1 / 17

MWM: Map-based World Model for Wireless Sensor Networks

MWM: Map-based World Model for Wireless Sensor Networks. Abdelmajid Khelil , Faisal Karim, Brahim Ayari, Neeraj Suri. AUTONOMICS’ 08, Turin, Italy. World model @ Sink. How to convert raw data into information?. World model @ Network.

Download Presentation

MWM: Map-based World Model for Wireless Sensor Networks

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. MWM: Map-based World Model for Wireless Sensor Networks Abdelmajid Khelil, Faisal Karim, Brahim Ayari, Neeraj Suri AUTONOMICS’ 08, Turin, Italy

  2. World model @ Sink How to convert raw data into information? World model @ Network Wireless Sensor Networks (WSN): Bridge to Physical World Alarm Users, Admins.. User info. Event Query Sink: If report(s) received  fire  notify user Else: no fire Sink App. info. Update model Query model Sensor nodes: If (avg)temp > threshold report fire Else: no report Example: Detect forest fires Sensor network Create model Change world Deploy wireless battery-powered nodes with temperature sensors Raw data Raw data Raw data Physical world .. Independent from raw data, application and users!

  3. Three Main System-level Design Paradigms • WSN as Network • Inherent node redundancy  Convergecast, filtering • Limited resources • Cross-layer • WSN as Database • Query dissemination • In-network aggregation • E.g. tinyDB • WSN as Event Service • Nodes provide/consume services • E.g. pub/sub WSN Query Result Abstraction level  These paradigms still address single sensor nodes and ignore spatial correlation of sensor readings  less accepted

  4. Problem Statement and Objectives • Widely Accepted Abstraction Level is Needed • How to convert sensor data into information which is: Understandable, contextual, interactive and actionable. • Abstraction Should Consider • Inherent spatial correlation of sensor readings (Inherent node redundancy in WSN) • Requirements • Generalized • Unified incorporation of • Physical world and • Network world • Frugal and lightweight (creation, management etc.)  Our Approach: Map-based World Model (MWM)

  5. Outline • System Model • Map-based World Model • Design Methodology • Two Case Studies • Detecting and predicting fires • Predicting network partitioning • Related Work • Conclusions

  6. System Model • Nodes • Large number of static resource-limited sensor nodes (SNs): Motes.. • A few static powerful sinks • A few mobile resource-moderate assist nodes (ANs): PDAs, robots.. • Nodes Know their Own Geographic Position • Clocks are Synchronized • Nodes Functionality • SNs create the model • ANs manage the model • Sinks represent operator(s) SN AN

  7. The MWM Approach • Appropriately Group Spatially-Correlated Readings into Regions and Maps • Maps • Natural way to represent the physical world (spatio-temporal data) • Efficient techniques exist • MWM: A Set of Relevant Maps • User maps (uMAP), e.g., temperature map • Network maps (nMAP), e.g., map of residual energy Region border nodes

  8. Existing Map Construction Algorithms • The eScan Approach [1] • Map-construction along the aggregation-tree • Map is partial at SNs & complete at sink • Data with low time validity (chemicals etc.) • The Isoline Approach [2] • Local flood to label border nodes • Map is partial at SNs & complete at sink • Data with low time validity • The gMAP Approach [3] • AN collects data and construct map • Map at AN • Data with high time validity (energy etc.) [1] Y. Zhao et al. Residual Energy Scan for Monitoring Sensor Networks. In IEEE WCNC, 2002. [2] I. Solis and K. Obraczka. Isolines: Energy-efficient Mapping in Sensor Networks. In ISCC, 2005. [3] A. Khelil et al. gMAP: An Efficient Construction of Global Maps for Mobility-Assisted WSN, TR, 2007.

  9. The MWM Architecture • Main Idea: Address Regions Instead of Nodes • Architecture Retains Existing Abstractions • Substitute node by a region • TinyDB (database) • Pub/sub (service) • Cross layer (network) • Architecture Simplifies • Design of application, • Design of network • Etc.

  10. event attr1 > th1 attr2 > th2 event Queries and Events in MWM • Queries • SQL-like language, query regions instead of sensor nodes • Example: SELECT region, temp FROM tempMAP WHERE temp > threshold • Trade-offs: • Centralized vs. decentralized MWM • Pro-active vs. reactive regioning • Query dissemination [1] • Events • Event: Predicate P(attr1, .. attrk), attri of mapk, e.g., attr1 > th1 • Event composition ≡ geometric operation, e.g., attr1 > th1 & attr2 > th2 [1] R. Sarkar et al. Iso-Contour Queries and Gradient Routing with Guaranteed Delivery in Sensor Networks. infocom’08.

  11. MWM-based WSN Design Methodology (Geometric) abstraction level acceptable by users, application designers and network developers  Simplifies requirement engineering, debugging, standardization etc. Step 1: Identify situations and events of interest (Geometric) Step 2: Identify the required maps (MWM) and define events and their operations in MWM (Geometric) Step 3: Sketch a solution assuming global MWM (Geometric) Step 4: Distribute the required MWM knowledge on nodes (Geometric) Step 5: Select requisite communication primitives

  12. fire fire Case Study 1: Detecting and Predicting Fires Step 1: Fire and pre-fire regions Step 2: Temperature map. Step 3: Fire-temp threshold, pre-fire-temp threshold, regions report to sink Step 4: Border nodes report position and temp value Step 5: Local flood for isoline construction. Each border node unicasts to sink (Not all sensor nodes are illustrated) Existing techniques [1][2] do not • Provide for prediction • Deliver fire perimeter [1] M. Hefeeda et al. Wireless Sensor Networks for Early Detection of Forest Fires. In MASS, 2007. [2] D.M. Doolin et al. Wireless Sensors for Wildfire Monitoring. In SPIE, 2005.

  13. Case Study 2: Predicting Network Partitioning Step 1: Predict coverage drops and isolated regions Step 2: Starting with connected network we require the residual energy map Step 3: Regions of weak energy should report to sink; Sink predicts partitioning Step 4: Border nodes report position and energy value Step 5: Local flood for isoline construction; Each border node unicasts to sink (Not all sensor nodes are illustrated) Existing techniques [1][2] do not • Provide for prediction • Provide important details (partition shape etc.) • Support all shapes/types of partitions [1] N. Shrivastava et al. Detecting Cuts in Sensor Networks. In IPSN, 2005. [2] K.P. Shih et al. PALM: A Partition Avoidance Lazy Movement Protocol for Mobile Sensor Networks. In WCNC, 2007.

  14. Predictive Monitoring and Pro-active Reconfiguration • Predictive Monitoring of both Physical and Network Worlds • Combine (map) data from spatial and temporal domains • Event prediction • Pro-active Network Reconfiguration • Examples: Node displacement • To provide self-healing and graceful degradation • - E.g., by delaying network partition • MWM simplifies • Spatial intervention • Event-triggered autonomous reconfiguration  Predictability and pro-activeness enhance system autonomicity

  15. Related Work • Modeling Technique in WSN • Network models, simulation models etc.: Complex and domain-specific • Geographic Information Systems (GIS) and spatial temporal databases • Modeling languages: SensorML, REACTIVEML and LUSSENSOR  MWM specification • Existing Real World Models • Context-awareness models: Complex, rely on powerful infrastructure, and involve user. • Sentient computing: Focus on indoor scenarios • Augmented and virtual reality models • Real-world models in autonomic computing • All models are „embedded“ in the infrastructure ; We argue for a model distribution • All models dynamically involve the user

  16. Conclusions • The Ongoing Evolution of the Web Map • Interoperability/standardization between WSNs: SensorWeb, SensorGrid etc. • Enhances autonomicity of sensing and reacting • Implementation in OMNET++ simulator • Maps Provide a Widely Accepted Abstraction • We Developed Map-based System Architecture for WSNs • Unified Model for Both Physical and Network Worlds • Powerful Tool for Both Design and Deployment • A novel design methodology • Two case studies WSN 1 WSN 2 WSN 3 WSN 4 Geographic map Queries events

  17. Thanks for your attention! Abdelmajid Khelil, Faisal Karim Shaikh, Brahim Ayari, Neeraj Suri Department of Computer Science TU Darmstadt, Germany {khelil, fkarim, brahim, suri}@informatik.tu-darmstadt.de

More Related