PPZTEM: An efficient approximate trajectory extraction method with error bound constraint for wireless sensor networks

https://doi.org/10.1016/j.comcom.2012.02.006Get rights and content

Abstract

Trajectory extraction has been studied in many research areas, including traditional spatio-temporal databases, advanced vehicle information systems, and military surveillance. In wireless sensor networks, several factors make it difficult to acquire an object’s trajectory, including imprecise and stream-oriented localized locations, limited sensor storage, and limited bandwidth. This paper proposes the Possible Presence Zone Trajectory Extraction Method (PPZTEM) with an error bound control mechanism to extract the approximate object trajectory from imprecise localized locations. PPZTEM constructs a trajectory that describes the most probable path of an object in wireless sensor networks. The constructed trajectory of PPZTEM satisfies the given error bound constraint and requires only a small amount of data. Experiments on a broad variety of synthetic and real-world object trajectories reveal that PPZTEM significantly reduces the data size of the trajectory by fusing the localized locations. At the same time PPZTEM achieves user-specified error constraints on the estimated locations.

Introduction

Advances in distributed spatio-temporal technologies and localizations in sensor devices have enabled location-based services in wireless sensor networks [1], [2], [3], such as military surveillance, mountain hiker tracking, health care patient monitoring, and wildlife monitoring [4], [5], [6]. For the location-based services, trajectory extraction is critical because it is used to determine the movement behavior of the moving objects based on the estimated locations in a wireless sensor network. For example, in wildlife habitation monitoring applications [4], [5], the trajectories of a group of wildlife can be extracted from localized locations. Then, the extracted trajectories can be sent to the laboratory. By using trajectory extraction, ecologists can study wildlife behavior remotely instead of spending time in risky outdoor environments.

Wireless sensor networks inherently have unstable data transmission, scarce resources, and frequent data collection from the physical world. Thus, in order to obtain fine-quality results, a sensor-based system has to adapt to the hardware and applications. In wireless sensor networks, extracting object trajectories has two environmental constraints:

  • Localization imprecision: This is inherent to low cost sensor devices which incur noisy sensor readings [7], [8], [9]. In localization-related literature [10], [11], [12], multiple possible object locations or an uncertain region are frequently observed in noisy sensor readings. If all possible object locations are considered, trajectory extraction will require a lot of computation to construct all possible trajectories. Thus, an efficient method of processing the imprecise locations is necessary.

  • Stream-oriented localization data: This is a major concern for trajectory extraction, since the object location data are collected periodically at a rapid speed. However, the memory of a sensor device is limited [13], [14]. The continuously arriving location data can easily exceed the size of the sensor memory. Hence, using a technique for trajectory data reduction, such as the data fusion technique [15], is critical. If trajectory data is huge, sensor nodes need to frequently swap out trajectory data to a remote server to reserve sensor memory for future data. In this case, the communication cost is large, and thus, the lifetime of the wireless sensor network is shortened. Therefore, a trajectory extraction method needs to consider the effect of continuously received localization data.

In current research on trajectory extraction, most works assume that the sensors offer precise locations of the objects, and then use these precise locations to determine the traveling paths of the objects [16]. However, in practice, when acquiring an object’s location, multiple possible object locations are output by the low-cost sensors due to noisy sensor readings [9]. This paper faces the above mentioned challenges and focuses on constructing the traveling path for a moving object under the localization imprecision effect and the stream-oriented property of localization data.

In this paper, we propose a trajectory extraction method, called the Possible Presence Zone-based Trajectory Extraction Method (PPZTEM), with an error bound control mechanism to continuously extract the most probable traveling path of an object from the imprecise locations generated from low-cost sensor devices and to describe the trajectory with a minimal data size to cope with the great amount of stream-oriented localization data. To handle multiple possible object locations, we design a structure called the durational traveling path to represent the most probable traveling path of an object, and a structure called the possible presence zone to represent all possible locations of the traveling path of an object. For processing the stream-oriented location data, PPZTEM incorporates the incoming imprecise locations for the longest duration into a possible presence zone to minimize the data size of a trajectory. Like most compression methods, if PPZTEM only pursues the goal of minimizing the trajectory data size, the trajectory with the minimal data size will be overly simplified. Such a constructed trajectory could greatly deviate from the actual traveling path of the object; it is not acceptable for applications that require high accuracy. Hence, PPZTEM devises a trajectory quality control mechanism to ensure that the constructed trajectory can satisfy a user-defined error bound constraint. The accuracy of the approximate trajectory generated by PPZTEM is acceptable for most applications.

The major contributions of this paper are summarized as follows.

  • Our work considers the environmental constraints, including localization imprecision and stream-oriented localization data, and provides a practical and efficient solution for the trajectory extraction problem in wireless sensor networks.

  • We propose a trajectory extraction method, the Possible Presence Zone-based Trajectory Extraction Method (PPZTEM), for continuously extracting a moving object trajectory by considering multiple possible object locations from the wireless sensor network.

  • We design a quality control mechanism by specifying an error constraint for PPZTEM. Hence, PPZTEM can minimize the data size of the constructed trajectory and satisfies the specified error constraint simultaneously.

  • We conducted comprehensive experiments to verify the performance of the proposed PPZTEM using a synthetic moving model and real-world GPS data of a vehicle. The results show that PPZTEM achieves a significant trajectory data size reduction with respectable accuracy of the trajectory for describing the localized locations.

The rest of this paper is organized as follows. Section 2 introduces the technology of wireless sensor networks. In Section 3, we discuss the system model for trajectory extraction in wireless sensor networks. Section 4 describes the problem statement of trajectory extraction. In Section 5, we present the design of PPZTEM. Then, we conduct performance experiments in Section 6. Section 7 concludes the paper with directions for future work.

Section snippets

Technology background

A broad range of methods have been proposed to process moving object trajectories in traditional spatial–temporal databases [17]. Most studies have focused on post-location-data analysis, management, and mining. However, in wireless sensor networks, storing stream-oriented locations is a bottleneck for a sensor system with limited storage and energy. Two issues are presented in this section, including (i) localization imprecision and (ii) streaming data processing. In this section, we also

Object-tracking sensor network

Fig. 1 shows the system model for trajectory extraction. In order to extract the trajectory of a moving object in a wireless sensor network, three key components need to work cooperatively: (i) wireless sensor network, (ii) object agent, and (iii) central server.

Problem statement

In wireless sensor networks, an approximate trajectory is extracted to describe the movement of a moving object based on the periodical localization results. As mentioned in the last section, a localization result with localization error is represented in the form of an uncertainty region. In this paper, the formal definition of the approximate trajectory is extended from the previous work [22] as follows.

Definition 4.1 Approximate trajectory

Let Si,j be the traveling of an object moving from timestamp ti to tj (position li to lj).

Overview

The objective of the Possible Presence Zone Trajectory Extraction Method (PPZTEM) is to continuously extract an approximate trajectory of a moving object, such that the approximate trajectory consumes the minimal storage while the estimated locations of the approximate trajectory satisfy a given error bound constraint. The basic idea of PPZTEM is to use a few trajectory segments, e.g., least-squares fitting segments [40], to represent a massive number of raw trajectory segments that connect

Experiment environments

The simulation in this study models a sensor network of 10,000 sensor nodes in a monitored region of 1000 × 1000 m2 [16], [41], [42]. The average distance between two nodes is 10 m. For some sensor products, e.g., [43], the communication distance could be less than 20 m in indoor environments, even less than 10 m in outdoor environments. In our previous publications that we have done the real-world experiments [44], we even found the effective localization distance is around 5–6 m. Thus, the average

Conclusions and future work

Trajectory extraction in wireless sensor networks is a critical technology. However, wireless sensor networks have several inherent hardware limitations, which lead to the localization imprecision effect and the stream-oriented location processing issue. In this paper, we proposed the Possible Presence Zone-based Trajectory Extraction Method (PPZTEM), to construct the object trajectory in wireless sensor networks. For the localization imprecision effect, we designed algorithms to maintain the

Acknowledgements

Authors thank anonymous reviewers for their valuable comments on improving the quality and presentation of the paper. This work was partially supported by National Science Council of Taiwan and National Cheng-Kung University under Grants NSC100-2221-E-006-267-MY2, NSC99-2221-E-218-035, and D100-36014.

References (48)

  • M. Vemula et al.

    Sensor self-localization with beacon position uncertainty

    Signal Process.

    (2009)
  • G. Stupp et al.

    The expected uncertainty of range-free localization protocols in sensor networks

    Theor. Comput. Sci.

    (2005)
  • X. Meng et al.

    Contour maps: monitoring and diagnosis in sensor networks

    Comput. Netw.

    (2006)
  • J. Chen, X. Meng, Y. Guo, S. Grumbach, H. Sun, Modeling and predicting future trajectories of moving objects in a...
  • Y. Yao, X. Tang, E.-P. Lim, Continuous monitoring of knn queries in wireless sensor networks, in: Proceedings of the...
  • Y. Yao, X. Tang, E.-P. Lim, In-network processing of nearest neighbor queries for wireless sensor networks, in:...
  • A. Mainwaring, D. Culler, J. Polastre, R. Szewczyk, J. Anderson, Wireless sensor networks for habitat monitoring, in:...
  • P. Juang et al.

    Energy-efficient computing for wildlife tracking: design tradeoffs and early experiences with ZebraNet

    SIGARCH Comput. Archit. News

    (2002)
  • J. Huang, S. Mishra, A sensor based tracking system using witnesses, in: Proceedings of the 25th IEEE International...
  • D. Moore et al.

    Robust distributed network localization with noisy range measurements

    (2004)
  • K. Whitehouse et al.

    A practical evaluation of radio signal strength for ranging-based localization

    SIGMOBILE Mob. Comput. Commun. Rev.

    (2007)
  • K. Whitehouse et al.

    The Effects of Ranging Noise on Multihop Localization: An Empirical Study

    (2005)
  • V. Cevher, R. Chellappa, J.H. McClellan, Gaussian approximations for energy-based detection and localization in sensor...
  • I. Akyildiz et al.

    A survey on sensor networks

    IEEE Commun. Mag.

    (2002)
  • C.-Y. Chong et al.

    Sensor networks: evolution, opportunities, and challenges

    Proc. IEEE

    (2003)
  • D. Smith et al.

    Approaches to multisensor data fusion in target tracking: a survey

    IEEE Trans. Knowl. Data Eng.

    (2006)
  • Y. Xu et al.

    Compressing moving object trajectory in wireless sensor networks

    Int. J. Distrib. Sen. Netw.

    (2007)
  • J.F. Roddick et al.

    Spatial, temporal and spatio-temporal databases – hot issues and directions for phd research

    SIGMOD Rec.

    (2004)
  • T. He et al.

    Range-free localization schemes for large scale sensor networks

  • M. Rudafshani et al.

    Localization in wireless sensor networks

  • N. Bulusu et al.

    GPS-less low-cost outdoor localization for very small devices

    IEEE Pers. Commun.

    (2000)
  • A. Savvides et al.

    Dynamic Fine-Grained Localization in Ad-Hoc Networks of Sensors

    (2001)
  • A. Galstyan, B. Krishnamachari, K. Lerman, S. Pattem, Distributed online localization in sensor networks using a moving...
  • X. Li

    Collaborative localization with received-signal strength in wireless sensor networks

    IEEE Trans. Veh. Technol.

    (2007)
  • Cited by (0)

    View full text