2008SS-08

7/30/2019 2008SS-08

1/15

Intelligent Agents

Meet the Semantic Webin Smart Spaces

Harry Chen,Tim Finin, Anupam Joshi, and Lalana Kagal

University of Maryland, Baltimore County

Filip Perich

Cougaar Software

Dipanjan Chakraborty

IBM India Research Laboratory

IEEE INTERNET COMPUTING, NOVEMBER, OCTOBER 2004, Published by the IEEE Computer Society

2008. 04.18

Summarized by Dongjoo Lee, IDS Lab., Seoul National University

Presented by Dongjoo Lee, IDS Lab., Seoul National University

7/30/2019 2008SS-08

2/15

Copyright 2008 by CEBT

Contents

EasyMeeting

Vigil

Services

Architecture

Context Broker Architecture (Cobra)

COBRA-ONT

Context Reasoning

Privacy Protection

Conclusion

2

7/30/2019 2008SS-08

3/15


EasyMeeting

A pervasive computing system that supports users in a smart meeting-room environment in

which a distributed system of intelligent agents, services, devices, and sensors share acommon goal;

Goal

Provide relevant services and information to meeting participants on the basis of their contexts.

Context Broker

Provide a centralized model of context that all devices, services, and agents in the space can share

Acquire contextual information from sources that are unreachable by the resource-limited devices

Reason about contextual information that cant be directly acquired from the sensors

Detect and resolve inconsistent knowledge sotred in the shared context model

Protect privacy by enforcing policies that users have defined to control the sharing and use of theircontextual information

Differences

Uses OWL for expressing ontologies to

support context modeling and knowledge sharing

detect and resolve inconsistent context knowledge

protect the users privacy.

3

7/30/2019 2008SS-08

4/15


EasyMeeting - Vigil

Specialized server entities that facilitate system communication, client-

role management, and service-access control.

Clients, services, and Vigil managers

Role-based inference mechanism to control access to services

Role-permission definition

Reasoning of the role-assignment manager is built on the Rei

framework.

Deonticconcept

Rights, prohibitions, obligations, and dispensations

4

7/30/2019 2008SS-08

5/15


EasyMeeting - Services

Speech understanding

CCML (Centaurus Capability Markup Language)

IBM WebSphere Voice Server SDK, Voice XML

Presentation

AppleScript commands

Lighting control

X10 technology

Music

MP3 music player software

Greeting

Profile display

Web-based server application

5

7/30/2019 2008SS-08

6/15


EasyMeeting - Architecture

6

Presentation Schedule

On 8 September 2004,1:00 to 2:30 P.M.Room 338

device profileIn Standard Device Ontology

Harry (speaker)

Hrabowski(the distinguished audience)

7/30/2019 2008SS-08

7/15


Context Broker Architecture (Cobra)

Jena reasoning API OWL ontologies

Jess rule-based engine domain specific reasoning

7

7/30/2019 2008SS-08

8/15


COBRA-ONT

Why OWL ?

Expressive knowledge-representation

language

Have a normative syntax in RDF and

XML

Has many predefined classes and

properties

COBRA-ONT imports from SOUPA

Time, space, policy, social networks,

actions, location context, documents, and

events

8

Integrated from other ontologies

FOAF DAML-Time & the Entry Sub-ontology of

Time

OpenCyc Spatial Ontologies & RCC

COBRA-ONT & MoGATU BDI Ontology

Rei Policy Ontology

7/30/2019 2008SS-08

9/15


User Profile Example

9

7/30/2019 2008SS-08

10/15


Context Reasoning

Jena rule engine ontology axioms

Java Expert System Shell (JESS) forward-chaining inference

Algorithm

Ontology inference

1) Jess rule execution

2) select the type of context it attempt to infer

3) decide whether it can infer this type of context using only ontology reasoning

Logic inference

4) Find all essential supporting facts by querying the ontology model

5) Convert RDF representation into the Jess representation

6) Executing the predefined forward-chaining procedure

7) Add newly deduced facts to ontology model

10

7/30/2019 2008SS-08

11/15


Context Reasoning - Assumption-based reasoning

11

Harry is in Room RM338

Harry intends to give a presentation in meetting m1203

7/30/2019 2008SS-08

12/15


Privacy Protection

Users can define customized policy rules to permit or forbid access to their

private information in various granularity.

12

7/30/2019 2008SS-08

13/15


Privacy Protection - Example

13

RDF Notation 3 Syntax

7/30/2019 2008SS-08

14/15


Feedback from Demonstrations

From three external groups

UMBC university administrators, visitors from commercial companies and

other universities

Critics

The system has a limited ability to handle unexpected situational changes

The workflow process was too rigid and could be unsuitable for everydayusage

Using policy to control how private information is shared doesnt address

other kinds of privacy concerns such as the logging and persistent storage of

a users private information by the agents, and the possibility for the agents

acquiring certain private user information by reasoning over an aggregatedcollection of their public information.

14

7/30/2019 2008SS-08

15/15


Conclusion

The EasyMeeting and Cobra prototypes demonstrate the feasibility of

using OWL ontologies to let distributed agents

share knowledge

reason about contextual information

express policies for user privacy protection

Challenging issues

Scalability of knowledge sharing in a distributed and dynamic environment

Performance and time complexity of context reasoning of a vast amount of

sensing data

User-interface issues associated with editing and maintaining user privacy

policies

15

Documents

2008SS-08