Semantic Web Services for Smart Devices in a

1
Semantic Web Servicesfor Smart Devicesin a
Global Understanding Environment(SmartResource
)

• Vagan Terziyan
• Industrial Ontologies Group
• Agora Center, University of Jyväskylä
• HCISWWA , November 7, 2003, Catania (Sicily),
  Italy

http//www.cs.jyu.fi/ai/OntoGroup/index.html
2
Content

• Resources in Semantic Web and Beyond
• Global Understanding Environment
• Resource Adaptation
• Remote Diagnostics of Resources
• Resource Maintenance and Networking

3
MAIN RESEARCH OBJECTIVE

• Our intention is to make resources (Web
  documents and services, industrial devices, human
  experts, etc.) active in a sense that they can
  analyze their state independently from other
  systems and applications, initiate and control
  own maintenance proactively. Resource state can
  provide knowledge about resource condition,
  whereas both resource condition and goal of the
  resource will result in certain behavior of
  active resource towards effective and predictive
  maintenance.

4
Self-maintenance

• Do not expect that someone cares about you, take
  care yourself even if you are just an industrial
  device !
• You should be proactive enough to realize that
  you exist and want to be in a good shape
• You should be sensitive enough to feel your own
  state and condition
• You should be smart enough to understand that
  you need some maintenance.

5
Resource Agents
1. I feel bad, temperature 40, pain in stomach,
Who can advise what to do ?
3. Hey, I have some pills for you
2. Yeah, your condition is not good. You need
urgent help
6
Industrial Resources

• Classes of resources in maintenance systems
• Device - machines, equipment, etc.
• Processing Unit embedded, local and remote
  systems, for monitoring, diagnostics and control
  over devices
• Human (Expert) users of the system, operators,
  maintenance experts

7
Research Challenges

• Resource Adaptation and Interoperability
  (Semantic Web)
• Unify data representation for heterogeneous
  environment
• Provide basis for communication
• Resource Proactivity (Agent Technology)
• Design of framework for delivering
  self-maintained resources to industrial systems
• Resource Interaction (Peer-to-Peer, Web Services
  technologies)
• Design of goal-driven co-operating resources
• Resource-to-Resource communication models in
  distributed environment (in the context of
  industrial maintenance)
• Design of communication infrastructure

8
GUN Concept
Global Understanding eNvironment
9
RESOURCE ADAPTATION

• First Slice of Gun Architecture

10
Targets
11
Diversity of Resources
GUN (Global Understanding eNvironment) concept
considers notion of resource in a very general
sense. Types of resources that can be integrated
into GUN are not limited only to digital
documents and database content. Real-world
objects can be also represented as resources
capable, for example, to accept and respond to
queries, interact with other resources in order
to achieve own goals.
Generic GUN-resource
12
Generic Resource Adapter
Semantic wrapping of resource actions
translation of external messages into
resource-native formats
Generic Adapter
configuration
Semantic Layer
Resource-specific messaging
Messaging Layer
Communication-specific connector of a resource
Connectivity Layer
GUN environment
GUN-resource
The integration requires development of the
Generic Resource Adapter, which will provide
basic tools for adaptation of the resource to
Semantic Environment. It should have open modular
architecture, extendable for support of variety
low- and high-level protocols of the resources
and semantic translation modules specific for
every resource (e.g. human, device, database).
Generic Resource Adapter must be configurable
for individual resource. Configuration includes
setting up of communication specific parameters,
choosing messaging mechanism, establishing
messaging rules for the resource and providing a
semantic description of the resource interface.
13
Semantic adapter for Devices
If to consider field devices as data sources,
then information to be annotated is data from
sensors, control parameters and other data that
presents relevant state of the device for the
maintenance process. Special piece of
device-specific software (Semantic Adapter) is
used for translation of raw diagnostic data into
standardized maintenance data based on shared
ontology.
Adapter
Shared ontology
Device-specific calls
Semantic message
14
Semantic adapters for Services
The purpose of Service Semantic Adapter is to
make service component semantic web enabled,
allowing communication with service on semantic
level regardless of the incompatibility on
protocol levels, both low-level (data
communication protocol) and high-level (messaging
rules, message syntax, data encoding, etc.).
Adapter
Shared ontology
Service-specific calls
Semantic message
15
Semantic Adapters for Human-experts
Human in the system is an initiator and
coordinator of the resource maintenance process.
The significant challenge is development of
effective and handy tools for human interaction
with Semantic Web-based environment. Human will
interact with the environment via special
communication and semantic adapter.
GUN-resource
Semantic message that will be visualized
Shared ontology
Action translated into semantic message
User interface
Human
16
REMOTE DIAGNOSTICS

• Second Slice of Gun Architecture

17
Goals

• Development of remote diagnostic model with
• semantic-based communication
• expert (human) and diagnostic (Web) service
• with learning capabilities

18
Device local platform

• Device is a sample of a device, which state is to
  be automatically annotated with diagnosis. It
  is supplied with Local Platform, which contains
  Local Alarm Service and History Data Storage.

Device

• Local Alarm Service is a local device-specific
  algorithm capable to detect alarm states of the
  Device

• History Data is collected by Device via the
  maintenance ontology for history data
  representation

Local Alarm Service
Device state data
History Data Storage
Local Platform
19
Services (are able to learn)

• Service is a standalone diagnostic algorithm
  capable to learn Diagnostic (Classification,
  Prediction) Model of an expert based on labelled
  history data about the device state.

Service
Diagnostic model
Learning sample
20
Device Expert interactions
Expert

• Accepts semantic description of device state and
  can respond with classification label (semantic
  description of diagnosis)
• Can make semantic query to request device-state
  data (also labeled history data), get response
  from Device and provide own label for observed
  device state

Device
History data
Expert
21
Device Service interactions

• Service presents to a Device possibility to use
  it as a tool for self-diagnostics.

Service accepts semantic description of device
state from a Device and responds with
classification label obtained using existing
learned classification model

• If classification model has to be built first (no
  model yet) than perform learning

• Request data required for learning using semantic
  query
• Build (via a machine learning technique) a
  classification model
• Notify Device about readiness to perform
  diagnostics

22
Device Service, learning
Device
History data
Service
23
Device Service, servicing
Device
History data
Service
Diagnostic model
24
System structure
Expert
Simple remote diagnostic model with
semantic-based communication, expert and
diagnostic service with learning capabilities.
Labelled data
Watching and querying diagnostic data
Querying diagnostic results
Device
Service
Labelled data
History data
Querying data for learning
Learning sample and Querying diagnostic results
Diagnostic model
25
MAINTENANCE NETWORKING

• Third Slice of Gun Architecture

26
Networking
Expert
Device
Service
27
Goals

• Develop network infrastructure for resource
  maintenance system
• Support global experience reuse
• Support automated search of potential partners
  for services and resources (devices)
• Support collaborative resource diagnostics by
  multiple services and servicing multiple
  resources by one service.

28
P2P networking
- highly scalable
- fault-tolerable

• supports dynamic changes
• of network structure

Why to interact?

• does not need
• administration

resource summarizes opinions from multiple
services
service learns from multiple teachers
one service for multiple similar clients
services exchange lists of clients
resources exchange lists of services
29
Notice boards
Component advertisement solution
Client 3
Client 2
Allows search for new partners
Source of new entry points into P2P network
Client 1
Allows automated search based on semantic
profiles
Service 3
Service 1
Service 2
30
P2P semantic resource discovery

• P2P network formation through Notice Boards
• Search for necessary partners in P2P network
  according to their semantic descriptions
• Establishment of additional P2P links via
  exchanging addresses between partners

31
Discovery sample scenario

• Number of queried peers is restricted due to
• superhub based structure
• query forwarding mechanism based on
• analysis of semantic profile

Resource
32
Devices multiple services
Device will support service composition in form
of ensembles using own models of service quality
estimation. Service composition is made with goal
of increasing diagnostic performance.
Device
Labelled data
Service
Service
33
Services multiple devices
Service
Diagnostic model
1
Diagnostic model
n
Device
Device
Device
Labelled data
Device
Device
Device
Labelled data
Labelled data
Labelled data
Labelled data
Labelled data
34
Results of Networking

• Decentralized environment that integrates
• many devices,
• many services,
• many human experts
• and supports

Establishment of new peer-to-peer links through
NoticeBoards, advertisement mechanism
Exchange of contact lists between neigbor peers
Semantic based discovery of necessary network
components
Interaction One service many devices
Interaction One device many services
35
Device-to-Device opinion exchange
Device will be able to derive service quality
estimates basing on analysis of opinions of
other devices and trust to them.
Service 1
Service 2
Service quality evaluations
?
Device
?
6
8
trust 100
Device 1
trust 2
1
Device 2
4
36
Service-to- Service model exchange and
integration
Diagnostic models integration entails creation of
a more complex model extension or a service with
new diagnostic model
37
Certification
Support for certification authorities in the
network. Certificates gained by services will be
used by devices for optimal service search and
selection. Device makes its decision taking into
account also its own service quality evaluations.
Service 1
Service 2
Service 3
5
3
4
Device
6
1
trust
2
Own evaluations
Certifying party
38
Maintenance executive services
Support for maintenance services that can
influence on device state and perform
maintenance actions upon it (automated control
system, maintenance personnel). They complete the
minimal working set of maintenance system
components.
Service
diagnosis
data
Control
Device
control
39
Business Models
Noticeboard owner
Service
opinion cost 80
1-day advertisement 300
platform package 3000
Device
search service 80/item
expert support 40/hour
service cost 10/hour
certification 3000
service teaching 45/min
platform hosting 5/day
1000 new service addresses 40
?
Certifying party
New players are possible
40
Concluding Remark

• Among recent initiatives aimed at development of
  adoption of open information standards for
  operations and maintenance and implementation of
  interoperable cooperative industrial environments
  are
• MIMOSA (Machinery Information Management Open
  System Alliance)1. The project consortium
  pretends to build an open, industry-built, robust
  Enterprise Application Integration and
  condition-based maintenance specifications.
• PROTEUS2, funded by industrial companies and
  led with a goal to develop a generic
  maintenance-oriented platform for industry.
• These initiatives are very expensive, labor and
  resource consuming, and still does not attempt to
  apply and benefit from the Semantic Web
  technology. We believe however that without
  comprehensive metadata description framework,
  ontologies and open knowledge/semantics
  representation standards their results will be
  just next consortium-wide standards, rather than
  comprehensive, flexible and extensible framework.
• 1 http//www.mimosa.org/
• 2 http//www.proteus-iteaproject.com/