Views from the coalface: chemo-sensors, sensor networks and the semantic sensor web

1
Views from the coalfacechemo-sensors, sensor
networks and the semantic sensor web
Jer Hayes CLARITY / IBM Dublin, Ireland
2
Who?

• The Adaptive sensors group (ASG) is the sensor
  element of the CLARITY Centre for Sensor Web
  Technologies, a joint DCU-UCD research
  partnership funded by Science Foundation Ireland
  under 07/CE/I1147.
• CLARITY is a research centre that  focuses on the
  intersection between two important research areas
  - Adaptive Sensing and Information Discovery.
• IBM is a CLARITY industry partner. Various
  Irish-based centres under the Innovative
  Environmental Solutions grouping.

3
ASG

• Novel sensing

4
About me

• I currently work for IBM within CLARITY

Remote sensing sea surface temp.
Testing wireless sensor networks at sea
Food technology spoilage sensor
5
Outline

• Sensor networks
• Problems with sensors bias?
• Core problems an example sensor system
• Intelligence in the network
• Summary

6
Sensor networks

• The semantic sensor web offers the unique
  opportunity to unify the real and virtual world.
• We are on the cusp of unifying real-world and
  virtual world
• Large scale sensor-networks will be built around
  internet-enable devices (in some cases only the
  base-station may be internet enabled).

7
Physical sensor bias

• Biased towards considering sensors to be like
  thermistors which is understandable as they
  exhibit almost ideal behaviour
• low cost, long-life, very low-power, small form
  factor, high accuracy and precision, rugged,
  reliable, etc.
• Bias colours the expectations of SSW/WSN
  researchers in that they expect all sensors to
  conform to this ideal.
• Sensors arent always reliable
• leaching of active components from sensing
  membranes, physical damage, lack of selectivity,
  baseline drift and biofouling (particularly in
  the marine environment!).

8
Dealing with raw data streams

• Given any sensor we can ask - what does this data
  stream mean ?
• Generally speaking data streams are not self
  identifying
• We require outside information, metadata, to
  understand the stream.
• The main driver for the use of metadata so far
  has been data sharing.
• Scientists generate large amounts of data and
  often we wish to share this data with other
  researchers.

9
Data sharing
SEACOOSsoutheastern Atlantic coastal ocean
observatory system It involves 13 universities
and institutions
NETCDF file format Distributed Oceanographic Data
Systems (DODS) Open Source Project for a Network
Data Access Protocol (OPeNDAP)
10
Core problems from our perspective

1. The heterogeneity of data sources and data
   transport methods that all must neatly fit into
   the SSW.
2. The quality of the data must be described and
   understood.
3. Data streams from different sources and
   modalities (esp. contextual information) which
   vary across many dimensions, including spatial,
   temporal, granularity of data, must be
   integrated.
4. The SSW must be capable of supporting analytics
   (e.g. decision making) across the SSW nodes.

11
Phosphate system

• Component of SmartCoast project, which aims to
  develop a smart water quality monitoring system,
  to aid compliance with increased monitoring
  requirements under the Water Framework Directive.
• Phosphate is a key limiting nutrient in
  freshwater ecosystems.
• Eutrophication
• A major water quality problem in Ireland and many
  other countries
• Elevated nutrient levels lead to excessive
  growth of algae and aquatic plants
• Oxygen depletion ? fish kills
• Algal blooms ? toxicity in water bodies

12
Objective and Requirements

• Develop an autonomous, remotely controlled
  phosphate sensor capable of monitoring PO43- at
  appropriate levels at remote locations over long
  deployments
• Requirements
• Sensitive
• Stable chemistry
• Communicate wirelessly
• Low power
• Robust portable
• Low cost low maintenance requirements

13
Principle of Operation

• Yellow method for phosphate detection
• Forms vanadomolybdophosphoric acid (yellow)
• Absorption proportional to phosphate conc.
• Advantages
• Excellent reagent stability
• Fast reaction time (minutes)
• Microfluidic technology
• Minimizes reagent consumption, storage
  requirements and pumping power
• UV-LED and photodiode
• Low powered, inexpensive sensitive optical
  detection

14
Current Status

• Mark II sensor designed to build on the successes
  and address the limitations of the original.
• Improvements
• Lower power, more flexible fluid handling system.
• More sensitive optical detection system.
• More reliable and lower powered communications
  using GSM modem in SMS mode.
• 2 point calibration protocol.
• Solar panel for energy harvesting during long
  deployments.
• Improved ruggedisation.

Yeah, so what? What about the semantic sensor
web?!!!
15
Problems at the coalface

• How do we plug this sensor into a sensor network
  and / or the semantic sensor web?
• What? Where? When? Who?
• What is context for this sensor? How do we find
  it and can we trust it?
• Weather? Other water quality parameters.
• Data from models? Topology, soil type, land use
  in river basin district
• For other sensors what context is could be
  complex.
• Can the network control the device properly? Can
  it change sampling rates etc.?
• Do we just pass on data or can we control the
  device? Who is allowed to control it?

16
Problems at the coalface

• We are looking for a complete picture so data
  streams will come from hardware software, e.g.
  modelling.
• Air quality
• Ground based in-situ sensors
• Remote sensing
• Models chemical transport, weather etc.

Software sensors (Models/Database Hardware
sensors?
17
Analysed using IR gas sensor
Chemometric program analyses data and decides if
concentrations are within threshold limits
Gas sample extracted
Borehole well
If thresholds are exceeded, a message to sent to
personnel onsite to investigate further
CH4
VOCs
CO2
Landfill gas generation
18
Intelligence?

• OGC's Sensor Web Enablement (SWE)

Where does the landfill site fit in?
Where should the analytics take place?
How do we know contextual information is accurate?
Should bad data be released?
19
Summary

• Sensors arent as reliable as wed like to think.
• Need to account for data quality.
• Contextual information is required for the
  complete picture.
• From a large variety of possible sources