Autonomic Computing and Communications Systems. Third International ICST Conference, Autonomics 2009, Limassol, Cyprus, September 9-11, 2009, Revised Selected Papers

Research Article

MPM: Map Based Predictive Monitoring for Wireless Sensor Networks

Download291 downloads
  • @INPROCEEDINGS{10.1007/978-3-642-11482-3_6,
        author={Azad Ali and Abdelmajid Khelil and Faisal Shaikh and Neeraj Suri},
        title={MPM: Map Based Predictive Monitoring for Wireless Sensor Networks},
        proceedings={Autonomic Computing and Communications Systems. Third International ICST Conference, Autonomics 2009, Limassol, Cyprus, September 9-11, 2009, Revised Selected Papers},
        proceedings_a={AUTONOMICS},
        year={2012},
        month={4},
        keywords={Predicitve Monitoring Time Series Analysis Wireless Sensor Networks Event Prediction},
        doi={10.1007/978-3-642-11482-3_6}
    }
    
  • Azad Ali
    Abdelmajid Khelil
    Faisal Shaikh
    Neeraj Suri
    Year: 2012
    MPM: Map Based Predictive Monitoring for Wireless Sensor Networks
    AUTONOMICS
    Springer
    DOI: 10.1007/978-3-642-11482-3_6
Azad Ali1,*, Abdelmajid Khelil1,*, Faisal Shaikh1,*, Neeraj Suri1,*
  • 1: Technische Universität Darmstadt
*Contact email: azad@informatik.tu-darmstadt.de, khelil@informatik.tu-darmstadt.de, fkarim@informatik.tu-darmstadt.de, suri@informatik.tu-darmstadt.de

Abstract

We present the design of a Wireless Sensor Networks (WSN) level event prediction framework to monitor the network and its operational environment to support proactive self* actions. For example, by monitoring and subsequently predicting trends on network load or sensor nodes energy levels, the WSN can proactively initiate self-reconfiguration. We propose a Map based Predictive Monitoring (MPM) approach where a selected WSN attribute is first profiled as WSN maps, and based on the maps history, predicts future maps using time series modeling. The ”attribute” maps are created using a gridding technique and predicted maps are used to detect events using our regioning algorithm. The proposed approach is also a general framework to cover multiple application domains. For proof of concept, we show MPM’s enhanced ability to also accurately ”predict” the network partitioning, accommodating parameters such as shape and location of the partition with a very high accuracy and efficiency.