Challenges in Sensor Networks

1
Challenges in Sensor Networks

• Karl Aberer
• School of Computer and Communication Science,
  EPFL
• National Competence Centre in Mobile Information
  and Communication Systems (NCCR-MICS)

2
Sensor Data Publishing

• Wireless sensor networks, e.g. SensorScope
• Published over the Web, e.g. SensorMap
• Many data formats and schemas
• e.g. NASA Global Change Master Directorycontains
  36000 datasets on env. research
• The usual challenge data integration
• Possible solutions (usual and less usual
  suspects)
• Standardization, e.g. SensorML
• Manual data integration, e.g. SkyServer
• Peer-to-peer data integration
• Community based, e.g. Flickr
• Distributed, collaborative reasoning on metadata
  and schema mappings
• Emergent semantics

3
Semantic Integration of Sensor Data

• Assume we want to integrate sensor data from a
  variety of scientific experiments, research
  groups, disciplines, across spatial and temporal
  scales
• A big opportunity common semantic grounding
• Measurements relate to shared physical phenomena
• Spatial and temporal correspondences
• Rich correspondences and redundancies at data
  level facilitates integration
• A big challenge different interpretations of the
  data
• Sensor data rarely viewed in raw form
  model-based views
• Data cleaning, calibration, interpolation,
  physical models
• Different views depending on available data,
  knowledge, purpose
• A new source of heterogeneity!

4
Model-based Integration of Sensor Data