11 pages

A knowledge-oriented meta-framework for integrating sensor network infrastructures

of 11
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
A knowledge-oriented meta-framework for integrating sensor network infrastructures
  A knowledge-oriented meta-framework for integrating sensornetwork infrastructures Nafaaˆ Jabeur, James D. McCarthy, Xitao Xing, Phil A. Graniero  Department of Earth and Environmental Sciences, University of Windsor, 401 Sunset Avenue, Windsor, Ontario, Canada N9B 3P4 a r t i c l e i n f o  Article history: Received 4 May 2007Received in revised form6 April 2008Accepted 12 April 2008 Keywords: Sensor websSensor network integrationAgent-based softwareInteroperability a b s t r a c t In this paper, we describe a meta-framework that helps guide development of sensornetwork (SN) cyberinfrastructure in a way that enables emerging sensor infrastructures,including advances in sensor hardware, communication, monitoring applications, andknowledge representation, to interoperate. This framework is guided by the  DAST  principle . That is, the overall goal of any SN infrastructure is essentially the same: toacquire the right  Data  from the right  Area  using the right  Sensors  at the right  Time . Inconformity with this principle, our meta-framework integrates SN infrastructures alongaxes related tothe answers to five questions: Why has processing been requested? Whatare the goals of the processing? Where is it carried out? How is it carried out? And,when will the results be provided? The infrastructure components are integrated byusing various data standards and technologies currently available from various SNresearch groups, and mapping them to an overarching knowledge-based meta-framework. In concrete terms, we show in this paper how four distinct sensortechnology projects under development in our research lab are used to fit these five axesof SN infrastructure and how they can be indirectly integrated through the use of software agent-based tools, which embody the meta-framework: an ontology-baseddecision support system that applies models of SN infrastructure to its evaluationtechniques; SN configuration tools that enable network configurations to be exportedinto common geospatial standards; a transformation engine that converts these SNconfigurations, along with collected data, into a representation that meshes withour infrastructure models so that they may be used within our decision supportenvironment; and a Virtual SN to handle many of the management and controlaspects of SNs. &  2008 Elsevier Ltd. All rights reserved. 1. Introduction Interest in enhancing sensor networks (SNs) continuesto increase worldwide, but new solutions are still in theirinfancy due to several constraints, particularly limitedhardware capabilities of sensors (Schneider et al., 2005).Most of these constraints are weakening with advances inhardware and software, enabling sensors to be smaller,less expensive, more energy-efficient, and more reliable.Thanks to these advances, new approaches for developinglarge-scale SNs are becoming possible for use in a widevariety of monitoring domains, such as wildlife andhabitat, transportation, safety and security, environment,and hazards. These networks are able to collect hugeamounts of data; thousands of daily measurements fordensely instrumented sites. These data will likely beheterogeneous in nature, coming from many unrelatedsources and being stored in different formats. As such, this Contents lists available at ScienceDirectjournal homepage: Computers & Geosciences ARTICLE IN PRESS 0098-3004/$-see front matter  &  2008 Elsevier Ltd. All rights reserved.doi:10.1016/j.cageo.2008.04.006  Corresponding author. Tel.: +15192533000x2485;fax: +15199737081. E-mail addresses: (N. Jabeur), (J.D. McCarthy), (X. Xing), (P.A. Graniero).Computers & Geosciences 35 (2009) 809–819  data may not match users’ requirements since differentapplications will store, transmit, and consume data indifferent ways and under different monitoring conditions.In order to reduce resource consumption and better targetdata acquisition efforts, a new architectural approach toinfrastructure is needed where subsystems designed fordealing with a complete family of SN issues (e.g. sensors,their spatial context, occurring phenomena, and requiredtypes and formats of data and results) collaborate. Thiscollaborative infrastructure can be supported by organiza-tional frameworks built upon knowledge representationstandards.Several works have been proposed for integratingknowledge representation standards with the sensinginformation that flows from phenomena, to the SN, tothe decision-making and application domain. In thisrespect, Russomanno et al. (2005) proposed an ontologythat models the physical SN based upon the recommen-dations of the Committee on new sensor technologies(NRC,1995). Eid et al. (2006) augmented this ontology for use in processing raw data streams in order to bettersearch and query SNs. SN data may be enriched bysemantics (Wang et al., 2004) in order to improvenetworking operations. Jurdak et al. (2004) proposed aframework for modeling SNs by characterizing themaccording to some general features (e.g. topology, networksetting, data flow) and providing a set of performancegoals (e.g. power, quality of measurements). Despite thesevaluable initiatives, there is no complete and common SNinfrastructure able to blend current and future researchworks. We contend that such infrastructure must supportthe  DAST principle : provide the right  Data  from the right  Area  using the right  Sensors  at the right  Time . We believethat this goal can be met using a combination of intelligent agents and knowledge representation stan-dards. In this respect, we propose in this paper aknowledge-oriented meta-framework for integrating SNinfrastructures. We outline four distinct subsystemsdeveloped in our research lab that enable differentcomponents of a complex SN infrastructure and demon-strate how they can be conceptually connected within ourmeta-framework.In the remainder of this paper, Section 2 discusses therequirements for developing a knowledge-oriented SNcyberinfrastructure. Section 3 presents our models torepresent SNs, phenomena, measures/results, and geo-graphic space, respectively. Section 4 describes how thesemodels have been incorporated within IWTS, REASON,and ENGINE, which are SN and decision support technol-ogies we have developed. Section 5 describes how thesecomponents may be integrated and leveraged at a higherlevel via a proposed Virtual Sensor Network (VSN) or theuse of existing geospatial and Sensor Web Enablementstandards. 2. Knowledge requirements for a SN cyberinfrastructure Several interesting works have addressed specificissues of SNs spanning hardware (e.g. Ekanayake et al.,2004; Hempstead et al., 2005), network protocols (e.g. Intanagonwiwat et al., 2000; Woo et al., 2003), applica- tions (e.g. Szewczyk et al., 2004; Simon et al., 2004), and operating systems (e.g. Han et al., 2005; Abrach et al., 2003; Levis and Culler, 2002). Continuous improvements of sensors with respect to energyefficiency, reliability, andminiaturization have led to the possibility of developinglarge-scale SNs. However, progress in SN does not relystrictly on hardware and software advances. Indeed, aprimary factor currently limiting progress is the lack of anoverall SNarchitecture (Culleret al., 2005) able toimproveupon SNs’ efficiency and design (Baldwin, 2003). Acommonly agreed-upon, concrete SN architecture is notlikely in the near future, and for good reason. Much of thecurrent research is in exploratory or emergent phases,which means the ‘landscape’ of candidate architectures isheavily fragmented (Handziski et al., 2003). As such, thecommunity can greatly benefit from discussing anddeveloping new‘meta-standards’ that focus on connectingsolutions being developed for  families  of specific SN issues,e.g. modeling sensors, their behaviors with respect tophenomena, their spatial context, etc. Recent progress inknowledge representation provides data standards thatoffer favorable partial frameworks for reliable dataprocessing and exchange in a network environment.We think that any successful attempt to design andimplement a complete SN infrastructure must connect orreuse solutions that model the network, its context(spatial, temporal, and physical), constraints, stimulatingphenomena, and types of required data and results. In thisrespect, Handziski et al. (2003) identified hardwareabstraction for prototypes, characterization of powerconsumption, and protocol architecture as threeapproaches that could contribute to better reuse of existing solutions.Integrating data models and ontologies across the fullflow of SN information should enable the community tomove towards a new generation of networks complyingwith the  DAST principle : provide the right  Data  from theright  Area  using the right  Sensors  at the right  Time . In thisgeneration collaborating autonomous subsystems, builtusing existing or emerging data standards mapped toknowledge representation meta-standards, connect tocreate an infrastructure chain where each of the down-stream components can add higher value information tothe upstream components and augment their mutualcapabilities.In this paper, we propose a meta-framework thatbuilds a SN cyberinfrastructure along five intuitive axes(Fig. 1): Why has processing been fired? What are thegoals of this processing? Where is it carried out? How is itcarried out? And when will the results be provided? Wepropose models for phenomena, data/results, geographicspace, and the SN itself. The modeling of requirements,which depend upon the intended use of the SN, is beyondthe scope of this paper. Our models are connected byfour subsystems under development in our researchlab (Fig. 2). REASON serves as our decision supportenvironment, allowing phenomena and events measuredby the Integrated Watershed Telemetry System (IWTS)and managed by a VSN to be identified and analyzed. TheENGINE system ensures that data exchanged between the ARTICLE IN PRESS N. Jabeur et al. / Computers & Geosciences 35 (2009) 809–819 810  subsystems are transformed into appropriate representa-tions and delivered in the correct format. 3. Meta-framework models  3.1. Sensor network model The OGC’s SensorML model (Botts, 2005) defines thehardware components of a SN including transducers,sensors, stations, and their connections. Jurdak et al.(2004) proposed an ontology that goes beyond modelinghardware aspects of the SN. Their ontology containsinformation about the network features and performanceof the network. We converge and extend these twomodels by adding network aggregation, expanding con-nection modes, and supporting management aspects of the network (Fig. 3). There are several nested composi-tions possible within the model, whether as a collection of Stations in a Subnet, or as a collection of Instruments in a ARTICLE IN PRESS Fig. 1.  Five axes of sensor network cyberinfrastructure development. Fig. 2.  Overview and interaction of our main sensor network cyberinfrastructure components. N. Jabeur et al. / Computers & Geosciences 35 (2009) 809–819  811  Station, Multiplexer, or Multi-probe. These aggregationsmay themselves be nested within other aggregations.Our model gives more complete details about connec-tions between network components along three axes:physical, logical, and semantic. A physical connectiondefines a wired link between one component and another,for example transducers wired together in an array.A logical connection describes a less constrained, config-ured link between Stations within a given Subnet.For example, several Stations may communicate indepen-dently via cellular modems and feed through aTCP/IP-based carrier, yet are considered to operate withina single subnetwork. Semantic connections enable usto connect our Sensor Network and Geographic SpaceModels. A Station is semantically connected to anotherStation if they are collaborating to acquire observations of the same phenomenon, matching Delin’s (2002) notion of a macro-instrument. These connections may be supportedby different protocols, such as O&M, Modbus, or SWL (Nickerson et al., 2005) depending on the hardware,software, communication, and decision-support needs.Management aspects within our model governoperational responsibilities for a Station or Subnet. Theresponsibility defines a component’s current data acquisi-tion policies and introduces its authority to changethe policies of other components. The responsibility is a‘placeholder’ concept that may be extended, and its usewill vary depending upon several users’ or organizations’structures and definitions. For example, a conservationauthority may operate a complete network, but through jurisdictional agreements it may also be managed as asubnetwork by a provincial ministry. Each may havedifferent mandated data collection and reportingpolicies that must be reconciled and complied withduring network operation.  3.2. Phenomena model A model of phenomena is important to better specifyhow the SN should deal with real-world stimuli. Ourmodel (Fig. 4) is comprised of two main parts. The firstpart concerns the phenomena occurring in the world andmonitored by sensors. These phenomena can be internal(associated with the operation of the monitoring system)or external (associated with the environment beingmonitored). Internal phenomena may be related to hard-ware (e.g. battery level or failure of a sensor) or software(e.g. exceeding available memory). External phenomenamatch with the  definition  attribute of a  swe:Quantity  tagwithin the  Input   section of a SensorML specification,and the phenomenon ontology may be extended to beconsistent with an externally published definition list.The second part concerns interactions, which can wrapthe observed state of phenomena monitored by thenetwork. These interactions stimulate the network tocarry out relevant processing tasks, and create the bridgeneeded to glue the phenomena model to the other piecesof our meta-framework. In a collaborative design, actorsgive and ask for information, and this information iscommunicated using two types of interactions: Notifica-tions and Requests, each of which imply a direction to theinteraction. Requests are initiated by a client of thenetwork in order to obtain or react to the observed stateof some phenomenon. Notifications are initiated by thenetwork itself based on some request or a predefined stateof the observed phenomenon.Consider the situation where the administrator of thenetwork wishes to create a rule alerting them wheneverthe water level at a stream gauging station drops below0.6m. An example sequence of instances of the rules,alerts, and state of the observed phenomenon is shownbelow. We have implemented an OWL ontology thatrepresents these concepts, however, it is too large to showhere adequately so we have used a more compact, partialnotation to show the sequence. Event : The administrator sets the new rule via thenetwork design software: Client :  Request.Rule [123, Phenomenon.External [‘‘Wa-terLevel’’], Operator  [ o ],  State [0.6, ‘‘m’’]] SN :  Notification.Acknowledgement  [ Request  [123]]//Ack-nowledges receipt of the request SN :  Notification.Answer.Accept  [ Request  [123]]//Requestis activated in the network Event : A change in the observed water level occurswhich matches the rule: SN : Notification.Alert[Rule[123]] Client :  Request.Data [124, Phenomenon.External [‘‘Wa-terLevel’’],  Time [‘‘now’’]] ARTICLE IN PRESS Fig. 3.  A model of sensor network components and structure. N. Jabeur et al. / Computers & Geosciences 35 (2009) 809–819 812  SN : Notification.Acknowledgement[Request[124]] SN : Notification.Answer.Data[Request[124], State[0.57,‘‘m’’]]Within the operating SN, these concept instances aretransformed to messages using a protocol that matchesthe specific hardware (Section 4.1).  3.3. Measures/results model Current sensor measurement models focus largely onencoding these measurements for discovery and deliverypurposes. Some applications make use of specializeddatabases for storing sensor data (e.g. Bonnet et al.,2001). XML has emerged as the main format for deliveringmessages to applications due to the large number of existing tools for handling XML data. Encodings such asObservations and Measurements (O&M) (Cox, 2006) useXML to encode sensor observations of various types in amachine-readable fashion. These approaches tend to storenot only a measurement value, but also any associatedparameters such as unit of measurement and a procedureused to perform the measurement. This additionalinformation is enough for certain applications, but whenthere is a need to extract knowledge from the data, thisrequires knowledge about the sensor that goes beyondsimple metadata (Sen, 2004). Data consumers need toknow precisely what the measurement  represents  toensure that the data are compatible with the analysisproblem. Our Measures/Results model (Fig. 5) expressesdifferent types of observations and measurements(numeric, textual, binary, etc.) and relates them to themeasuring instrument  and  the measured phenomenon toprovide context. For example, this links informationcontained in SensorML and O&M documents in a mannerneeded by a monitoring application to perform knowl-edge-based problem solving.  3.4. Geographic space model Once the phenomena and the required measurementsare identified and analyzed, the SN can reduce its resourceconsumption by focusing its sensing efforts on areas of interest. These areas are identified according to theircontents (e.g. targeted data, sensors) and locations (e.g.position regarding targets, communication pathways). Inorder to speed up their localization, we model thegeographic space (Fig. 6) along two axes: (1) dividing thespace using a spatial grid; and (2) dividing the contents of space into objects and groups of objects. The spatial griddivides the space into indexable tiles whose sizes dependonthedensityofsensorsaswellasthescaleoftheobservedphenomena. Categorization of space into objects identifiesnotable features of the landscape. An object, which mayspan more than one tile, can be identified by its geometric,biophysical,and semantic attributesaswellasitssensitivity(how much the object may be affected by or affect currentdata collection or phenomenon tracking).Semantics, like sensitivity, are a subjective notion thatdepends upon the intended use of the network. Forexample if a contaminant is detected in a stream, amonitoring or decision support system could query forsources on the adjoining or upstream hillslopes that mayrelease this contaminant. A system could also locate anyobject considered as sensitive for this contaminant (e.g.lakes or habitats). According to their importance, objectsmay be grouped into thematic layers (e.g. lakes orbuildings) or layers of interest. A layer of interest ( Jabeur,2006) maycontain objectsfromseveral thematic layers buthave the same importance. In addition to our currentmodel, several geographic space models (e.g. Gbei et al.,2003; Cecconi, 2003; Be ´dard and Bernier, 2002; Jabeur,2006) can be integrated and extended to better supportusers’ requirements. In particular, extensions may resultfrom deeper treatment of sensitivity and semantic notions. 4. Operational infrastructure 4.1. Sensor network: IWTS  The Integrated Watershed Telemetry System (IWTS)enhances current conventional telemetry technology in ARTICLE IN PRESS Fig. 4.  A model for sensor network phenomena. N. Jabeur et al. / Computers & Geosciences 35 (2009) 809–819  813
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!