2 Artificial Intelligence Applications Institute, University of Edinburgh, UK Institute for Human...

2

Artificial Intelligence Applications Institute, University of Edinburgh, UKInstitute for Human and Machine Cognition, Pensacola, Florida

CoSAR-TSCoSAR-TS

Coalition Search and RescueTask Support

Demo – 16-Oct-2003

DARPA DAML ProgramAIAI, Edinburgh & IHMC, Florida

3

Artificial Intelligence Applications Institute, University of Edinburgh, UKInstitute for Human and Machine Cognition, Pensacola, Florida

DescriptionDescription

To study and develop a demonstrator for Task Support in a realistic Coalition Search and Rescue scenario

To show the value of linking research at AIAI work on I-X Task Support with IHMC work on KAoS Agent Policy and Domain Services

SONAT Elements of National Power (ENP) database extended with Binni data

OWL descriptions of country infrastructure DAML-S descriptions of agents and services are being

used Feedback to the DAML-S development community are

being provided Input to standards efforts (e.g. SWSL)

Binni SAR ScenarioBinni SAR Scenario

Binni SAR ScenarioBinni SAR Scenario

6

Artificial Intelligence Applications Institute, University of Edinburgh, UKInstitute for Human and Machine Cognition, Pensacola, Florida

SAR Joint PublicationsSAR Joint Publications

CoSAR-TS Demo ArchitectureCoSAR-TS Demo Architecture

CoABS Grid Manager

KAoS Policy and Domain Services

Task Support Tools

Process Panel

Domain Editor

Activity Editor

Messenger I-Space

Map Tool

11

Artificial Intelligence Applications Institute, University of Edinburgh, UKInstitute for Human and Machine Cognition, Pensacola, Florida

DAML CoSAR-TS OntologyDAML CoSAR-TS Ontology

12

Artificial Intelligence Applications Institute, University of Edinburgh, UKInstitute for Human and Machine Cognition, Pensacola, Florida

DAML-SDAML-S

KAoS DAML Policy FragmentKAoS DAML Policy Fragment <rdfs:Class rdf:ID='policy-6ef00f45-00f4-0000-8000-0000deadbeefAction' daml:label='coginst_ActionClassName'> <daml:subClassOf rdf:resource='http://ontology.coginst.uwf.edu/CoSAR-TS/RescueResources.daml#ReturnResultQueryRescueResourcesAction'/> <daml:subClassOf> <daml:Restriction rdf:ID='policy-6ef00f45-00f4-0000-8000-0000deadbeefActorRestriction0'> <daml:onProperty rdf:resource='http://ontology.coginst.uwf.edu/Action.daml#performedBy'/> <daml:toClass rdf:resource='#MembersOfDomainBinni'/> </daml:Restriction> </daml:subClassOf> <daml:subClassOf> <daml:Restriction rdf:ID='policy-6ef00f45-00f4-0000-8000-0000deadbeefTargetRestriction0'> <daml:onProperty> <daml:ObjectProperty rdf:about='http://ontology.coginst.uwf.edu/CoSAR-TS/RescueResources.daml#hasHomeCountry'/> </daml:onProperty> <daml:toClass rdf:resource='#policy-6ef00f45-00f4-0000-8000-0000deadbeefTargetInstances0'/> </daml:Restriction> </daml:subClassOf> <daml:subClassOf> <daml:Restriction rdf:ID='policy-6ef00f45-00f4-0000-8000-0000deadbeefTargetRestriction1'> <daml:onProperty> <daml:ObjectProperty rdf:about='http://ontology.coginst.uwf.edu/CoSAR-TS/RescueResources.daml#hasRescueDestination'/> </daml:onProperty> <daml:toClass> <rdfs:Class rdf:about='#policy-6ef00f45-00f4-0000-8000-0000deadbeefTargetInstances1'/> </daml:toClass> </daml:Restriction> </daml:subClassOf> </rdfs:Class>

<RDFNsId1:NegAuthorizationPolicy rdf:ID='policy-6ef00f45-00f4-0000-8000-0000deadbeef' RDFNsId1:hasName='CoSAR-TS1' RDFNsId1:hasPriority='1' RDFNsId1:hasUpdateTimeStamp='1049833307991'> <RDFNsId1:controls rdf:resource='#policy-6ef00f45-00f4-0000-8000-0000deadbeefAction'/> </RDFNsId1:NegAuthorizationPolicy></rdf:RDF>

14

Artificial Intelligence Applications Institute, University of Edinburgh, UKInstitute for Human and Machine Cognition, Pensacola, Florida

KAoS DAML Policy ExampleKAoS DAML Policy Example

Example defines negative authorization for CMU Matchmaker to return Gao Rescue Resources if the destination hospital is in Arabello

Why not just manage policies within the Matchmaker DAML-S descriptions themselves?

– Perspective independence: For various reasons, either the MM service, the Coalition, or Gao may be motivated to define and manage such a policy

– Confidentiality: It may be in the defining party’s interest not to disclose the policy to the other parties

– Resource management independence: The policy will need to be enforced by some component that belongs to the defining party, not necessarily to the subject or target of the policy

– Redundancy: The independence of the policy reasoning mechanisms provides a check against buggy, malicious, or poorly-designed code in the agents and services

15

Artificial Intelligence Applications Institute, University of Edinburgh, UKInstitute for Human and Machine Cognition, Pensacola, Florida

DemonstrationDemonstration

16

Artificial Intelligence Applications Institute, University of Edinburgh, UKInstitute for Human and Machine Cognition, Pensacola, Florida

Achievements to DateAchievements to Date

Direct OWL file processing from I-X of BBN SONAT ENP database via HP Jena Toolkit

SOAP Access to SAR Resources from KAoS and I-X via Katia Sycara’s CMU Matchmaker

I-X linked to Norman Sadeh’s CMU Notification Agent for personalized context-aware notifications

Declarative representation and reasoning about KAoS policy and domain information in DAML (soon OWL)

Complexities of coalition organization represented in KAoS domains

KAoS policy-governed access to Semantic Web information and services

Use of KAoS to dynamically generate I-X agent relationships (peers, subordinates, superiors

17

Artificial Intelligence Applications Institute, University of Edinburgh, UKInstitute for Human and Machine Cognition, Pensacola, Florida

http://www.aiai.ed.ac.uk/project/cosar-ts/

http://www.coginst.uwf.edu/kaos/

http://i-x.info

http://binni.org

Further InformationFurther Information

18

Artificial Intelligence Applications Institute, University of Edinburgh, UKInstitute for Human and Machine Cognition, Pensacola, Florida

Extra Slides

Additional DetailsAdditional Details

Binni SAR ScenarioBinni SAR Scenario