Environmental Cyberinfrastructure Needs for Distributed Sensor Networks

1
Environmental Cyberinfrastructure Needs for
Distributed Sensor Networks
Presentation by Adam Cornwell
2
Background

• Workshop convened to identify problems and pose
  creative solutions to issues related to
  environmental sensor networks, including
• communication and data transfer,
• remote management,
• collecting/managing/archiving data,
• data analysis/visualization/quality control,
• security,
• Development and deployment of new sensors

3
Sensor Net?

• Sensors for the primary purpose of collecting
  environmental information, can be deployed as an
  array of multiple sensors
• Sensor categories
• Physical- temperature, wind speed, etc
• Chemical- carbon dioxide, pH, etc
• Biological- Video camera, microphone, RFID, etc
• Technology has enabled sensors to send back data
  in real time, remote management, and even
  alteration of experiments remotely

4
More on sensors

• Sensors do not yet exist for many types of data,
  integration of multiple sensors in a network and
  appropriate analytical methods may be used to
  infer the desired information
• Desired properties of sensors
• Reliable for long term applications
• High power efficiency
• Able to communicate with other net nodes (and
  provide data about itself)
• Affordable
• Low environmental impact

5
Sensor Arrays

• Have become attractive for applications where
  single sensor remote sensing is not adequate or
  possible
• Spatially distributed phenomena, difficult to
  sense from more remote locations with single
  sensors
• Fault tolerance and validation multiple sensors,
  maybe even a few of the same type in one area,
  prevents a sensor failure from affecting the
  results of the system, and provides a way to
  verify sensor output, as well as a reference for
  sensor calibration.

6
Goals Related to Informatics

• Increase collaboration between bio / eco / geo /
  IT / CS / engineering fields to
• Develop lower cost but still reliable and power
  efficient sensors, new types of sensors
• Increase portability and flexibility of software
• Automate tasks (sensor calibration, data quality
  control, etc) to reduce hardware and software
  maintenance cost
• All of these will be achieved more easily after
  standards are developed and adopted

7
Current RD for a Sensor Net Platform

• Collaboration between UC Berkeley and Intel has
  produced a prototype of a complete platform
• Mote Single chip device with a processor,
  memory, and a bluetooth radio which runs the OS
  and interfaces to the sensor-specific boards
• TinyOS Open source OS developed for embedded
  devices in wireless sensor networks
• TinyDB Allows for querying data from motes with
  SQL-like commands and a java-based client API

8
Intel Mote
Taken from http//www.intel.com/research/explorato
ry/motes.htm
9
TinyDB Query Window
Taken from http//telegraph.cs.berkeley.edu/tinydb
/screenshots.htm
10
Security

• Quality of data of provided by a sensor network
  is more closely related to the security of the
  networked systems than in many bioinformatics-rela
  ted systems
• Remote locations of sensor networks in
  uncontrolled environments allow physical security
  to be compromised
• Wireless nature of most sensor arrays means data
  can be more easily intercepted, or attacked by
  methods such as Denial of Service
• Lightweight encryption provided in TinyOS network
  stack, TinySEC

11
Managing the Data

• Environmental data sets are becoming larger and
  more diverse
• Data needs to be transferred and stored, and
  shared when applicable
• Work with other groups and organizations that are
  addressing the same issues, to collaboratively
  develop open standards
• Data sharing and finding relevant data remotely
  should become less of an issue when solutions
  such as Web Services mature.
• Might be reasonable to establish a central
  publicly accessible resource for environmental
  data- like an NCBI, but for ecoinformatics
• Data models should designed with the scientific
  objectives in mind

12
Metadata

• In the past, metadata has been considered to be
  of little use- the individuals in a group
  performing an experiment know where the relevant
  data is and what it means
• As more data accumulates from an increasing
  number of sensors, and the multi-disciplinary
  sharing of data becomes more common, metadata
  becomes more important, despite there being a
  cost of generating it
• Metadata in this case could the sensors self
  description data that is sent with its data
  (sensor type, ID number, etc) and information
  about the syntax and semantics of the data itself
• A metadata standard would need to be extensible
  in a hierarchical manner, and easy to view on the
  internet- XML based solutions seem to fit the bill

13
Metadata
Page 49
14
Conclusions

• Problems being faced now in other scientific
  fields with increasing data set sizes are often
  similar if not the same as the problems that
  caused the need for bioinformatics in biology
• Share many of the same requirements for
  transformation, storage, and access of data.
  Development of standards needed.
• No need to reinvent the wheel- collaborate when
  possible
• As data types become more varied, there is more
  of a need for multidiscipline ventures
• Education is the key to acceptance, effective
  use, and future developments- includes
  scientists, policy makers, and the public