This tutorial presents tools and techniques for effectively utilizing the Internet of Things (IoT) for building advanced applications, including the Physical-Cyber-Social (PCS) systems. The issues and challenges related to IoT, semantic data modelling, annotation, knowledge representation (e.g. modelling for constrained environments, complexity issues and time/location dependency of data), integration, analy- sis, and reasoning will be discussed. The tutorial will de- scribe recent developments on creating annotation models and semantic description frameworks for IoT data (e.g. such as W3C Semantic Sensor Network ontology). A review of enabling technologies and common scenarios for IoT applications from the data and knowledge engineering point of view will be discussed. Information processing, reasoning, and knowledge extraction, along with existing solutions re- lated to these topics will be presented. The tutorial summarizes state-of-the-art research and developments on PCS systems, IoT related ontology development, linked data, do- main knowledge integration and management, querying large- scale IoT data, and AI applications for automated knowledge extraction from real world data.
Related: Semantic Sensor Web: http://knoesis.org/projects/ssw
Physical-Cyber-Social Computing: http://wiki.knoesis.org/index.php/PCS
Data Processing and Semantics for Advanced Internet of Things (IoT) Applications: modeling, annotation, integration, and perception
1. Data Processing and Semantics for Advanced
Internet of Things (IoT) Applications:
modeling, annotation, integration, and perception
Pramod Anantharam1
, Payam Barnaghi2
, Amit Sheth1
1
Kno.e.sis Center, Wright State University
2
Centre for Communication Systems Research, University of Surrey
Madrid, Spain, June 12-14, 2013
http://aida.ii.uam.es/wims13/keynotes.php#tutorials
Special Thanks to:
Cory Henson, Kno.e.sis Research Center, Wright State University
Wei Wang, Centre for Communication Systems Research, University of Surrey
3. Internet of Things
“sensors and actuators embedded in physical
objects — from containers to pacemakers —
are linked through both wired and wireless
networks to the Internet.”
“When objects in the IoT can sense the
environment, interpret the data, and
communicate with each other, they become
tools for understanding complexity and for
responding to events and irregularities swiftly”
source: http://www.iot2012.org/
4. “Thing” connected to the internet
“In 2013 cumulative shipments of Bluetooth-enabled devices will surpass 10 billion
and Wi-Fi enabled devices will surpass 10 billion cumulative shipments in 2015,”
- Peter Cooney, wireless analyst with ABI Research
“73% of Countries with 4G Services Have Dropped Their 4G Tariffs by an
Average of 30% in the Past 6 Months”1
- ABI Research (07 Feb 2013)
1http://www.abiresearch.com/
10. 10
Wireless Sensor Networks (WSN)
Sink
node Gateway
Core network
e.g. InternetGateway
End-user
Computer services
- The networks typically run Low Power Devices
- Consist of one or more sensors, could be different type of sensors (or actuators)
11. 11
Key characteristics of IoT devices
Often inexpensive sensors (actuators) equipped with a radio
transceiver for various applications, typically low data rate ~ 10-250
kbps.
Deployed in large numbers
The sensors should coordinate to perform the desired task.
The acquired information (periodic or event-based) is reported back
to the information processing centre (or sometimes in-network
processing is required)
Solutions are application-dependent.
11
12. 12
Beyond conventional sensors
Human as a sensor (citizen sensors)
e.g. tweeting real world data and/or events
Virtual (software) sensors
e.g. Software agents/services
generating/representing data
Road block, A3
Road block, A3
Suggest a different route
17. 17
Making Sense of Data
In the next few years, sensor networks will produce
10-20 time the amount of data generated by social
media. (source: GigaOmni Media)
18. 18
Things, Data, and lots of it
image courtesy: Smarter Data - I.03_C by Gwen Vanhee
19. 19
Big Data and IoT
"Big data" is a term applied to data sets whose size is beyond the ability of
commonly used software tools to capture, manage, and process the data
within a tolerable elapsed time. Big data sizes are a constantly moving
target, as of 2012 ranging from a few dozen terabytes to many petabytes of
data in a single data set.” (wikipedia)
Every day, we create 2.5 quintillion bytes of data — so much that 90% of
the data in the world today has been created in the last two years alone.
(source IBM)
20. 20
The seduction of data
Turn 12 terabytes of Tweets created each day into sentiment analysis
related to different events/occurrences or relate them to products and
services.
Convert (billions of) smart meter readings to better predict and balance
power consumption.
Analyze thousands of traffic, pollution, weather, congestion, public transport
and event sensory data to provide better traffic management.
Monitor patients, elderly care and much more…
Adapted from: What is Bog Data?, IBM
22. 22
“Raw data is both an oxymoron and bad
data”
Geoff Bowker, 2005
ource: Kate Crawford, "Algorithmic Illusions: Hidden Biases of Big Data", Strata 2013.
23. 23
IoT Data in the Cloud
Image courtesy: http://images.mathrubhumi.com
http://www.anacostiaws.org/userfiles/image/Blog-Photos/river2.jpg
25. 25
Change in communication paradigm
Sink
node Gateway
Core network
e.g. Internet End-user
Data
Sender
Data
Receiver
A sample data communication in conventional networks
A sample data communication in WSN
Fire! Some bits
01100011100
26. 26
Collaboration and in-network processing
In some applications a single sensor node is not able to handle the
given task or provide the requested information.
Instead of sending the information form various source to an external
network/node, the information can be processed in the network itself.
e.g. data aggregation, summarisation and then propagating the processed
data with reduced size (hence improving energy efficiency by reducing the
amount of data to be transmitted).
Data-centric
Conventional networks often focus on sending data between two
specific nodes each equipped with an address.
Here what is important is data and the observations and measurements
not the node that provides it.
Required mechanisms
27. “People want answers, not numbers”
(Steven Glaser, UC Berkley)
Sink
node Gateway
Core network
e.g. Internet
What is the temperature at home?Freezing!
28. 28
IoT Data alone is not enough
Domain knowledge
Machine interpretable meta data
Delivery, sharing and representation services
Query, discovery, aggregation services
Publish, subscribe, notification, and access interfaces/services
30. 30
IoT Data Challenges
Discovery: finding appropriate device and data sources
Access: Availability and (open) access to IoT resources and data
Search: querying for data
Integration: dealing with heterogeneous device, networks and data
Interpretation: translating data to knowledge usable by people and
applications
Scalability: dealing with large number of devices and myriad of data
and computational complexity of interpreting the data.
31. Energy consumption of the nodes
Batteries have small capacity and recharging
could be complex (if not impossible) in some
cases.
The main consumers of the energy are: the
controller, radio, to some extent memory and
depending on the type, the sensor(s).
A controller can go to:
“active”, “idle” and “sleep”
A radio modem could turn transmitter,
receiver, or both on or off,
sensors and memory can be also turned on
and off.
35. What are the main issues?
Heterogeneity
Interoperability
Mobility
Energy efficiency
Scalability
Security
36. What is important?
Robustness
Quality of Service
Scalability
Seamless integration
Security, privacy, Trust
37. In-network processing
Mobile Ad-hoc Networks are supposed to deliver
bits from one end to the other
WSNs, on the other end, are expected to provide
information, not necessarily original bits
Gives addition options
E.g., manipulate or process the data in the network
Main example: aggregation
Applying aggregation functions to a obtain an average
value of measurement data
Typical functions: minimum, maximum, average, sum, …
Not amenable functions: median
source: Protocols and Architectures for Wireless Sensor Networks, Protocols and Architectures for Wireless Sensor Networks
Holger Karl, Andreas Willig, chapter 3, Wiley, 2005 .
38. In-network processing- example
Applying Symbolic Aggregate Approximation (SAX)
SAX Pattern (blue) with word length of 20 and a vocabulary of 10 symbols
over the original sensor time-series data (green)
39. Data-centric networking
In typical networks (including ad hoc networks),
network transactions are addressed to the identities
of specific nodes
A “node-centric” or “address-centric” networking paradigm
In a redundantly deployed sensor networks, specific
source of an event, alarm, etc. might not be important
Redundancy: e.g., several nodes can observe the same
area
Thus: focus networking transactions on the data
directly instead of their senders and transmitters !
data-centric networking
Principal design change
source: Protocols and Architectures for Wireless Sensor Networks, Protocols and Architectures for Wireless Sensor Networks
Holger Karl, Andreas Willig, chapter 3, Wiley, 2005 .
40. Implementation options for
data-centric networking
Overlay networks & distributed hash tables (DHT)
Hash table: content-addressable memory
Retrieve data from an unknown source, like in peer-to-peer networking – with
efficient implementation
Some disparities remain
Static key in DHT, dynamic changes in WSN
DHTs typically ignore issues like hop count or distance between nodes when
performing a lookup operation
Publish/subscribe
Different interaction paradigm
Nodes can publish data, can subscribe to any particular kind of data
Once data of a certain type has been published, it is delivered to all subscribes
Subscription and publication are decoupled in time; subscriber and published are
agnostic of each other (decoupled in identity);
There is concepts of Semantic Sensor Networks- to annotate sensor resources
and observation and measurement data!
Adapted from: Protocols and Architectures for Wireless Sensor Networks, Protocols and Architectures for Wireless Sensor Networks
Holger Karl, Andreas Willig, chapter 3, Wiley, 2005 .
41. IoT and Semantic technologies
The sensors (and in general “Things”) are increasingly
being integrated into the Internet/Web.
This can be supported by embedded devices that
directly support IP and web-based connection (e.g.
6LowPAN and CoAp) or devices that are connected
via gateway components.
Broadening the IoT to the concept of “Web of Things”
There are already Sensor Web Enablement (SWE)
standards developed by the Open Geospatial
Consortium that are widely being adopted in industry,
government and academia.
While such frameworks provide some interoperability,
semantic technologies are increasingly seen as key
enabler for integration of IoT data and broader Web
information systems.
42. Semantics and IoT resources and
data
Semantics are machine-interpretable metadata (for mark-up),
logical inference mechanisms, query mechanism, linked data
solutions
For IoT this means:
ontologies for: resource (e.g. sensors), observation and
measurement data (e.g. sensor readings), domain concepts (e.g.
unit of measurement, location), services (e.g. IoT services) and
other data sources (e.g. those available on linked open data)
Semantic annotation should also supports data represented
using existing forms
Reasoning /processing to infer relationships and hierarchies
between different resources, data
Semantics (/ontologies) as meta-data (to describe the IoT
resources/data) / knowledge bases (domain knowledge).
44. SSW Introduction
lives
in
has
pet
is ahas pet
Perso
n
Perso
n
Anima
l
Anima
l
Concrete Facts
Resource Description Framework
Concrete Facts
Resource Description Framework
Semantic Web
(according to Farside)
General Knowledge
Web Ontology Language
General Knowledge
Web Ontology Language
“Now! – That should clear up a few things around
is a
53. Enabling the Internet of Things
Situational awareness
enables:
Devices/things to function
and adapt within their
environment
Devices/things to work
together
54. Sensor systems are
too often stovepiped.
Closed centralized
management of sensing
resources
Closed inaccessible
data and sensors
55. We want to set this data free
With freedom comes
responsibility
Discovery, access, and search
Integration and interpretation
Scalability
56. Drowning in Data
A cross-country flight from New York to Los Angeles on a
Boeing 737 plane generates a massive 240 terabytes of
data
- GigaOmni Media
58. Challenges
To fulfill this vision, there are difficult challenges to overcome such
as the discovery, access, search, integration, and interpretation of
sensors and sensor data at scale
Discovery finding appropriate sensing resources and data sources
Access sensing resources and data are open and available
Search querying for sensor data
Integration dealing with heterogeneous sensors and sensor data
Interpretation translating sensor data to knowledge usable by people
and
applications
Scalability dealing with data overload and computational complexity
of interpreting the data
59. Solution
Semantic Sensor Web
Internet Computing, July/Aug.
2008
Uses the Web as platform for
managing sensor resources and
data
Uses semantic technologies for
representing data and knowledge,
integration, and interpretation
62. Solution
Interpretation
Abstraction – converting low-level data to high-level knowledge
Machine Perception – w/ prior knowledge and abductive
reasoning
IntellegO – Ontology of Perception
63. Solution
Scalability
Data overload – sensors produce too much data
Computational complexity of semantic interpretation
“Intelligence at the edge” – local and distributed integration and
interpretation of sensor data
65. 65
Part 2: Data and knowledge
modelling requirements,
semantic annotation
Image source: CISCO
66. Recall of the Internet of Things
A primary goal of interconnecting devices and
collecting/processing data from them is to
create situation awareness and enable
applications, machines, and human users to
better understand their surrounding
environments.
The understanding of a situation, or context,
potentially enables services and applications
to make intelligent decisions and to respond to
the dynamics of their environments.Barnaghi et al 2012, “Semantics for the Internet of Things: early progress and back to the future”
67. IoT challenges
Numbers of devices and different users and interactions required.
Challenge: Scalability
Heterogeneity of enabling devices and platforms
Challenge: Interoperability
Low power sensors, wireless transceivers, communication, and networking
for M2M
Challenge: Efficiency in communications
Huge volumes of data emerging from the physical world, M2M and new
communications
Challenge: Processing and mining the data, Providing secure access and
preserving and controlling privacy.
Timeliness of data
Challenge: Freshness of the data and supporting temporal requirements in
accessing the data
Ubiquity
Challenge: addressing mobility, ad-hoc access and service continuity
Global access and discovery
Challenge: Naming, Resolution and discovery
68. IoT: one paradigm, many visions
Diagram adapted from L. Atzori et al., 2010, “the Internet of Things: a Survey”
69. Semantic oriented vision
“The object unique addressing and the representation
and storing of the exchanged information become the
most challenging issues, bringing directly to a ‘‘Semantic
oriented”, perspective of IoT”, [Atzori et al., 2010]
Data collected by different sensors and devices is
usually multi-modal (temperature, light, sound, video,
etc.) and diverse in nature (quality of data can vary with
different devices through time and it is mostly location
and time dependent [Barnaghi et al, 2012]
some of challenging issues: representation, storage, and
search/discovery/query/addressing, and processing IoT
resources and data.
70. What is expected?
Unified access to data: unified descriptions
Deriving additional knowledge (data mining)
Reasoning support and association to other entities and
resources
Self-descriptive data an re-usable knowledge
In general: Large-scale platforms to support discovery
and access to the resources, to enable autonomous
interactions with the resources, to provide self-
descriptive data and association mechanisms to reason
the emerging data and to integrate it into the existing
applications and services.
71. Semantic technologies and IoT
There are already Sensor Web Enablement
(SWE) standards developed by the Open
Geospatial Consortium that are widely adopted.
While such frameworks provide certain levels of
interoperability, semantic technologies are seen
as key enabler for integration of IoT data and
and existing business information systems.
Semantic technologies provide potential support
for:
Interoperability and machine automation
IoT resource and data annotation, logical inference, query
and discovery, linked IoT data
73. IoT domain concepts - Entity
Physical entities (or entity of
interests): objects in the physical world,
features of interest that are of interests to
users (human users or any digital artifacts).
Virtual entities: virtual representation of the
physical entities.
Entities are the main focus of interactions
between humans and/or software agents.
This interaction is made possible by a hardware
component called Device.
Definition adapted from De et al, 2012, “Service modeling for the Internet of Things”
74. IoT domain concepts –
Device, Resource and Service
A Device mediates the interactions between users
and entities.
The software component that provides information
on the entity or enables controlling of the device, is
called a Resource.
A Service provides well-defined and standardised
interfaces, offering all necessary functionalities for
interacting with entities and related processes.
Definition adapted from De et al, 2012, “Service modeling for the Internet of Things”
75. Other concepts need to considered
Gateways
Directories
Platforms
Systems
Subsystems
…
Relationships among them
And links to existing knowledge base and linked data
76. Don’t forget the IoT data
Sensors and devices provide observation and
measurement data about the physical world objects
which also need to be semantically described and can
be related to an event, situation in the physical world.
The processing of data into knowledge/ perception and
using it for decision making, automated control, etc.
Huge amount of data from our physical world that need
to be
Annotated
Published
Stored (temporary or for longer term)
Discovered
Accessed
Proceeded
Utilised in different applications
77. Semantics for IoT resources and data
Semantics are machine-interpretable metadata, logical inference
mechanisms, query and search mechanism, linked data…
For IoT this means:
ontologies for: resource (e.g. sensors), observation and
measurement data (e.g. sensor readings), services (e.g. IoT
services), domain concepts (e.g. unit of measurement, location)
and other data sources (e.g. those available on linked open
data)
Semantic annotation should also supports data represented using
existing forms
Reasoning/processing to infer relationships between different
resources and services, detecting patterns from IoT data
78. Characteristics of IoT resources
Extraordinarily large number
Limited computing capabilities
Limited memory
Resource constrained environments (e.g.,
battery life, signal coverage)
Location is important
Dynamism in the physical environments
Unexpected disruption of services
…
79. Characteristics of IoT data
Stream data (depends on time)
Transient nature
Almost always related to a phenomenon or
quality in our physical environments
Large amount
Quality in many situations cannot be assured
(e.g., accuracy and precision)
Abstraction levels (e.g., raw, inferred or
derived)
…
80. Utilise semantics
Find all available resources (which can provide
data) and data related to “Room A” (which is an
object in the linked data)?
What is “Room A”? What is its location? returns “location” data
What type of data is available for “Room A” or that “location”?
(sensor category types)
Predefined Rules can be applied based on
available data
(TempRoom_A > 80°C) AND (SmokeDetectedRoom_A position==TRUE)
FireEventRoom_A
Learning these rules needs data mining or pattern recognition
techniques
82. Semantic modelling
Lightweight: experiences show that a lightweight
ontology model that well balances expressiveness and
inference complexity is more likely to be widely adopted
and reused; also large number of IoT resources and
huge amount of data need efficient processing
Compatibility: an ontology needs to be consistent with
those well designed, existing ontologies to ensure
compatibility wherever possible.
Modularity: modular approach to facilitate ontology
evolution, extension and integration with external
ontologies.
83. Existing models for resources and data
W3C Semantic Sensor Network Incubator
Group’s SSN ontology (mainly for sensors
and sensor networks, observation and
measurement, and platforms and systems)
Quantity Kinds and Units
Used together with the SSN ontology
based on QUDV model OMG SysML(TM)
Working group of the SysML 1.2 Revision Task
Force (RTF) and W3C Semantic Sensor Network
Incubator Group
84. Existing models for services
OWL-S and WSMO are heavy weight models: practical
use?
Minimal service model
Deprecated
Procedure-Oriented Service Model (POSM) and Resource-
Oriented Service Model (ROSM): two different models for
different service technologies
Defines Operations and Messages
No profile, no grounding
SAWSDL: mixture of XML, XML schema, RDF and OWL
hRESTS and SA-REST: mixture of HTML and reference
to a semantic model; sensor services are not anticipated
to have HTML
86. Some existing IoT models and
ontologies
FP7 IoT-A project’s Entity-Resource-Service
ontology
A set of ontologies for entities, resources, devices
and services
Based on the SSN and OWL-S ontology
FP7 IoT.est project’s service description
framework
A modular approach for designing a description
framework
A set of ontologies for IoT services, testing and
QoS/QoI
92. IoT.est service profile highlight
ServiceType class represents the service
technologies: RESTful and SOAP/WSDL services.
serviceQos and serviceQoI are defined as
subproperty of serviceParameter; they link to concepts
in the QoS/QoI ontology.
serviceArea: the area where the service is provided;
different from the sensor observation area
Links to the IoT resources through “exposedBy”
property
Future extension:
serviceNetwork, servicePlatform and
serviceDeployment
Service lifecycle, SLA…
94. Linked data principles
using URI’s as names for things: Everything is
addressed using unique URI’s.
using HTTP URI’s to enable people to look up
those names: All the URI’s are accessible via
HTTP interfaces.
provide useful RDF information related to
URI’s that are looked up by machine or
people;
including RDF statements that link to other
URI’s to enable discovery of other related
concepts of the Web of Data: The URI’s are
linked to other URI’s.
95. Linked data in IoT
Using URI’s as names for things;
- URI’s for naming IoT resources and data (and also streaming
data);
Using HTTP URI’s to enable people to look up those
names;
- Web-level access to low level sensor data and real world
resource descriptions (gateway and middleware solutions);
Providing useful RDF information related to URI’s that are looked
up by machine or people;
- publishing semantically enriched resource and data descriptions
in the form of linked RDF data;
Including RDF statements that link to other URI’s to enable
discovery of other related things of the web of data;
- linking and associating the real world data to the existing data on
the Web;
96. Linked data layer for not only IoT…
Images from Stefan Decker, http://fi-ghent.fi-week.eu/files/2010/10/Linked-Data-scheme1.png; linked data diagram: http://richard.cyganiak.de/2007/10/lod/
97. Creating and using linked sensor data
http://ccsriottb3.ee.surrey.ac.uk:8080/IOTA/
99. Semantics in IoT - reality
If we create an Ontology our data is interoperable
Reality: there are/could be a number of ontologies for a domain
Ontology mapping
Reference ontologies
Standardisation efforts
Semantic data will make my data machine-understandable and my system
will be intelligent.
Reality: it is still meta-data, machines don’t understand it but can interpret it. It still
does need intelligent processing, reasoning mechanism to process and interpret the
data.
It’s a Hype! Ontologies and semantic data are too much overhead; we deal
with tiny devices in IoT.
Reality: Ontologies are a way to share and agree on a common vocabulary and
knowledge; at the same time there are machine-interpretable and represented in
interoperable and re-usable forms;
You don’t necessarily need to add semantic metadata in the source- it could be added
to the data at a later stage (e.g. in a gateway);
100. 100
Part 5: Semantic Sensor Web
and
Perception
Image source: semanticweb.com; CISCO
101. What is the Sensor Web?
Sensor Web is an additional layer connecting sensor
networks to the World Wide Web.
Enables an interoperable usage of sensor resources by
enabling web based discovery, access, tasking, and
alerting.
Enables the advancement of
cyber-physical applications through
improved situation awareness.
102. Why is the Sensor Web important?
In general
Enable tight coupling of the cyber and
physical world
In relation to IoT
Enable shared situation awareness (or
context) between devices/things
103. Bridging the Cyber-Physical Divide
Psyleron’s Mind-Lamp (Princeton U),
connections between the mind and the
physical world.
Neuro Sky's mind-controlled headset to
play a video game.
MIT’s Fluid Interface Group: wearable
device with a projector for deep
interactions with the environment
104. Bridging the Cyber-Physical Divide
Foursquare is an online application which
integrates a persons physical location and
social network.
Community of enthusiasts that share experiences of
self-tracking and measurement.
FitBit Community allows the
automated collection and
sharing of health-related data,
goals, and achievements
106. How do we design the Sensor Web?
Integration through shared semantics
OGC Sensor Web Enablement
W3C SSN ontology and Semantic Annotation
Interpretation through integration of
heterogeneous data and reasoning with prior
knowledge
Semantic Perception/Abstraction
Linked Open Data as prior knowledge
Scale through distributed local interpretation
“intelligence at the edge”
109. Vision of Sensor Web
Quickly discover sensors (secure or public) that can meet
my needs – location, observables, quality, ability to task
Obtain sensor information in a standard encoding that is
understandable by me and my software
Readily access sensor observations in a common manner,
and in a form specific to my needs
Task sensors, when possible, to meet my specific needs
Subscribe to and receive alerts when a sensor measures a
particular phenomenon
110. Principles of Sensor Web
Sensors will be web accessible
Sensors and sensor data will be discoverable
Sensors will be self-describing to humans and software
(using a standard encoding)
Most sensor observations will be easily accessible in real
time over the web
111. OGC SWE Services
Sensor Observation Service (SOS)
access sensor information (SensorML) and sensor
observations (O&M
Sensor Planning Service (SPS)
task sensors or sensor systems
Sensor Alert Service (SAS)
asynchronous notification of sensor events (tasks,
observation of phenomena)
Sensor Registries
discovery of sensors and sensor data
113. OGC SWE Languages
Sensor Model Language (SensorML)
Models and schema for describing sensor
characteristics
Observation & Measurement (O&M)
Models and schema for encoding sensor
observations
117. Sensor Web + Semantic Web
Semantic Web
The web of data where web content is processed
by machines, with human actors at the end of the
chain.
The web as a huge, dynamic, evolving database
of facts, rather than pages, that can be interpreted
and presented in many ways (mashups).
Fundamental importance of ontologies to describe
the fact that represents the data. RDF(S)
emphasises labelled links as the source of meaning:
essentially a graph model . A label (URI) uniquely
identifies a concept.
OWL emphasises inference as the source of
meaning: a label also refers to a package of logical
axioms with a proof theory.
Usually, the two notions of meaning fit.
Goal to combine information and services
for targeted purpose and new knowledge
Sensor Web
The internet of things made up of Wireless Sensor
Networks, RFID, stream gauges, orbiting satellites,
weather stations, GPS, traffic sensors, ocean buoys,
animal and fish tags, cameras, habitat monitors,
recording data from the physical world.
Today there are 4 billion mobile sensing devices
plus even more fixed sensors. The US National
Research Council predicts that this may grow to
trillions by 2020, and they are increasingly connected
by internet and Web protocols.
Record observations of a wide variety of
modalities: but a big part is time-series‟ of numeric
measurements.
The Open Geospatial Consortium has some web-
service standards for shared data access (Sensor
Web Enablement).
Goal is to open up access to real-time and
archival data, and to combine in applications.
118. So, what is a Semantic Sensor Web?
Reduce the difficulty and open up sensor networks by:
Allowing high-level specification of the data collection process;
Across separately deployed sensor networks;
Across heterogeneous sensor types; and
Across heterogeneous sensor network platforms;
Using high-level descriptions of sensor network capability; and
Interfacing to data integration methods using similar query and
capability descriptions.
To create a Web of Real Time Meaning!
119. W3C SSN Incubator Group
SSN-XG commenced: 1 March 2009
Chairs:
Amit Sheth, Kno.e.sis Center, Wright State University
Kerry Taylor, CSIRO
Amit Parashar Holger Neuhaus Laurent Lefort, CSIRO
Participants: 39 people from 20 organizations, including:
Universities in: US, Germany, Finland, Spain, Britain, Ireland
Multinationals: Boeing, Ericsson
Small companies in semantics, communications, software
Research institutes: DERI (Ireland), Fraunhofer (Germany),
ETRI (Korea), MBARI (US), SRI International (US), MITRE
(US), US Defense, CTIC (Spain), CSIRO (Australia), CESI
(China)
120. W3C SSN Incubator Group
Two main objectives:
The development of an ontology for describing
sensing resources and data, and
The extension of the SWE languages to support
semantic annotations.
126. Stimulus-Sensor-Observation
The SSO Ontology Design Pattern is developed following the principle of
minimal ontological commitments to make it reusable for a variety of application
areas.
Introduces a minimal set of classes and relations centered around the notions
of stimuli, sensor, and observations. Defines stimuli as the (only) link to the
physical environment.
Empirical science observes these stimuli using sensors to infer information
about environmental properties and construct features of interest.
129. SSN Sensor
A sensor can do (implements) sensing: that is, a sensor is any entity that can
follow a sensing method and thus observe some Property of a
FeatureOfInterest.
Sensors may be physical devices, computational methods, a laboratory setup
with a person following a method, or any other thing that can follow a Sensing
130. SSN Measurement Capability
Collects together measurement properties (accuracy, range, precision, etc) and
the environmental conditions in which those properties hold, representing a
specification of a sensor's capability in those conditions.
131. SSN Observation
An Observation is a Situation in which a Sensing method has been used to estimate or
calculate a value of a Property.
Links to Sensing and Sensor describe what made the Observation and how; links to
Property and Feature detail what was sensed; the result is the output of a Sensor; other
metadata gives the time(s) and the quality.
Different from OGC’s O&M, in which an “observation” is an act or event, although it also
provides the record of the event.
133. What SSN does not model
Sensor types and models
Networks: communication, topology
Representation of data and units of measurement
Location, mobility or other dynamic behaviours
Animate sensors
Control and actuation
….
134. Semantic Annotation of SWE
Recommended
technique via Xlink
attributes requires no
change to SWE
xlink:href - link to
ontology individual
xlink:role - link to
ontology class
xlink:arcrole - link to
ontology object property
135. How do we design the Sensor Web?
Integration through shared semantics
OGC Sensor Web Enablement
W3C SSN ontology and Semantic Annotation
Interpretation through integration of
heterogeneous data and reasoning with prior
knowledge
Semantic Perception/Abstraction
Linked Open Data as prior knowledge
Scale through distributed local interpretation
“intelligence at the edge”
136. 136
Interpretation of data
A primary goal of interconnecting devices and
collecting/processing data from them is to
create situation awareness and enable
applications, machines, and human users to
better understand their surrounding
environments.
The understanding of a situation, or context,
potentially enables services and applications
to make intelligent decisions and to respond to
the dynamics of their environments.
137. 137
Observation and measurement data
Source: W3C Semantic Sensor Networks, SSN Ontology presentation, Laurent Lefort et al.
138. 138
How to say what a sensor is and
what it measures?
Sink
node
Gateway
139. 139
Data/Service description frameworks
There are standards such as Sensor Web Enablement (SWE) set
developed by the Open Geospatial Consortium that are widely being
adopted in industry, government and academia.
While such frameworks provide some interoperability, semantic
technologies are increasingly seen as key enabler for integration of
IoT data and broader Web information systems.
141. 141
W3C SSN Ontology
makes observations
of this type
Where it is
What it
measures
units
SSN-XG ontologies
SSN-XG annotations
SSN-XG Ontology Scope
142. 142
Semantics and IoT data
Creating ontologies and defining data models is not enough
tools to create and annotate data
data handling components
Complex models and ontologies look good, but
design lightweight versions for constrained environments
think of practical issues
make it as compatible as possible and/or link it to the other existing
ontologies
Domain knowledge and instances
Common terms and vocabularies
Location, unit of measurement, type, theme, …
Link it to other resources
Linked-data
URIs and naming
143. 143
Semantics and sensor data
Source: W. Wang, P. Barnaghi, "Semantic Annotation and Reasoning for Sensor Data", In proceedings of the 4th European Conference on Smart
Sensing and Context (EuroSSC2009), 2009.
144. 144
Semantics and Linked-data
The principles in designing the linked data are
defined as:
using URI’s as names for things;
using HTTP URI’s to enable people to look up
those names;
provide useful RDF information related to URI’s
that are looked up by machine or people;
including RDF statements that link to other URI’s
to enable discovery of other related concepts of
the Web of Data;
146. 146
Myth and reality
#1: If we create an Ontology our data is interoperable
Reality: there are/could be a number of ontologies for a domain
Ontology mapping
Reference ontologies
Standardisation efforts
#2: Semantic data will make my data machine-understandable and my
system will be intelligent.
Reality: it is still meta-data, machines don’t understand it but can interpret it. It still does
need intelligent processing, reasoning mechanism to process and interpret the data.
#3: It’s a Hype! Ontologies and semantic data are too much overhead; we
deal with tiny devices in IoT.
Reality: Ontologies are a way to share and agree on a common vocabulary and knowledge;
at the same time there are machine-interpretable and represented in interoperable and re-
usable forms;
You don’t necessarily need to add semantic metadata in the source- it could be added to the
data at a later stage (e.g. in a gateway);
Legacy applications can ignore it or to be extended to work with it.
149. 149
SAX representation
SAX Pattern (blue) with word length of 20 and a vocabulary of 10 symbols
over the original sensor time-series data (green)
P. Barnaghi, F. Ganz, C. Henson, A. Sheth, "Computing Perception from Sensor Data",
in Proc. of the IEEE Sensors 2012, Oct. 2012.
fggfffhfffffgjhghfff
jfhiggfffhfffffgjhgi
fggfffhfffffgjhghfff
150. 150
Data Processing Framework
fggfffhfffffgjhghfff dddfffffffffffddd cccddddccccdddccc aaaacccaaaaaaaaccccdddcdcdcdcddasddd
PIR Sensor Light Sensor
Temperature
Sensor
Raw sensor data
stream
Raw sensor data
stream
Raw sensor data
stream
Attendance Phone
Hot
Temperature
Cold
Temperature
Bright
Day-time
Night-time
Office room
BA0121
On going
meeting
Window has
been left open
….
Temporal data
(extracted from
SSN descriptions)
Spatial data
(extracted from
SSN descriptions)
Thematic
data
(low level
abstractions)
Parsimonious
Covering Theory
Observations
Perceptions
Domain knowledge
SAX Patterns
Raw Sensor Data
(Annotated with SSN
Ontology)
…
….
Perception
Computation
High-level
Perceptions
151. 151
SensorSAX
F. Ganz, P. Barnaghi, F. Carrez, “Information Abstraction for Heterogeneous Real World Internet Data”, Feb. 2013.
152. 152
Evaluation results of abstraction
creation
F. Ganz, P. Barnaghi, F. Carrez, “Information Abstraction for Heterogeneous Real World Internet Data”, Feb. 2013.
153. 153
Data size reduction
F. Ganz, P. Barnaghi, F. Carrez, “Information Abstraction for Heterogeneous Real World Internet Data”, Feb. 2013.
154. 154
Enabling the Internet of Things
- Diversity range of applications
- Interacting with large number of
devices with various types
-Multiple heterogeneous networks
-Deluge of data
-Processing and interpretation of the
IoT data
155. 155
Challenges and opportunities
Providing infrastructure
Publishing, sharing, and access solutions on a global scale
Indexing and discovery (data and resources)
Aggregation and fusion
Trust, privacy and security
Data mining and creating actionable knowledge
Integration into services and applications in e-health, the public sector,
retail, manufacturing and personalized apps.
Mobile apps, location-based services, monitoring control etc.
New business models
156. 156
Part 6: Cognitive aspects of
knowledge representation, reasoning
and perception
Image source: CISCO
157. Abstraction provides the ability to interpret and synthesize information
in a way that affords effective understanding and communication of
ideas, feelings, perceptions, etc. between machines and people.
Abstraction
158. People are excellent at abstraction;
of sensing and interpreting stimuli
to understand and interact with the
world.
The process of interpreting stimuli
is called perception; and studying
this extraordinary human capability
can lead to insights for developing
effective machine perception.
Abstraction
162. What can we learn from
Cognitive Models of
Perception?
A-priori background knowledge is a key
enabler
Perception is a cyclical, active process
Ulric Neisser (1976)Ulric Neisser (1976) Richard Gregory (1997)Richard Gregory (1997)
163. Is Semantic Web up to the task
of modeling perception?
Representation
Heterogeneous sensors, sensing, and observation
records
Background knowledge (observable properties,
objects/events, etc.)
Inference
Explain observations (hypothesis building)
Focus attention by seeking additional stimuli (that
discriminate between explanations)
Difficult Issues to Overcome
Perception is an inference to the best explanation
Handle streaming data
Real-time processing (or nearly)
164. Both people and machines are capable of observing
qualities, such as redness.
* Formally described in a sensor/ontology (SSN ontology)
observes
Observer Quality
165. The ability to perceive is afforded through the use of
background knowledge, relating observable qualities to
entities in the world.
* Formally described in
domain ontologies
(and knowledge bases)
inheres in
Quality
Entity
166. With the help of sophisticated inference, both people and
machines are also capable of perceiving entities, such as
apples.
the ability to degrade gracefully with incomplete information
the ability to minimize explanations based on new
information
the ability to reason over data on the Web
fast (tractable)
perceives
EntityPerceiver
168. The ability to perceive efficiently is afforded through the
cyclical exchange of information between observers and
perceivers.
Traditionally called the
Perceptual Cycle
(or Active Perception)
sends
focus
sends
observation
Observer
Perceiver
170. 1970’s – Perception is an active, cyclical process of
exploration and interpretation.
- Nessier’s Perception Cycle
1980’s – The perception cycle is driven by
background knowledge in order to generate and test
hypotheses.
- Richard Gregory (optical illusions)
1990’s – In order to effectively test hypotheses,
some observations are more informative than others.
- Norwich’s Entropy Theory of Perception
Cognitive Theories of Perception
171. Key Insights
Background knowledge plays a crucial role in perception; what we
know (or think we know/believe) influences our perception of the
world.
Semantics will allow us to realize computational models of
perception based on background knowledge.
Contemporary Issues
Internet/Web expands our background knowledge to a global
scope; thus our perception is global in scope
Social networks influence our knowledge and beliefs, thus
influencing our perception
172. observes
inheres in
Integrated together, we have an general model – capable of
abstraction – relating observers, perceivers, and background
knowledge.
perceives
sends
focus
sends
observation
Observer Quality
EntityPerceiver
173. Ontology of Perception – as an extension of SSN
Provides abstraction of sensor data through perceptual
inference of semantically annotated data
174. Prior Knowledge
W3C SSN Ontology Bi-partite Graph
Prior knowledge conformant to SSN ontology (left),
structured as a bipartite graph (right)
175. Explanation is the act of accounting for sensory observations (i.e.,
abstraction); often referred to as hypothesis building.
Observed Property: A property that has been observed.
ObservedProperty ≡ ∃ssn:observedProperty—
.{o1} ⊔ … ⊔
∃ssn:observedProperty—
.{on}
Explanatory Feature: A feature that explains the set of observed
properties.
ExplanatoryFeature ≡ ∃ssn:isPropertyOf—
.{p1} ⊓ … ⊓
∃ssn:isPropertyOf—
.{pn}
Semantics of Explanation
176. Example
Assume the properties elevated blood pressure and
palpitations have been observed, and encoded in RDF
(conformant with SSN):
ssn:Observation(o1), ssn:observedProperty(o1, elevated blood
pressure)
ssn:Observation(o2), ssn:observedProperty(o2, palpitations)
Given these observations, the following
ExplanatoryFeature class is constructed:
ExplanatoryFeature ≡ ∃ssn:isPropertyOf—
.{elevated blood
pressure} ⊓ ∃ssn:isPropertyOf—
.{palpitations}
Given the KB, executing the query
ExplanatoryFeature(?y) can infer the features,
Hypertension and Hyperthyroidism, as explanations:
ExplanatoryFeature(Hypertension)
ExplanatoryFeature(Hyperthyroidism)
Semantics of Explanation
177. Discrimination is the act of deciding how to narrow down the multitude of
explanatory features through further observation.
Expected Property: A property is expected with respect to (w.r.t.) a set
of features if it is a property of every feature in the set.
ExpectedProperty ≡ ∃ssn:isPropertyOf.{f1} ⊓ … ⊓ ∃ssn:isPropertyOf.{fn}
NotApplicable Property: A property is not-applicable w.r.t. a set of
features if it is not a property of any feature in the set.
NotApplicableProperty ≡ ¬∃ssn:isPropertyOf.{f1} ⊓ … ⊓
¬∃ssn:isPropertyOf.{fn}
Discriminating Property: A property is discriminating w.r.t. a set of
features if it is neither expected nor not-applicable.
DiscriminatingProperty ≡ ¬ExpectedProperty ⊓ ¬NotApplicableProperty
Semantics of Discrimination
178. Example
Given the explanatory features from the previous
example, Hypertension and Hyperthyroidism, the
following classes are constructed:
ExpectedProperty ≡ ∃ssn:isPropertyOf.{Hypertension} ⊓
∃ssn:isPropertyOf.{Hyperthyroidism}
NotApplicableProperty ≡ ¬∃ssn:isPropertyOf.{Hypertension} ⊓
¬∃ssn:isPropertyOf.{Hyperthyroidism}
Given the KB, executing the query
DiscriminatingProperty(?x) can infer the property clammy
skin as discriminating:
DiscriminatingProperty(clammy skin)
Semantics of Discrimination
179. How do we design the Sensor Web?
Integration through shared semantics
OGC Sensor Web Enablement
W3C SSN ontology and Semantic Annotation
Interpretation through integration of
heterogeneous data and reasoning with prior
knowledge
Semantic Perception/Abstraction
Linked Open Data as prior knowledge
Scale through distributed local interpretation
“intelligence at the edge”
180. Efficient Algorithms for IntellegO
Use of OWL-DL reasoner too resource-intensive for use in
resource constrained devices (such as sensor nodes, mobile
phones, IoT devices)
Runs out of resources for problem size (prior knowledge) > 20
concepts
Asymptotic complexity: O(n3
) [Experimentally determined]
To enable their use on resource-constrained devices, we now
describe algorithms for efficient inference of explanation and
discrimination.
These algorithms use bit vector encodings and operations,
leveraging a-priori knowledge of the environment.
181. Efficient Algorithms for IntellegO
Semantic (RDF)
Encoding
Bit Vector Encoding
Lower
Lift
First, developed lifting and
lowering algorithms to translate
between RDF and bit vector
encodings of observations.
182. Efficient Algorithms for IntellegO
Explanation Algorithm
Discrimination Algorithm
Utilize bit vector operators to efficiently
compute explanation and
discrimination
Explanation: Use of the bit vector AND
operation to discover and dismiss those
features that cannot explain the set of
observed properties
Discrimination: Use of the bit vector AND
operation to discover and indirectly assemble
those properties that discriminate between a
set of explanatory features. The discriminating
properties are those that are determined to be
neither expected nor not-applicable
183. Efficient Algorithms for IntellegO
Evaluation: The bit vector encodings and algorithms yield significant and
necessary computational enhancements – including asymptotic order of
magnitude improvement, with running times reduced from minutes to
milliseconds, and problem size increased from 10’s to 1000’s.
186. 186
J. McCarthyJ. McCarthy M. WeiserM. Weiser
D.
Engelbart
D.
Engelbart J. C. R. LickliderJ. C. R. Licklider
Similar Visions of Computing
187. 187
If people do not believe that mathematics is simple, it is only
because they do not realize how complicated life is.
- John von Neumann
If people do not believe that mathematics is simple, it is only
because they do not realize how complicated life is.
- John von Neumann
Computational paradigms have always dealt with a
simplified representations of the real-world…
Computational paradigms have always dealt with a
simplified representations of the real-world…
Algorithms work on these simplified
representations
Algorithms work on these simplified
representations
Solutions from these algorithms are transcended back
to the real-world by humans as actions
Solutions from these algorithms are transcended back
to the real-world by humans as actions
http://ngs.ics.uci.edu/blog/?p=1501
Grand challenges in the real-world are
complex
188. What has changed now?
Data Processing and Semantics for Advanced Internet of Things (IoT) Applications, WIMS
2013, Madrid, Spain
There are more devices connected to the internet than
the entire human population.
There are more devices connected to the internet than
the entire human population.
Today we have around 10 billion devices connected
to the internet making it an era of IoT (Internet of Things)
Today we have around 10 billion devices connected
to the internet making it an era of IoT (Internet of Things)
http://www.cisco.com/web/about/ac79/docs/innov/IoT_IBSG_0411FINAL.pdf
189. 189
What has not changed?What has not changed?
We need computational paradigms to tap
into the rich pulse of the human populace
We need computational paradigms to tap
into the rich pulse of the human populace
We are still working on the simpler representations of the real-
world!
We are still working on the simpler representations of the real-
world!
Represent, capture, and compute with richer and
fine-grained representations of real-world
problems
Represent, capture, and compute with richer and
fine-grained representations of real-world
problems
What should change?What should change?
190. PCS Computing: Asthma Scenario
190
Sensordrone – for monitoring
environmental air quality
Wheezometer – for monitoring
wheezing sounds
Can I reduce my asthma attacks at night?
What are the triggers?
What is the wheezing level?
What is the propensity toward asthma?
What is the exposure level over a day?
What is the air quality indoors?
Commute to Work
Personal
Public Health
Population Level
Closing the window at home
in the morning and taking an
alternate route to office may
lead to reduced asthma attacks
Actionable
Information
Actionable
Information
191. Personal, Public Health, and
Population Level Signals for Monitoring
Asthma
Data Processing and Semantics for Advanced Internet of Things (IoT) Applications, WIMS
2013, Madrid, Spain
ICS= inhaled corticosteroid, LABA = inhaled long-acting beta2-agonist, SABA= inhaled short-acting beta2-agonist ;
*consider referral to specialist
Asthma Control
and Actionable Information
Sensors and their observations
for understanding asthma
192. Asthma Early Warning Model
Data Processing and Semantics for Advanced Internet of Things (IoT) Applications, WIMS
2013, Madrid, Spain
192
Personal
Level
Signals
Personal
Level
Signals
Societal Level
Signals
Societal Level
Signals
(Personal Level Signals)
(Personalized
Societal Level Signal)
(Societal Level Signals)
Societal Level
Signals Relevant to
the Personal Level
Societal Level
Signals Relevant to
the Personal Level
Personal Level Sensors
(mHealth**)
QualifyQualify QuantifyQuantify
Action
Recommendation
Action
Recommendation
What are the features influencing my asthma?
What is the contribution of each of these features?
How controlled is my asthma? (risk score)
What will be my action plan to manage asthma?
StorageStorage
Societal Level Sensors
Asthma Early Warning Model (AEWM)
Query AEWM
Verify & augment
domain knowledge
Recommended
Action
Action
Justification
*http://www.slideshare.net/jain49/eventshop-120721, ** http://www.youtube.com/watch?v=btnRi64hJp4
(EventShop*)
193. Health Signal Extraction to
Understanding
Data Processing and Semantics for Advanced Internet of Things (IoT) Applications, WIMS
2013, Madrid, Spain
Population Level
Personal
Wheeze – Yes
Do you have tightness of chest? –Yes
ObservationsPhysical-Cyber-Social System Health Signal Extraction Health Signal Understanding
<Wheezing=Yes, time, location>
<ChectTightness=Yes, time, location>
<PollenLevel=Medium, time, location>
<Pollution=Yes, time, location>
<Activity=High, time, location>
Wheezing
ChectTightness
PollenLevel
Pollution
Activity
Wheezing
ChectTightness
PollenLevel
Pollution
Activity
RiskCategory
<PollenLevel, ChectTightness, Pollution,
Activity, Wheezing, RiskCategory>
<2, 1, 1,3, 1, RiskCategory>
<2, 1, 1,3, 1, RiskCategory>
<2, 1, 1,3, 1, RiskCategory>
<2, 1, 1,3, 1, RiskCategory>
.
.
.
Expert
Knowledge
Background
Knowledge
tweet reporting pollution level
and asthma attacks
Acceleration readings from
on-phone sensors
Sensor and personal
observations
Signals from personal, personal
spaces, and community spaces
Risk Category assigned by doctors
Qualify
Quantify
Enrich
Outdoor pollen and pollution
Public Health
Well Controlled - continue
Not Well Controlled – contact
nurse
Poor Controlled – contact doctor
194. 194
J. McCarthyJ. McCarthy M. WeiserM. Weiser
D.
Engelbart
D.
Engelbart J. C. R. LickliderJ. C. R. Licklider
197. SSN Application: Weather
50% savings in sensing
resource requirements during
the detection of a blizzard
Order of magnitude resource
savings between storing observations
vs. relevant abstractions
200. Real-Time Feature Streams
Demo: http://www.youtube.com/watch?
v=_ews4w_eCpg
Data Processing and Semantics for Advanced Internet of Things (IoT) Applications, WIMS
2013, Madrid, Spain
201. SSN Application: Fire Detection
Weather Application
SECURE: Semantics-empowered Rescue
Environment
(detect different types of fires)
DEMO: http://www.youtube.com/watch?v=in2KMkD_uqg
202. SSN Application: Health Care
MOBILEMD: Mobile app to help reduce re-
admission of patients with Chronic Heart Failure
204. SSN Application: Health Care
Active Monitoring PhaseActive Monitoring Phase
Are you feeling lightheaded?Are you feeling lightheaded?
Are you have trouble taking deep
breaths?
Are you have trouble taking deep
breaths?
yesyes
yesyes
Have you taken your Methimazole
medication?
Have you taken your Methimazole
medication?
Do you have low blood pressure?Do you have low blood pressure?
yesyes
• Abnormal heart rate
• Clammy skin
• Lightheaded
• Trouble breathing
• Low blood pressure
• Panic Disorder
• Hypoglycemia
• Hyperthyroidism
• Heart Attack
• Septic Shock
Observed Symptoms Possible Explanations
nono
Active Sensors – blood pressure, weight scale, pulse
oxymeterDemo: http://www.youtube.com/watch?v=btnRi64hJp4
205. Domain ExpertsDomain Experts
ColdWeatherColdWeather
PoorVisibilityPoorVisibility
SlowTrafficSlowTraffic
IcyRoadIcyRoad
Declarative domain knowledge
Causal
knowledge
Linked Open Data
ColdWeather(YES/NO)IcyRoad (ON/OFF) PoorVisibility (YES/NO)SlowTraffic (YES/NO)
1 0 1 1
1 1 1 0
1 1 1 1
1 0 1 0
Domain ObservationsDomain Observations
Domain
Knowledge
Domain
Knowledge
Structure and parameters
205
WinterSeasonWinterSeason Otherknowledge
Correlations to causations using
Declarative knowledge on the
Semantic Web
kTraffic
206. Traffic jamTraffic jam
Link
Description
Link
Description
Scheduled
Event
Scheduled
Event
traffic jamtraffic jambaseball gamebaseball game
Add missing random variables
Time of dayTime of day
bad weather CapableOf slow traffic
bad weatherbad weather
Traffic data from sensors deployed on road
network in San Francisco Bay Area
time of daytime of day
traffic jamtraffic jambaseball gamebaseball game
time of daytime of day
slow trafficslow traffic
Three Operations: Complementing graphical model structure extraction
Add missing links bad weatherbad weather
traffic jamtraffic jambaseball gamebaseball game
time of daytime of day
slow trafficslow traffic
Add link direction
bad weatherbad weather
traffic jamtraffic jambaseball gamebaseball game
time of daytime of day
slow trafficslow traffic
go to baseball game Causes traffic jam
Knowledge from ConceptNet5
traffic jam CapableOfoccur twice each day
traffic jam CapableOf slow traffic
206
207. 207
Scheduled
Event
Scheduled
Event
Active EventActive Event
Day of weekDay of week
Time of dayTime of day
delaydelay
Travel
time
Travel
time
speedspeed
volumevolume
Structure extracted form
traffic observations
(sensors + textual) using
statistical techniques
Scheduled
Event
Scheduled
Event
Active EventActive Event
Day of weekDay of week
Time of dayTime of day
delaydelay
Travel
time
Travel
time
speedspeed
volumevolume
Bad WeatherBad Weather
Enriched structure which has
link directions and new
nodes such as “Bad Weather”
potentially leading to better
delay predictions
kTraffic
208. SemMOB
Data Processing and Semantics for Advanced Internet of Things (IoT) Applications, WIMS
2013, Madrid, Spain
First responders have limited time to analyze
sensor (on and around them) observations.
First responders have limited time to analyze
sensor (on and around them) observations.
Fire fighters need to combine prior knowledge of
fires and its behavior, extinguishers, and floor
plan of the building for rescue strategy and operations.
Fire fighters need to combine prior knowledge of
fires and its behavior, extinguishers, and floor
plan of the building for rescue strategy and operations.
There are a variety of sensors used to monitor
vitals of firefighters, location, and poisonous gases.
There are a variety of sensors used to monitor
vitals of firefighters, location, and poisonous gases.
O2
O2
Heart rateHeart rate
COCO
CO2
CO2
GPSGPS
AccelerometerAccelerometer CompassCompass
FootstepsFootsteps
There is a team of them making
it further difficult for analysis
There is a team of them making
it further difficult for analysis
Semantic Web allow us to describe the domain, sensors, and
first responders -- apply reasoning techniques to derive actionable
insights
Semantic Web allow us to describe the domain, sensors, and
first responders -- apply reasoning techniques to derive actionable
insights
Sensors on each responder provides
different view of the event and they need
to register dynamically as the firefighters
arrive.
Sensors on each responder provides
different view of the event and they need
to register dynamically as the firefighters
arrive.
209. SemMOB
Data Processing and Semantics for Advanced Internet of Things (IoT) Applications, WIMS
2013, Madrid, Spain
PlatformPlatform SensingDeviceSensingDevice
System_AGT
C1
System_AGT
C1
Sensor_LSM1
1
Sensor_LSM1
1
SensorOutputSensorOutput
O2Concentraction_outputO2Concentraction_output
Measuremen
tCapability
Measuremen
tCapability
MeasurementCa
pability_LSM11
MeasurementCa
pability_LSM11
PropertyProperty
concentratio
n
concentratio
n
onPlatform
hasOutput
hasMeasurementCapability
Observes
ObservationObservation
O2ConcentrationO2Concentration
ObservedProperty
O2Obs_1505201
2
O2Obs_1505201
2
observationResultTimehttp://geonames.org/629864
0
http://geonames.org/629864
0
hasLocation
name
xsd:floatxsd:float
latlong
Dayton, Dayton-Wright Brothers Airport^^xsd:string
FeatureFeature
Demo: http://www.youtube.com/watch?v=nX11OqgtQlw
210. LOKILAS
(LOst Key Identification, Localization and Alert System)
Data Processing and Semantics for Advanced Internet of Things (IoT) Applications, WIMS
2013, Madrid, Spain
211. LOKILAS
Data Processing and Semantics for Advanced Internet of Things (IoT) Applications, WIMS
2013, Madrid, Spain
212. LOKILAS
Data Processing and Semantics for Advanced Internet of Things (IoT) Applications, WIMS
2013, Madrid, Spain
213. Future work
Creating ontologies and defining data models are not
enough
tools to create and annotate data
Tools for publishing linked IoT data
Designing lightweight versions for constrained
environments
think of practical issues
make it as much as possible compatible and/or link it to the other
existing ontologies
Linking to domain knowledge and other resources
Location, unit of measurement, type, theme, …
Linked-data
URIs and naming
214. Some of the open issues
Efficient real-time IoT resource/service
query/discovery
Directory
Indexing
Abstraction of IoT data
Pattern extraction
Perception creation
IoT service composition and compensation
Integration with existing Web services
Service adaptation
215. Selected references
Payam Barnaghi, Wei Wang, Cory Henson, Kerry Taylor, "Semantics for the Internet of Things: early progress and back to the future",
(to appear) International Journal on Semantic Web and Information Systems (special issue on sensor networks, Internet of Things and
smart devices), 2012.
Atzori, L., Iera, A. & Morabito, G. , “The Internet of Things: A survey”, Computer Networks, Volume 54, Issue 15, 28 October 2010, 2787-
2805.
Suparna De, Tarek Elsaleh, Payam Barnaghi , Stefan Meissner, "An Internet of Things Platform for Real-World and Digital Objects",
Journal of Scalable Computing: Practice and Experience, vol 13, no.1, 2012.
Suparna De, Payam Barnaghi, Martin Bauer, Stefan Meissner, "Service modelling for the Internet of Things", in Proceedings of the
Conference on Computer Science and Information Systems (FedCSIS), pp.949-955, Sept. 2011.
Cory Henson, Amit Sheth, and Krishnaprasad Thirunarayan, “Semantic Perception: Converting Sensory Observations to Abstractions”,
IEEE Internet Computing, Special Issue on Context-Aware Computing, March/April 2012.
Payam Barnaghi, Frieder Ganz, Cory Henson, Amit Sheth, “Computing Perception from Sensor Data”, In proceedings of the 2012 IEEE
Sensors Conference, Taipei, Taiwan, October 28-31, 2012.
Michael Compton et al, “The SSN Ontology of the W3C Semantic Sensor Network Incubator Group”, Journal of Web Semantics, 2012.
Harshal Patni, Cory Henson, and Amit Sheth , “Linked Sensor Data”, in Proceedings of 2010 International Symposium on Collaborative
Technologies and Systems (CTS 2010), Chicago, IL, May 17-21, 2010.
Amit Sheth, Cory Henson, and Satya Sahoo , “Semantic Sensor Web IEEE Internet Computing”, vol. 12, no. 4, July/August 2008, pp. 78-
83.
Wei Wang, Payam Barnaghi, Gilbert Cassar, Frieder Ganz, Pirabakaran Navaratnam, "Semantic Sensor Service Networks", (to appear)
in Proceedings of the IEEE Sensors 2012 Conference, Taipei, Taiwan, October 2012.
Wang W, De S, Toenjes R, Reetz E, Moessner K, "A Comprehensive Ontology for Knowledge Representation in the Internet of Things",
International Workshop on Knowledge Acquisition and Management in the Internet of Things (KAMIoT 2012) in conjunction with IEE
IUCC-2012, Liverpool, UK. Liverpool. 25-27 June, 2012.
216. Some useful links related to IoT
Internet of Things, ITU
http://www.itu.int/osg/spu/publications/internetofthings/InternetofThings_summary.pdf
IoT Comic Book
http://www.theinternetofthings.eu/content/mirko-presser-iot-comic-book
Internet of Things Europe, http://www.internet-of-things.eu/
Internet of Things Architecture (IOT-A)
http://www.iot-a.eu/public/public-documents
W3C Semantic Sensor Networks
http://www.w3.org/2005/Incubator/ssn/XGR-ssn-20110628/
Hinweis der Redaktion
Scalability and interoperability problems
Intelligence at the edge or hub; still no good answer; could be app or design dependent
Traditional networking: host to host Now: data-oriented communication; looking for data, not the host providing the data unless you want to manage the node
IoT data different from traditional content; transient, small. The more important thing is how to find the service that can provide the data; could be huge stream of data
Declarative knowledge + statistical correlation This slide illustrates the three operations to enrich the correlation structure extracted using statistical methods These operations utilize declarative knowledge form ConceptNet5 as shown in each step
Statistical correlation structure shown above The enriched structure is shown below The enrichment of the graphical model will potentially allow us to capture the domain precisely and also improve our prediction as the model would get closer to the underlying probabilistic distribution in the real-world Log-Likelihood score is one way of quantifying how good a structure is based on the observed data There may be many candidate structures extracted from data which result in the log likelihood score Declarative knowledge will help us ground statistical models to reality which will allow us to pick one structure over the other Pramod Anantharam, Krishnaprasad Thirunarayan and Amit Sheth, 'Traffic Analytics using Probabilistic Graphical Models Enhanced with Knowledge Bases,' 2nd International Workshop on Analytics for Cyber-Physical Systems (ACS-2013) at SIAM International Conference on Data Mining (SDM13), pp. 13--20, Texas, USA, May 2-4, 2013. We stopped at structure extraction for our workshop paper (SIAM ACS workshop) since the declarative knowledge we used (ConceptNet5) and statistical model (nodes and edges) are at the same level of abstraction