19
A Context Modeling Survey T. Strang, C. Linnhoff-Popien German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU) Workshop on Advanced Context Modeling, Reasoning and Management, UbiComp, 2004 2008-09-29 Presentation by KwangHyun Nam, IDS Lab.

A Context Modeling Survey T. Strang, C. Linnhoff-Popien German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU) Workshop on Advanced

Embed Size (px)

Citation preview

Page 1: A Context Modeling Survey T. Strang, C. Linnhoff-Popien German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU) Workshop on Advanced

A Context Modeling Survey

T. Strang, C. Linnhoff-Popien

German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU)

Workshop on Advanced Context Modeling, Reasoning and Management, UbiComp, 2004

2008-09-29

Presentation by KwangHyun Nam, IDS Lab.

Page 2: A Context Modeling Survey T. Strang, C. Linnhoff-Popien German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU) Workshop on Advanced

Copyright 2008 by CEBT

Contents

Introduction

Fundamentals

Modeling Approaches

Key-Value Models

Markup Scheme Models

Graphical Models

Object Oriented Models

Logic Based Models

Ontology Based Models

Evaluation

Summary, conclusion and outlook

2

Page 3: A Context Modeling Survey T. Strang, C. Linnhoff-Popien German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU) Workshop on Advanced

Copyright 2008 by CEBT

Introduction

Past research

Published with respect to location, identity, time

Current research

To develop uniform context model, representation and query languages as well as reasoning algorithms

– To facilitate context sharing and interoperability of applications

Aim of this paper

Survey of the most relevant current approaches to model-ing context for ubiquitous computing

3

Page 4: A Context Modeling Survey T. Strang, C. Linnhoff-Popien German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU) Workshop on Advanced

Copyright 2008 by CEBT

Fundamentals

Evolution Chain

Context dependency is a major issue in recent work in the area of ubiquitous computing systems

Ubiquitous computing is a specialization of distributed computing and mobile computing

4

Page 5: A Context Modeling Survey T. Strang, C. Linnhoff-Popien German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU) Workshop on Advanced

Copyright 2008 by CEBT

Requirement for ubiquitous com-puting

Distributed composition (dc)

UbiComp is a derivative of a distributed computing system

Lacks of a central instance being responsible for the cre-ation, deployment and maintenance of data and services, in particular context descriptions

Composition and administration of model varies with high dynamics in terms of time, network, topology and source

Partial validation (pv)

Desirable to enable to partially validate contextual knowl-edge on structure & instance level

This is particularly important

– Due to the complexity of contextual interrelationships, which make any modeling intention error-prone

5

Page 6: A Context Modeling Survey T. Strang, C. Linnhoff-Popien German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU) Workshop on Advanced

Copyright 2008 by CEBT

Requirement for ubiquitous com-puting

Richness and quality of information (qua)

The quality of a information and the richness of that may differ

Model should support quality and richness indication

Incompleteness and ambiguity (inc)

The set of contextual information at any point in time is usually incomplete and/or ambiguous

This should be covered by the model

– Example

By interpolation of incomplete data on the instance level

6

Page 7: A Context Modeling Survey T. Strang, C. Linnhoff-Popien German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU) Workshop on Advanced

Copyright 2008 by CEBT

Requirement for ubiquitous com-puting

Level of formality (for)

A challenge to describe contextual facts & interrelationships in a precise and traceable manner

“Print document on printer near to me”

– What ‘near’ means to ‘me’? -> need a precise definition of terms

Each participating party in an ubiquitous computing interac-tion shares the same interpretation of the data exchanged

– Shared understanding

Applicability to existing environments (app)

A context model must be applicable within existing the infra-structure of ubiquitous computing environment

Example

– A service framework

7

Page 8: A Context Modeling Survey T. Strang, C. Linnhoff-Popien German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU) Workshop on Advanced

Copyright 2008 by CEBT

Modeling approaches

Key-Value Models

Most simple data structure of models

Frequently used in distributed service frameworks

Described with a list of simple attributes in a key-value manner

Easy to manage

Not very efficient for more sophisticated structuring

8

Environment Variables: Key-Value Models

Page 9: A Context Modeling Survey T. Strang, C. Linnhoff-Popien German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU) Workshop on Advanced

Copyright 2008 by CEBT

Modeling approaches (cont’d)

Markup Scheme Models

Hierarchical data structure consisting of markup tags

Typical representatives: profiles

– Based upon a serialization of a derivative of SGML

Examples

– Defined as extension to

Composite Capabilities/Preference Profile (CC/PP)

User Agent Profile (UAProf)

– Comprehensive Structured Context Profiles (CSCP)

Unlike CC/PP, not define any fixed hierarchy

– Pervasive Profile Description Language (PPDL)

Allow to account for contextual information and dependencies when defining interaction pqtterns on a limited scale

– Centaurus Capability Markup Language (CCML)

9

Page 10: A Context Modeling Survey T. Strang, C. Linnhoff-Popien German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU) Workshop on Advanced

Copyright 2008 by CEBT

Modeling approaches (cont’d)

Graphical Models

Particularly useful for structuring, but usually not used on instance level

Examples

– Well known: UML

A strong graphical component (UML diagram)

Due to its generic structure, UML is appropriate to model the con-text

– Contextual Extended ORM

Basic modeling concept in ORM is the fact

The modeling of a domain involves indentifying proper fact types & roles

Extended ORM is allowed to categorize fact types either as static or as dynamic

10

Page 11: A Context Modeling Survey T. Strang, C. Linnhoff-Popien German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU) Workshop on Advanced

Copyright 2008 by CEBT

Modeling approaches (cont’d)

Object Oriented Models

Intention behind object orientation is (as always) encapsu-lation and reusability

Examples

– Representative: Cues (TEA project)

Provide an abstraction from physical and logical sensors

Regarded as a function Taking the value of a single physical/logical sensor up to a certain time as

input

Providing a symbolic/sub-symbolic output

– Active Object Model (GUIDE project)

All the details of data collection and fusing are encapsulated within the active objects Hidden to other components of the system.

11

Page 12: A Context Modeling Survey T. Strang, C. Linnhoff-Popien German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU) Workshop on Advanced

Copyright 2008 by CEBT

Modeling approaches (cont’d)

Logic Based Models

Logic defines conditions on which a concluding expression or fact may be derived from a set of other expressions or facts (reasoning)

Context is defined as facts, expressions and rules

High degree of formality

Examples

– McCarthy’s Formalizing Context

To give a formalization recipe which allows for simple axioms for common sense phenomena

– Akman & Surav’s Extended Situation Theory

Extend the Situation Theory by Barwise & Perry

To model the context with situation which are ordinary situations

12

Page 13: A Context Modeling Survey T. Strang, C. Linnhoff-Popien German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU) Workshop on Advanced

Copyright 2008 by CEBT

Modeling approaches (cont’d)

Ontology Based Models

Ontology is used as explicit specification of a shared concep-tualization

Strong in the field of normalization and formality

Context is modeled as concepts and facts

Examples

– ASC model of Context Ontology Language (CoOL)

Used to support context-awareness in distributed service frameworks for various applications

– CONON ontology

An upper ontology which captures general features of basic contextual entities and a collection of domain specific ontologies and features.

– CoBrA system

Provide a set of ontological concepts to characterize entities

13

Page 14: A Context Modeling Survey T. Strang, C. Linnhoff-Popien German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU) Workshop on Advanced

Copyright 2008 by CEBT

Evaluation

Key-Value Models

Weak on the requirements 1 to 5 (-)

The simplicity of key-value pair is a drawback if quality meta-information or ambiguity shall be considered (-)

Solely the applicability is a strength (+)

Markup Scheme Models

Strong concerning the partial validation requirement (++)

Standard CC/PP & UAProf have only restricted overriding and merging mechanisms (-)

Applicability to existing markup-centric infrastructures is a strength (++)

14

Page 15: A Context Modeling Survey T. Strang, C. Linnhoff-Popien German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU) Workshop on Advanced

Copyright 2008 by CEBT

Evaluation (cont’d)

Graphical Models

The strengths are definitely on the structure level

– Mainly used to describe the structure of contextual knowledge and drive code or an ER-model from model

Distributed composition requirement has some constraints on the structure level (-)

Object Oriented Models

Strong regarding the distributed composition requirement (++)

A higher level of formality is reached through the use of well-defined interfaces (+)

– The invisibility as consequence of encapsulation is a little draw-back

15

Page 16: A Context Modeling Survey T. Strang, C. Linnhoff-Popien German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU) Workshop on Advanced

Copyright 2008 by CEBT

Evaluation (cont’d)

Logic Based Models

Be composed distributed (++)

Formality is extremely high (++)

However, this model is weak with respect to other require-ments(-)

Ontology Based Models

Strong in the distributed composition requirement (++)

Inherit the strengths in the field of normalization and for-mality from ontologies (++)

All requirements for UbiComp enable to be covered by this model.

16

Page 17: A Context Modeling Survey T. Strang, C. Linnhoff-Popien German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU) Workshop on Advanced

Copyright 2008 by CEBT

Summary, Conclusion and Out-look

17

Miss Ubiquitous ContestMiss Ubiquitous Contest

I’m so nervous Me, too.

Winner : Ontology

Thank you. I give all these glory to you!!

But, all others are also valuable.

Page 18: A Context Modeling Survey T. Strang, C. Linnhoff-Popien German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU) Workshop on Advanced

Copyright 2008 by CEBT

Summary, Conclusion and Out-look

The most promising assets for context modeling for ubiquitous computing environments

Ontology category

But, the other approaches aren’t unsuitable for UbiComp

This list of context modeling approaches is comprehen-sive, but - as in all surveys - incomplete

18

Page 19: A Context Modeling Survey T. Strang, C. Linnhoff-Popien German Aerospace Center (DLR), Ludwig-Maximilians-University Munich (LMU) Workshop on Advanced

Copyright 2008 by CEBT

Discussion

Pros

Indicate definite criterion for comparison of models for ubiquitous computing

May help to identify appropriate approach for ubiquitous computing applications

Cons

Lack reasons of analysis decision with respect to criterion of some items

Typographical error

– Specialisation -> Specialization ( 1 Page, right-side 6th line)

– Categorised -> Categorized (3 Page, in content of Graphical Models)

19