2011 12th IEEE International Conference Mobile Data Management 2011 12th International Conference on on Mobile Data Management

Mobile Sensor Databases

Hoyoung Jeung Distributed Information Systems Laboratory ´ Ecole Polytechnique F´ed´erale de Lausanne (EPFL), Switzerland hoyoung.jeung@epfl.ch I. I NTRODUCTION

It then clearly lays out the concept of MSD, including data models, brief description for popular query types, and the optimization space in MSD. The introduction also provides an overview of the key research challenges that will be discussed at the following phases.

Sensors are becoming increasingly mobile, facilitated by technical advances in wireless sensing, positioning, and mobile devices. This sensor mobility has created significant benefits for a wide range of applications. Mobile sensors, for example, enable low-cost monitoring systems that can harvest high-resolution data over broad sensing areas covered by only small numbers of movable sensing toolkits. In addition, sensors installed on navigable objects, such as unmanned aerial vehicles and robot fish, allow us to measure physical phenomena over unapproachable areas. This rapid growing use of mobile sensors is also increasing the need for our capability to manage large volumes of data produced by mobile sensors, so-called mobile sensor databases (MSD). Unlike classic sensor databases or moving objects databases, MSD include positional data as well as timestamped sensor readings. As a result, it is very challenging to effectively store, index, query, and mine MSD, since MSD exhibit various data characteristics simultaneously, i.e., multi-dimensional time series, continuously streamed to the system while moving. The main objective of this advanced seminar is to provide a comprehensive and clear overview of the concepts and core research challenges in mobile sensor databases at this point in time. To this end, this seminar starts by introducing emerging applications and projects that produce and manage MSD. We then discuss about the state-of-the-art techniques and their limitations for the key research areas in MSD: data models, querying and indexing, data mining, and uncertainty management.

Phase 2: Query Processing This part of the seminar presents primary query types typically considered in MSD, such as stream, range, continuous, and join queries. It then focuses on the techniques for processing the queries, along with an emphasis on modelbased sensor data approximation [4] and processing. We also highlight the reasons why the complexity of query processing in MSD increases, compared to typical query processing mechanisms in sensor databases [5] and moving objects databases [6]. In addition, the seminar presents several noteworthy stateof-the-art indexing methods [7], [8] proposed in the literature. This part then reveals the fundamental restriction of the access methods, which applications may face when they attempt to index MSD using such techniques. Phase 3: Data Mining Mobile sensor data mining concerns discovering meaningful patterns from MSD. The literatures in trajectory data mining [9], time series analysis [10], data stream mining [11], and activity recognition [12] offer a rich body of useful concepts and tools to extract interesting knowledge out from MSD. This part of the seminar introduces such mining concepts and approaches that can potentially serve as bases in mining MSD. At the same time, the seminar also discusses the limitations of the existing approaches for covering a variety of data analytic tasks in MSD—while MSD include various data characteristics, those approaches generally consider only certain parts of the data characteristics.

II. S EMINAR O UTLINE The seminar consists of the following four phases: Phase 1: Introduction We begin with motivating the need for mobile sensor data management by introducing several real-world mobile sensor applications, such as the OpenSense project [1], the CarTel [2] project, and the Nokia data collection campaign [3]. This covers relevant backgrounds in sensor networks and positioning systems, and reviews specific requirements for data-intensive mobile sensor applications. Unrecognized Copyright Information 978-0-7695-4436-6/11 $26.00 © 2011 IEEE DOI 10.1109/MDM.2011.34

Phase 4: Uncertainty Management One typical characteristic of sensor data is in uncertain and erroneous nature, originating from various sources, such as discharged batteries, network failures, and imprecise readings from low-cost sensors. This uncertain nature of data becomes even more noticeable in MSD, because sensor 1

mobility often causes physically unstable conditions for sensing. This part of the seminar discusses the main reasons for uncertainty in MSD. It then introduces existing approaches to deal with uncertain sensor data, including uncertainty data models, queries [13], data-cleaning methods [14], and data quality measures [15] and quality-aware sensor data processing.

[7] S. Chen, C. S. Jensen, and D. Lin, “A benchmark for evaluating moving object indexes,” PVLDB, vol. 1, pp. 1574– 1585, 2008.

ACKNOWLEDGMENT

[10] R. Shumway and D. Stoffer, Time Series Analysis and Its Applications. Springer-Verlag, New York, 2005.

[8] Y. Cai and R. Ng, “Indexing spatio-temporal trajectories with Chebyshev polynomials,” in SIGMOD, 2004, pp. 599–610. [9] H. Jeung, M. L. Yiu, X. Zhou, C. S. Jensen, and H. T. Shen, “Discovery of convoys in trajectory databases,” PVLDB, vol. 1, no. 1, pp. 1068–1080, 2008.

This work was supported by the European Commission in the PlanetData NoE (contract nr. 257641), the Nano-Tera initiative (http://www.nano-tera.ch) in the OpenSense project (reference nr. 839-401), and NCCRMICS (http://www.mics.org), a center supported by the Swiss National Science Foundation (grant nr. 5005-67322).

[11] M. M. Gaber, A. Zaslavsky, and S. Krishnaswamy, “Mining data streams: a review,” SIGMOD Record, vol. 34, pp. 18–26, 2005. [12] J. A. Ward, P. Lukowicz, and H. W. Gellersen, “Performance metrics for activity recognition,” ACM Trans. Intell. Syst. Technol., vol. 2, 2011.

R EFERENCES [1] K. Aberer, S. Sathe, D. Chakraborty, A. Martinoli, G. Barrenetxea, B. Faltings, and L. Thiele, “Opensense: open community driven sensing of environment,” in GIS-IWGS, 2010, pp. 39–42.

[13] R. Cheng, D. V. Kalashnikov, and S. Prabhakar, “Evaluating probabilistic queries over imprecise data,” in SIGMOD, 2003, pp. 551–562. [14] E. Elnahrawy and B. Nath, “Cleaning and querying noisy sensors,” in WSNA, 2003, pp. 78–87.

[2] B. Hull, V. Bychkovsky, Y. Zhang, K. Chen, M. Goraczko, A. Miu, E. Shih, H. Balakrishnan, and S. Madden, “CarTel: a distributed mobile sensor computing system,” in SenSys, 2006, pp. 125–138.

[15] A. Klein and W. Lehner, “Representing data quality in sensor data streaming environments,” Journal of Data and Information Quality, vol. 1, no. 2, pp. 1–28, 2009.

[3] N. Kiukkonen, J. Blom, O. Dousse, and J. Laurila, “Towards rich mobile phone datasets: Lausanne data collection campaign,” in International Conference on Pervasive Services (ICPS), 2010. [4] J. Considine, F. Li, G. Kollios, and J. Byers, “Approximate aggregation techniques for sensor databases,” in ICDE, 2004, pp. 449–. [5] P. Bonnet, J. Gehrke, and P. Seshadri, “Towards sensor database systems,” in MDM, 2001, pp. 3–14. [6] R. H. G¨uting, M. H. B¨ohlen, M. Erwig, C. S. Jensen, N. A. Lorentzos, M. Schneider, and M. Vazirgiannis, “A foundation for representing and querying moving objects,” TODS, vol. 25, pp. 1–42, 2000.

2