industrial ontologies group university of jyväskylä context-policy-configuration: paradigm of...
Post on 21-Dec-2015
214 views
TRANSCRIPT
Industrial Ontologies Group
University of JyväskyläUniversity of Jyväskylä
CONTEXT-POLICY-CONFIGURATION:CONTEXT-POLICY-CONFIGURATION:Paradigm of Intelligent Autonomous System CreationParadigm of Intelligent Autonomous System Creation
CONTEXT-POLICY-CONFIGURATION:CONTEXT-POLICY-CONFIGURATION:Paradigm of Intelligent Autonomous System CreationParadigm of Intelligent Autonomous System Creation
Oleksiy KhriyenkoOleksiy Khriyenko
June 8 – 12, 2010, Funchal, Madeira - PortugalJune 8 – 12, 2010, Funchal, Madeira - Portugal
University of Jyväskylä, FinlandUniversity of Jyväskylä, Finland
Vagan TerziyanVagan Terziyan
IOG, Agora Center, MIT IOG, Agora Center, MIT DepartmentDepartment
(presenter)(presenter)
1212thth International Conference on Enterprise Information Systems International Conference on Enterprise Information Systems
~ICEIS 2010~ ~ICEIS 2010~
1212thth International Conference on Enterprise Information Systems International Conference on Enterprise Information Systems
~ICEIS 2010~ ~ICEIS 2010~
Sergiy NikitinSergiy Nikitin
ContentContentContentContent
System evolution trendsSystem evolution trends UBIWARE platformUBIWARE platform Policy based system…Policy based system… Context-Policy-Configuration…Context-Policy-Configuration… Role-based Policy ControlRole-based Policy Control
Conclusions and future opportunitiesConclusions and future opportunities
CPC extensionCPC extension Example…Example…
AcknowledgementsAcknowledgements
System evolution trendsSystem evolution trendsSystem evolution trendsSystem evolution trends
Semantic Web:Semantic Web: semantic technologies are viewed today as a key technology to resolve the semantic technologies are viewed today as a key technology to resolve the problems of interoperability and integration within the heterogeneous world of ubiquitously problems of interoperability and integration within the heterogeneous world of ubiquitously interconnected objects and systemsinterconnected objects and systems;;
To achieve the vision of ubiquitous knowledge, the next generation of integration To achieve the vision of ubiquitous knowledge, the next generation of integration systems will utilize different methods and techniques:systems will utilize different methods and techniques:
Agent Technologies:Agent Technologies: a agent based approach to software engineering is considered to be gent based approach to software engineering is considered to be
facilitating the design of complex systems.facilitating the design of complex systems.
Context awareness:Context awareness: to be smart, system should be able to behave accordingly to a state of to be smart, system should be able to behave accordingly to a state of
environment and react on dynamic changes of it.environment and react on dynamic changes of it.
Policy:Policy: highly valuable approach towards creation automatically controllable system.highly valuable approach towards creation automatically controllable system.
UBIWARE platformUBIWARE platformUBIWARE platformUBIWARE platform
The The UBIWARE PlatformUBIWARE Platform is a development framework for creating multi-agent is a development framework for creating multi-agent systems. systems.
OntologyOntologyOntologyOntology
Rules Rules
Beliefs Beliefs
ActionsActionsActionsActions
new Beliefsnew Beliefsnew Beliefsnew Beliefs
OntologyOntologyOntologyOntology
Rules Rules
Beliefs Beliefs
OntologyOntologyOntologyOntology
Rules Rules
Beliefs Beliefs
OntologyOntologyOntologyOntology
Rules Rules
Beliefs Beliefs
OntologyOntologyOntologyOntology
Rules Rules
Beliefs Beliefs
Proactive Goal-driven Dynamic ResourceProactive Goal-driven Dynamic Resource as a main entity of any system… as a main entity of any system…
GUN GUN (Global Understanding Environment) concept: (Global Understanding Environment) concept: environment where all the resources of the virtual and environment where all the resources of the virtual and
the real world are connected and interoperate with the real world are connected and interoperate with each other…each other…
S-APL S-APL (Semantic Agent Programming Language): (Semantic Agent Programming Language): solves description of beliefs, rules and understanding solves description of beliefs, rules and understanding of their semantics, the meaning of predicates used in of their semantics, the meaning of predicates used in
those rules by all the parties involved while using those rules by all the parties involved while using first-order logic as the basis for an APL…first-order logic as the basis for an APL…
Policy based system…Policy based system…Policy based system…Policy based system…
System with two different levels of programming/administration. System with two different levels of programming/administration.
““advanced user” programming/administration:advanced user” programming/administration: implies building of the rules to reach implies building of the rules to reach different goals that cover particular domain. It is a definition of a certain domain by Ontology of different goals that cover particular domain. It is a definition of a certain domain by Ontology of Goals (set of possible abstract goals that can be reached by Resource, including sub-goal Goals (set of possible abstract goals that can be reached by Resource, including sub-goal hierarchy), and by set of abstract Behaviour Rules that can be used to achieve these goals (sub-hierarchy), and by set of abstract Behaviour Rules that can be used to achieve these goals (sub-goals);goals);
high-level system programming/administration:high-level system programming/administration: stage where user has to put the stage where user has to put the constraints on abstractions, he/she should specify/create concrete instance of goal/goals and constraints on abstractions, he/she should specify/create concrete instance of goal/goals and provide necessary initial states of the system;provide necessary initial states of the system;
Types of policies: Types of policies:
““D&R-type” policy:D&R-type” policy: user defines domains and ranges that can be considered as a policy user defines domains and ranges that can be considered as a policy for the system that should be followed during goal achievement process;for the system that should be followed during goal achievement process;
“ “ W-type ” policy:W-type ” policy: vector of weights of properties’ significance or vector of vector of weights of properties’ significance or vector of weights/preferences of any abstraction in general caseweights/preferences of any abstraction in general case;;
Context-Policy-Configuration…Context-Policy-Configuration…Context-Policy-Configuration…Context-Policy-Configuration…
Context-dependent Policy-based ControlContext-dependent Policy-based Control is an approach, able to leave is an approach, able to leave Resource flexible, dynamic and controlled at the same time. Resource flexible, dynamic and controlled at the same time.
CONTEXTCONTEXT POLICYPOLICY CONFIGURATIONCONFIGURATION(CONSTRAINTS)(CONSTRAINTS)
State of the EnvironmentState of the Environment(Resources with their properties)(Resources with their properties)
Context-based Policy Creation SystemContext-based Policy Creation System Resource Configuration EngineResource Configuration EngineModule builds a model of Policy based on Module builds a model of Policy based on available contextual information…available contextual information…
Module applies Policy and configures Resource Module applies Policy and configures Resource (its believes, capabilities, ontology, etc.)(its believes, capabilities, ontology, etc.)
Policy modelsPolicy models Resource believes, ontology, etc.Resource believes, ontology, etc.(Restricted subset of knowledge (Restricted subset of knowledge
and capabilities)and capabilities)(Filtering model, mask for perception)(Filtering model, mask for perception)
context
context
contextcontext
context
contextcontext
Configured knowledge and Configured knowledge and abilities to behave…abilities to behave…
context
OntologyOntologyOntologyOntology
OntologyOntologyOntologyOntology
(accordingly to policy model based on correspondent context)
0x
0x
0x
0x
)(xf
)(xf )(xf
)(xf
...),,,,(~...)))),(,(,(,()( 00000000 xxxxfxfxfxfxfxf
),...,,,( 113
12
11
1nwwwwW
),...,,,( 223
22
21
2nwwwwW
),...,,,( 333
32
31
3nwwwwW
Role-based Policy ControlRole-based Policy ControlRole-based Policy ControlRole-based Policy ControlGenerally we deal with a system with big amount of entities (Resources) with own Generally we deal with a system with big amount of entities (Resources) with own behaviours and goals. To be able to control the system on general level, we have to put behaviours and goals. To be able to control the system on general level, we have to put constraints/policies on separate entities as well as on the system in whole.constraints/policies on separate entities as well as on the system in whole.
Role as a Context. Role as a Context. Any organization, union, company, society, group, individual and etc. Any organization, union, company, society, group, individual and etc. can be considered as a sub system that plays certain Role, which restricts it with can be considered as a sub system that plays certain Role, which restricts it with particular set of goals and knowledge/resources used for goal achievement. particular set of goals and knowledge/resources used for goal achievement.
OntologyOntologyOntologyOntology
Role mRole mRole kRole k
Role nRole n
Role nRole n11
Role nRole n22
Role mRole m11
Role mRole m22
Role kRole k11
CPC extensionCPC extensionCPC extensionCPC extension
To be compatible with widely used technology we extend To be compatible with widely used technology we extend RDF SchemaRDF Schema with with some classes and properties for policy description. some classes and properties for policy description.
Policy System Ontology
RestrictionContainer
D&R_RestrictionContainer W_RestrictionContainer
basisSystem_is
hasRestrictionhasConfiguredOntology
rdfs:domainrdfs:range
rdfs:domain rdfs:rangerdfs:domain
rdfs:range
rdfs:subClassOfrdfs:subClassOf
ContainerProperty Class Resource
rdfs:subClassOf
rdfs:type
rdfs:subClassOfrdfs:type
rdfs:Statement
rdfs:Statement…
rdfs:Statement
rdfs:Statement…
RDFSRDFS
CPC CPC extensionextension
applied tordfs:range
rdfs:domain
Figure shows us initial part of CPC-extension of RDFS.Figure shows us initial part of CPC-extension of RDFS. In the platform we utilize N3 representation in S-APL language. In the platform we utilize N3 representation in S-APL language.
Example…Example…Example…Example…
Statement #1subject objectpredicate
Policy #n
applied to
GreenFactory #m
inContext
ContextContainer #1
Statement #2subject objectpredicate
GreenFactory #m
hasRole
Role #k
Statement #3subject objectpredicate
Factory #k
hasConfiguredOntology
Ontology #i
rdfs:Statement
rdfs:range
property: useEnergy
Oil-energy
Wood-energy
Coal-energy
Nuclear -energy
Hydro-energy
Wind-energy
Sun-energy
basisSystem_is Factory #k
hasRestriction D&R_RestrictionContainer #1
hasRestriction W_RestrictionContainer #1
Policy #n
rdfs:Statement
rdfs:range
Nuclear -energy
Hydro-energy
Wind-energy
Sun-energy
property: useEnergy
rdfs:Statement
buyEnergyFrom
Supplier #jSystem
rdfs:Statement
hasRelevance0.7
Consider a System – Consider a System – “GreenFactory”“GreenFactory” as a subsystem of as a subsystem of “Factory” “Factory” System with only System with only difference that “GreenFactory” utilized only green kind of energy: Nuclear-, Hydro-, Wind-, difference that “GreenFactory” utilized only green kind of energy: Nuclear-, Hydro-, Wind-, Sun-energy, etc.Sun-energy, etc.
““GreenFactory”GreenFactory” join some industrial financial group and should follow a policy that join some industrial financial group and should follow a policy that demands at least 70% of energy to be bought from the energy supplier that belongs to the demands at least 70% of energy to be bought from the energy supplier that belongs to the same financial group even if it is more expensive then buy energy from other suppliers. same financial group even if it is more expensive then buy energy from other suppliers.
Conclusions and future opportunitiesConclusions and future opportunitiesConclusions and future opportunitiesConclusions and future opportunities
This research presents a policy-based approach for supporting the high-level This research presents a policy-based approach for supporting the high-level configuration of systems, integrated into the middleware platform. Policies are configuration of systems, integrated into the middleware platform. Policies are high-level, declarative statements governing choices in the behaviour of a high-level, declarative statements governing choices in the behaviour of a system. system.
With this approach we do not program system in a hardcoded way, but build With this approach we do not program system in a hardcoded way, but build it able to change internal functionality and behaviour on the fly when context is it able to change internal functionality and behaviour on the fly when context is changed.changed.
As a future steps we are planning to elaborate a machine learning module to As a future steps we are planning to elaborate a machine learning module to automate (provide a suggestion to the user) the process of policy creation automate (provide a suggestion to the user) the process of policy creation depending on correspondent context.depending on correspondent context.
AcknowledgementsAcknowledgements AcknowledgementsAcknowledgements
University of JyväskyläUniversity of Jyväskylä
www.cs.jyu.fi/ai/OntoGroupwww.cs.jyu.fi/ai/OntoGroup
www.cs.jyu.fi/ai/OntoGroup/UBIWARE_details.htmwww.cs.jyu.fi/ai/OntoGroup/UBIWARE_details.htm