Transcript Semantics & Sensors: The web of real

Image: Burdekin Sensor Network,
Pavan Sikka & Google
Using Explicit Semantic Representations for
User Programming of Sensor Devices
Kerry Taylor and Patrick Penkala
CSIRO ICT Centre
Melbourne, 1st December 2009
Context
• lots of pics of sensors
CSIRO. Australasian Ontology Workshop. Melbourne, 1 December 2009
SSN-XG: Semantic Sensor Network Incubator Group
Commenced 1 March 2009.
Two main objectives:
(a) the development of ontologies for describing sensors, and
(b) the extension of the Sensor Model Language (SML), one of
the four SWE languages, to support semantic annotations.
CSIRO. Australasian Ontology Workshop. Melbourne, 1 December 2009
Aim: To address real-time programming, tasking and
querying sensors and sensor networks
• Represent the semantics of the
command language in an
ontology
• Use generic software tools, plus
device-specific “transformer” and
communication code modules
• Assume a stateless model
(declarative queries)
• simplicity
• amenability to optimisation
• multi-user sharing (detect query
subsumption, for example)
CSIRO. Australasian Ontology Workshop. Melbourne, 1 December 2009
Case Study: an Automatic Weather Station
• Environdata
WeatherMaster1600
• sensors for:
•
•
•
•
air temperature,
relative humidity,
wind speed,
wind direction
• + 3 simulated sensors: voltages
of the battery and solar panel and
the activity of the serial port.
• proprietary command-line
language of about 50 commands
• request-response interaction
style over a serial port.
• Data is time-stamped and
logged: for each of the four
sensors at once.
• 104 kilobytes memory, FIFO
CSIRO. Australasian Ontology Workshop. Melbourne, 1 December 2009
Environdata Command Language
Main Commands:
• STORAGE to measure data and
log in memory
• “STORAGE 13 CURRENT 2 3 0
0 1 EHOUR 1 0”
• command 13 logs the current
wind direction in memory 2 every
hour.
• MEM to retrieve data from
memory
• “MEM 4 SPECIFIC 2010 11 30 09
00 00 2010 12 01 09 00 00”
• requests logged data in MEM 4
for the given 24 hour period
• R for current values for all
sensors
• “R”
CSIRO. Australasian Ontology Workshop. Melbourne, 1 December 2009
1. Model the Commands in an Ontology
queryCurrentData
queryPeriodData
setStorageFunction
CSIRO. Australasian Ontology Workshop. Melbourne, 1 December 2009
2. Phrase queries using ontology terms in a deviceindependent query tool
CSIRO. Australasian Ontology Workshop. Melbourne, 1 December 2009
3. Classify query and instantiate
CSIRO. Australasian Ontology Workshop. Melbourne, 1 December 2009
4. Execute and see the results!
CSIRO. Australasian Ontology Workshop. Melbourne, 1 December 2009
Benefits
• Offers a device-independent route to sensor programming, but
avoids standardising to lowest common model.
• Validates queries by classification
• Is self-documenting language through semantic context.
• Can accommodate (some) evolution without coding.
• Can also use the ontology modelling and DL reasoning to
•
•
•
•
Represent variation in query capability amongst similar devices
Allocate queries to devices that are sufficiently capable
Admit alternative “syntaxes” (or terminology) for same functions
Discover sensors by function, location, latency, frequency,
accuracy, data format, custodian,...
• Optimise wrt query subsumption (e.g. logging frequency)
• Can extend to composition, substitution, spatial and temporal
reasoning etc
(see Compton et al in Proc Semantic Sensor Networks 2009)
CSIRO. Australasian Ontology Workshop. Melbourne, 1 December 2009
Future Work
Phenomics: Start with a particular observable trait or phenotype and work back to
discover the causal gene.
CSIRO. Australasian Ontology Workshop. Melbourne, 1 December 2009
CSIRO ICT Centre
Kerry Taylor
Research Scientist
Phone: 02 6216 7038
Email: [email protected]
Web: www.ict.csiro.au
SSN-XG: www.w3.org/2005/Incubator/ssn/
Thank you
Contact Us
Phone: 1300 363 400 or +61 3 9545 2176
Email: [email protected] Web: www.csiro.au