minimal network of static sensors. Then an accurate event location boundary estimation technique is developed from the sensing feedbacks, which also facilitates guided expansion of the area of possible event location (APEL) to deal with sensing errors. Finally, user-defined event-localization quality guarantee is provisioned cost-effectively by inviting mobile sensors on-demand to target positions. Analytical solutions are provided whenever appropriate and comprehensive simulations are carried out to evaluate localization performance. The proposed event localization technique outperforms the state-of-the-art range-based counterpart (Xu et al., 2011) in realistic environment with path loss, shadow fading, and sensor positioning errors.

Introduction

Wireless Sensor Networks (WSNs) have been established as a potential technology to facilitate monitoring events or environmental phenomena, tracking moving objects, and emergency response. Sources of many events emit electromagnetic or acoustic or radiation signals that can not only facilitate their timely detection but also allow precise localization of the events by jointly utilizing signal measurement from a set of collaborative and distributed signal sensors. For many application scenarios, merely reporting detection of an event, without precise information about location of the source, makes little sense. Therefore, event localization (also referred to as source localization) has attracted significant research interest since the introduction of WSNs. When sensors are not equipped with any accurate positioning system or due to the random deployment nature their position is unknown, localization of sensors by explicit beacon transmissions is a precursor to effective event localization. Sensor localization, a well-researched area, is privileged as the power level of the beacon signal can be known a priori and effectively controlled to improve accuracy.

Motivation. Firstly, event localization is considered more challenging as the power level of signal at the source of an event is neither predictable with precision nor can be controlled. Since event localization is essentially triggered only after an event is detected, the quality of event detection is paramount. Existing event localization techniques assume that the event sensing radius is already known without considering the consequent event detection accuracy. The existing works invariably assume that an event occurs at a fixed intensity. In that case, there is no estimate of the probability at which an event with lower intensity remains undetected. As the emitted power of an event type can be modelled with its probability density function (pdf), it is desirable that users should be able to dictate some event detection accuracy guarantee level for sensors with known sensitivity. Considering the rapid improvement in sensing hardware, it is natural that future WSNs will have more cost-effective sensor deployment by leveraging much longer sensing range. Thus the importance of understanding the impact of sensing radius model on detection accuracy is imperative.

Relevant contributions: This paper presents a comprehensive analysis on this important issue by developing a stochastic model of event sensing range by duly considering sensitivity of sensors, decay in received power due to path loss, shadowing effect, and user-defined event detection accuracy guarantee. This novel sensing range model can be used by any localization and coverage techniques. The primary objective of developing this model is to provision user-defined event-detection quality cost-effectively by finding the maximum sensing radius such that an event of known intensity pdf can be detected within the desired probability threshold using as few static sensors as possible. Our analytical model considers ideal sensor deployment on regular grid to achieve the necessary single-coverage detection network with minimal sensing disk¹ overlapping. We have also carried out comprehensive simulations in realistic scenarios with imperfect sensing disks, due to significant signal fading and sensor positioning error, to confirm that the detection quality guarantee can still be achieved by taking advantage of the sensing disk overlapping (see Section 6.2).

Secondly, sensor and event localization also differ significantly from the application point of view. In sensor localization, location of a sensor is estimated to a point which is then used for future references such as geographic routing and target-based query processing. In event localization, location of the event is needed for immediate response and thus, it is more appropriate to estimate the possible boundary of the event to facilitate the response team. All the existing event/source localization or relevant moving-object tracking techniques [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14] are “range-based” as detecting sensors try to approximate the location of an event to a point using received signal strength (RSS) (i.e., event intensity) or time of arrival (TOA) or angle of arrival (AOA) information at the sensors. Among them TOA information is relatively less sensitive to environmental noises such as path loss and shadow fading; however, it requires synchronization of clocks, which can be avoided by taking pair-wise difference of TOA information. Most of the ranged-based event localization techniques, therefore, use such time difference of arrival (TDOA) information instead of using TOA information directly. Recently, by showing that TDOA approaches inadvertently strengthen the impact of environmental noise by 3 dB, Xu et al. [1] proposed using TOA information directly and tackling clock synchronization errors by modelling them as Gaussian additive noise. Nevertheless, high-precision expensive hardware requirement has always been the major drawback of range-based techniques [15]. In addition, localization quality of these techniques is observed highly sensitive to sensor position errors that are unavoidable in real scenarios due to physical obstacles and/or using imprecision positioning system. Consequently, the calculated event location point has to be relaxed with an error-bounded circular region (ECR) centred at that point. Moreover, if the search party fails to locate the event in the estimated ECR, the search region has to be expanded omni-directionally, resulting in unguided search.

Alternative “range-free” localization techniques that have been successfully used for sensor localization are expected to be significantly less sensitive to sensor position errors as they avoid explicit point-to-point distance approximation to pin-point an event and instead confine the location within the innermost intersection region (IIR) from the constraints of sensing coverage. Note that overlapping sensing disks of two or more sensors form a set of non-overlapping sectors that we term here as IIRs. Range-free techniques also require less expensive hardware due to much relaxed precision and no synchronization requirements. Event localization using range-free techniques is an open problem, which has not been investigated so far.

Relevant contributions: In this paper, we develop a range-free event localization technique by accurately defining the IIR boundary from the emitted signal sensing feedbacks, which effectively uses the same sensing range model we have developed for guaranteeing event-detection quality. Our novel IIR boundary estimation algorithm uses arc-coding, similar to binary sensing [16], [17] used for target tracking, to achieve high accuracy of IIR area estimation, low computational complexity, and an effective neighbourhood IIR ordering to facilitate guided search (elaborated in Section 4) had the sensed IIR differ from the IIR containing the event due to imperfect sensing feedback. The proposed guided expansion strategy is a significant advantage over the existing event localization techniques considering that the failure rate of initial localization result will be reasonably high due to the probabilistic nature of event intensity information. A comparative performance analysis on this aspect is performed against the state-of-the-art range-based source localization technique [1].

Thirdly, significant improvement in event localization accuracy, in terms of the area of the IIR confining an event, demands overlapping sensing areas from a large number of sensors. This is quite analogous to k-coverage problem where sensor density for reasonably large k remains prohibitive with only static sensors. High-density static WSN also suffer more from sensors’ positional-drift [19] when deployed in an active environment. Moreover, just like event-detection accuracy guarantee, it is desirable to provision event-localization accuracy guarantee as well for which static WSN are not flexible enough. Note that these user-defined guarantees are quite natural in this context where type of events will significantly influence these quality metrics. In the context of multiple coverage, [20], [21], [22], [23], [24], [25], proposed a viable alternative with the aid of a few mobility-enabled sensors coupled with a fixed low-density static WSN to improve scalability.

Relevant contributions: In this paper, we adopt this hybrid WSN (HWSN) approach to adaptively accommodate user-defined event-localization accuracy guarantee. In HWSN, static sensors constitute the base single-coverage network to trigger the detection of an event first and then multiple mobile sensors are invited to move closer to the detecting static sensor on-demand to provide multiple overlapping sensing disks to guarantee given localization accuracy. There are many potential application scenarios of the proposed HWSN-based event localization. During natural disasters, when normal communication infrastructures fail, it has been proposed to use unmanned aerial vehicles (UAVs) for monitoring critical events such as leakage of radioactive materials [31], [32], lost ocean vehicle search, military target acquisition such as landmine or explosive detection, industrial facility inspections, remote temperature sensing, surveillance, environmental phenomena such as arsenic outburst monitoring, etc. If several long-range static sensors were deployed with single coverage in the wide region where the nuclear reactors are located in Japan, the faults caused by Tsunami attack in power plants and resulting increase in temperature could have been detected early. After initial detection of an event, mobile sensors could arrive near the static sensor that triggered the detection to assist defining a much smaller region to be exhaustively searched for the source of fault. Recent calamities such as the tsunamis in Japan and Indonesia and the hurricanes in USA have led to exploration of alternative technologies to meet disaster-time monitoring needs. In these scenarios, where the events are static and persistent for some time, HWSN facilitates efficient use of a small number of mobile sensors to cost-effectively achieve high localization or detection accuracy.

While HWSN are clear choices for demand-driven event surveillance with quality of detection and localization guarantees similar to quality of service guarantees in networking, their wide applications depend on how successfully we can address some of the impending optimization issues for efficient and cost-effective deployment of expensive mobile sensors having the cost of traversed distance, controlling overhead, and response delay. From the efficient mobile sensor deployment point of view, we formulate target positions for a given number of mobile sensors so that the minimum possible event localization area is achieved. In order to solve this optimization problem to localize a random event occurring at any location with uniform probability, estimation of the expected area of possible event location is needed to assess localization quality in this context, which motivates us to accurately compute the areas of all IIRs using the proposed arc-coding based IIR boundary estimation. This insightful localization quality metric is then formally defined, a theoretical lower bound is established, a heuristic to find the target positions of a given number of mobile sensors, and some closed-form relations of the quality metric with the number of sensors and the sensing radius are derived to improve scalability. This solution is then extrapolated to provision a user-defined event localization quality guarantee by suggesting the minimum number of mobile sensors.

To evaluate the proposed event localization technique, we have carried out comprehensive simulation investigating the impact of possible environmental noise and positional error.

Key findings. (i) Under reasonably high environmental noise (zero-mean log-normal with standard deviation (std-dev) 4 dB) and high sensor position error (zero-mean Gaussian with std-dev 25% of sensing radius), the proposed range-free technique can localize a random event of specific nature, for which past events’ intensities follow Gaussian distribution with std-dev 20% of the mean, within an expected area no more than 20% of the event detection area with as few as seven mobile sensors, which is almost half of the search area possible with the state-of-the-art range-based technique having comparable sensor density; (ii) the proposed technique is capable of provisioning user-defined event detection accuracy as well as localization quality guarantees for such random events, a highly desirable aspect comparable to quality-of-service guarantee in communications; and (iii) localization quality degradation due to the proposed stochastic modelling of sensing radius and target position errors of mobile sensors are insignificant compared to that caused by environmental noise such as shadow fading in the proposed technique.

The rest of the paper is organized as follows. Section 2 discusses the system model. Section 3 presents a stochastic event sensing range model. Our proposed localization technique is explained in Section 4. Optimization strategies for event localization in the context of HWSN and movement optimization of the mobile sensors are discussed in Section 5. Section 6 presents performance of the proposed schemes with simulation results. Relevant research works are discussed in Section 7 and finally Section 8 concludes the paper.