Managing Energy and Data Quality in Large Sensor Swarms
Jinseok Yang, Sameer Tilak, Tajana Simunic Rosing

Citation
Jinseok Yang, Sameer Tilak, Tajana Simunic Rosing. " Managing Energy and Data Quality in Large Sensor Swarms". Talk or presentation, 29, September, 2013; Poster from the First International Workshop on the Swarm at the Edge of the Cloud (SEC'13 @ ESWeek), Montreal.

Abstract
Modern applications typically require measuring several variables for extended periods of time over a large area. To meet the application requirements, the design and deployment has to carefully balance two competing goals: (1) high spatio-temporal resolution to ensure the accuracy of the collected data, and (2) minimal energy consumption to maximize the network lifetime and limit node maintenance. The amount of data that each node collects and processes directly affects both its power consumption and the accuracy of the information obtained. Extending the uptime of a swarm node - especially when they are deployed in difficult to access locations – is an active topic of research. Common approaches try to enhance the battery life directly by harvesting energy from the environment and employing low-power hardware, or using improved wireless protocols and distributed computation for data processing. More recently, researchers are optimizing the battery life indirectly by reducing the overall amount of sensed data. Here, the data is selectively sampled according to a predetermined protocol, reducing the total amount of samples collected by the individual sensor nodes, thus minimizing the energy consumption. To maintain an acceptable amount of total measurements, the missing data is inferred according to statistical models that capture how the data evolves. In addition to enhancing the battery life, these approaches are also able to estimate any lost or corrupted data, making them a popular choice. We recently proposed an energy efficient, data driven technique to estimate missing data within a heterogeneous sensor network [1]. The latent variable factorization model which typically considers only dyadic interactions in data has been extended to multivariate spatio-temporal data, by applying tensor decomposition techniques. The key advantage of using a latent variable model is that it provides a compact representation of the gathered data that can be used to recover the missing samples. In order to perform well under extreme sampling conditions, we explicitly incorporate the spatial, temporal, and inter-sensor correlations. The study focuses on the trade-off between the accuracy in recovering the missing data and the energy consumption when sensor nodes duty cycle to save energy. The proposed technique drastically reduces the amount of sampled data at each node, thus allowing the nodes to spend more time in a low-power sleep mode and save energy. The lower amount of sampled data implies a lower amount of data to transmit from the node to a central gathering station, reducing also the power consumptions associated with the radio communications. Our experiments with the OMNeT++ network simulator using realistic wireless channel conditions, on data collected from two real-world sensor networks, show that we can sample just 20% of the data and can reconstruct the remaining 80% of the data with less than 9% mean error, outperforming similar state-of-the-art techniques such is distributed compressive sampling. In addition, energy savings ranging up to 76%, depending on the sampling rate and the hardware configuration of the node. The next step in our studies is to develop a technique to decide how often latent variables need to be recomputed online to adapt to ever-changing environmental conditions, and to then apply our strategy to a large, swarm size deployment, that covers most of San Diego County. The data from the deployment has 1000s of sensors over an area size of 100x100 sq miles, and has been collected over the period of last 10 years.

Electronic downloads

Citation formats  
  • HTML
    Jinseok Yang, Sameer Tilak, Tajana Simunic Rosing. <a
    href="http://www.terraswarm.org/pubs/132.html"><i>
    Managing Energy and Data Quality in Large Sensor
    Swarms</i></a>, Talk or presentation,  29,
    September, 2013; Poster from the <a
    href="http://www.terraswarm.org/conferences/13/swarm/index.htm"
    >First International Workshop on the Swarm at the Edge of
    the Cloud (SEC'13 @ ESWeek)</a>, Montreal.
  • Plain text
    Jinseok Yang, Sameer Tilak, Tajana Simunic Rosing. "
    Managing Energy and Data Quality in Large Sensor
    Swarms". Talk or presentation,  29, September, 2013;
    Poster from the <a
    href="http://www.terraswarm.org/conferences/13/swarm/index.htm"
    >First International Workshop on the Swarm at the Edge of
    the Cloud (SEC'13 @ ESWeek)</a>, Montreal.
  • BibTeX
    @presentation{YangTilakRosing13_ManagingEnergyDataQualityInLargeSensorSwarms,
        author = {Jinseok Yang and Sameer Tilak and Tajana Simunic
                  Rosing},
        title = { Managing Energy and Data Quality in Large Sensor
                  Swarms},
        day = {29},
        month = {September},
        year = {2013},
        note = {Poster from the <a
                  href="http://www.terraswarm.org/conferences/13/swarm/index.htm"
                  >First International Workshop on the Swarm at the
                  Edge of the Cloud (SEC'13 @ ESWeek)</a>, Montreal.},
        abstract = {Modern applications typically require measuring
                  several variables for extended periods of time
                  over a large area. To meet the application
                  requirements, the design and deployment has to
                  carefully balance two competing goals: (1) high
                  spatio-temporal resolution to ensure the accuracy
                  of the collected data, and (2) minimal energy
                  consumption to maximize the network lifetime and
                  limit node maintenance. The amount of data that
                  each node collects and processes directly affects
                  both its power consumption and the accuracy of the
                  information obtained. Extending the uptime of a
                  swarm node - especially when they are deployed in
                  difficult to access locations â is an active
                  topic of research. Common approaches try to
                  enhance the battery life directly by harvesting
                  energy from the environment and employing
                  low-power hardware, or using improved wireless
                  protocols and distributed computation for data
                  processing. More recently, researchers are
                  optimizing the battery life indirectly by reducing
                  the overall amount of sensed data. Here, the data
                  is selectively sampled according to a
                  predetermined protocol, reducing the total amount
                  of samples collected by the individual sensor
                  nodes, thus minimizing the energy consumption. To
                  maintain an acceptable amount of total
                  measurements, the missing data is inferred
                  according to statistical models that capture how
                  the data evolves. In addition to enhancing the
                  battery life, these approaches are also able to
                  estimate any lost or corrupted data, making them a
                  popular choice. We recently proposed an energy
                  efficient, data driven technique to estimate
                  missing data within a heterogeneous sensor network
                  [1]. The latent variable factorization model which
                  typically considers only dyadic interactions in
                  data has been extended to multivariate
                  spatio-temporal data, by applying tensor
                  decomposition techniques. The key advantage of
                  using a latent variable model is that it provides
                  a compact representation of the gathered data that
                  can be used to recover the missing samples. In
                  order to perform well under extreme sampling
                  conditions, we explicitly incorporate the spatial,
                  temporal, and inter-sensor correlations. The study
                  focuses on the trade-off between the accuracy in
                  recovering the missing data and the energy
                  consumption when sensor nodes duty cycle to save
                  energy. The proposed technique drastically reduces
                  the amount of sampled data at each node, thus
                  allowing the nodes to spend more time in a
                  low-power sleep mode and save energy. The lower
                  amount of sampled data implies a lower amount of
                  data to transmit from the node to a central
                  gathering station, reducing also the power
                  consumptions associated with the radio
                  communications. Our experiments with the OMNeT++
                  network simulator using realistic wireless channel
                  conditions, on data collected from two real-world
                  sensor networks, show that we can sample just 20%
                  of the data and can reconstruct the remaining 80%
                  of the data with less than 9% mean error,
                  outperforming similar state-of-the-art techniques
                  such is distributed compressive sampling. In
                  addition, energy savings ranging up to 76%,
                  depending on the sampling rate and the hardware
                  configuration of the node. The next step in our
                  studies is to develop a technique to decide how
                  often latent variables need to be recomputed
                  online to adapt to ever-changing environmental
                  conditions, and to then apply our strategy to a
                  large, swarm size deployment, that covers most of
                  San Diego County. The data from the deployment has
                  1000s of sensors over an area size of 100x100 sq
                  miles, and has been collected over the period of
                  last 10 years.},
        URL = {http://terraswarm.org/pubs/132.html}
    }
    

Posted by Christopher Brooks on 1 Oct 2013.

Notice: This material is presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each author's copyright.