Aspect-Oriented Modeling for Anomaly Detection and State Estimation
Ilge Akkaya, Edward A. Lee

Citation
Ilge Akkaya, Edward A. Lee. "Aspect-Oriented Modeling for Anomaly Detection and State Estimation". Talk or presentation, 13, February, 2014.

Abstract
The TerraSwarm vision of a swarm of networked sensors and actuators brings about many opportunities in monitoring real-time state in distributed heterogeneous systems and in decision making. However, the advantages are followed by immediate challenges tied to handling high data rates and aggregate failures that may remain undetected in a heterogeneous system setting. We explore machine learning based anomaly detection and state estimation techniques to leverage autonomous fault detection and low-cost system state tracking. We present our progress in a generic machine learning toolkit prototyped in Ptolemy II, that would enable sensor and application developers to utilize machine learning tools for on-line analysis of sensor data. We present a particle filter interface in Ptolemy II within a two-robot target localization example and introduce Hidden-Markov Model estimators that can be integrated with models in an aspect-oriented way.

Electronic downloads

Internal. This publication has been marked by the author for TerraSwarm-only distribution, so electronic downloads are not available without logging in.

Citation formats

• HTML

  Ilge Akkaya, Edward A. Lee. <a
  href="http://www.terraswarm.org/pubs/269.html"
  ><i>Aspect-Oriented Modeling for Anomaly Detection
  and State Estimation</i></a>, Talk or
  presentation,  13, February, 2014.

• Plain text

  Ilge Akkaya, Edward A. Lee. "Aspect-Oriented Modeling
  for Anomaly Detection and State Estimation". Talk or
  presentation,  13, February, 2014.

• BibTeX

  @presentation{AkkayaLee14_AspectOrientedModelingForAnomalyDetectionStateEstimation,
      author = {Ilge Akkaya and Edward A. Lee},
      title = {Aspect-Oriented Modeling for Anomaly Detection and
                State Estimation},
      day = {13},
      month = {February},
      year = {2014},
      abstract = {The TerraSwarm vision of a swarm of networked
                sensors and actuators brings about many
                opportunities in monitoring real-time state in
                distributed heterogeneous systems and in decision
                making. However, the advantages are followed by
                immediate challenges tied to handling high data
                rates and aggregate failures that may remain
                undetected in a heterogeneous system setting. We
                explore machine learning based anomaly detection
                and state estimation techniques to leverage
                autonomous fault detection and low-cost system
                state tracking. We present our progress in a
                generic machine learning toolkit prototyped in
                Ptolemy II, that would enable sensor and
                application developers to utilize machine learning
                tools for on-line analysis of sensor data. We
                present a particle filter interface in Ptolemy II
                within a two-robot target localization example and
                introduce Hidden-Markov Model estimators that can
                be integrated with models in an aspect-oriented
                way. },
      URL = {http://terraswarm.org/pubs/269.html}
  }
  

Posted by Ilge Akkaya on 18 Feb 2014.
Groups: tools