A Study of Context-Awareness- CASS, Hydrogen

2008. 7. 17Context Team

Summarized and Presented by Seungseok Kang

CASS (Context-awareness sub-struc-ture)• Middleware for Mobile Context-Aware Applications

– Server-based middleware– Provides support for context-aware application on hand-

held and other mobile devices– Supports high level context data abstraction– Overcomes the memory and processor constraints

• Small mobile computer platform

– Separation of context based inference and behaviors from application code

Background

• High-level context information– Higher-level context abstraction makes it easier for appli-

cation designer– The higher level of the context, the easier it should be use

• Requirements for context-aware mobile system– Supporting a large number of context sources– Provision for context history– Supporting for context interpretation– Supporting higher-level abstraction of contexts– Event-based extensible middleware– Supporting transparent use of distributed sources– Supporting the separation of application

Architecture

• CASS Middleware– Server-based middleware

• Does not suffer from the processor and memoryconstraints

– Supports application in theuse of local caching ofinformation• CASS application do not need to store low-level details of context sources

• CASS Design– RuleEngine– SensorListener : update and store context information– ContextRetriever : retrieve stored context– ChangeListener : listen for notification of context change events

• Data Management– Use database for persistent data storage

• Context, application, user data and domain knowledge

– Data can be read and manipulated by SQL

Context Inference Mechanism

• Inference engine– Works in conjunction with separated knowledge base

• Represents knowledge in a more natural fashion• Knowledge can be updated without changing the engine

– Forward chaining• Contexts lead to fewer higher-level contexts

• Knowledge base– Database tables with high-level context states and behav-

ior

• Abstraction and search– Reducing the state-space of individual context compo-

nents• Ex) degrees of temperature -> Cold, Normal, Hot (context on-

tology)

Applications

• Developing some applications currently

• MALLET– “Maintenance assignment listing lightweight electronic

tool”– Building maintenance to be organized in a context-aware

way

• STONE– “Start on entrance”– Avoiding awkward start to presentations and lectures

• Ex) presenter can enable some preparations to happen au-tomatically as soon as the presenter enters the presentation area(beam projector, presentation application, presentation file)

Sensors

• Works with sensors– Sensor-derived context management– Sensor failure– Sensor fusion

• Sensor software is extensible– To allow new types of sensors to be added

Hydrogen

• Context-awareness framework on mobile devices– Supports context-awareness for considering some con-

straints• Network connections• Limited computing power• Characteristics of mobile users

– Extensible framework to consider all kind of context in-formation

• Related works– ContextToolkit– GeoNotes– CampusSpace– ……

Background

• Context characteristics– Physical context

• Representing the level of environment sensors

– Logical context• Representing more abstract information• Enrich the semantics of physical context information

• Requirements for an architecture of a framework to support context-awareness on mobile devices– Lightweightness – limited processing power– Extensibility – limited sensor extension– Robustness – network disconnection– Meta-information – controversial information between

sensors– Context-Sharing – incomplete sensed context

Hydrogen Context-Framework

• Framework Architecture– Three layer

• Application layer• Management layer

– Providing and retrievingcontexts and sharingcontext informationwith other devices usingP2P communication

• Adaptor Layer– Separating context storing, sensing from other layers– Responsible to get information from sensors– Providing same context information to multiple applications

– All application have access to all context data by querying the ContextServer

– All layers are located on one device• Robust against network disconnections

Implementation

• Prototypical implementation– PersonalJava, Jeode, J2ME, iPAQs, PocketPC 2002

– ContextClient• Inter-layer communication with XML protocol• Provides sending and listening data functionally

Implementation (cont.)

• Context– Includes the context of the location / remote devices– Time, Location, Device, User, Network– Other context can be specialized by ContextObjects

• ContextClient– Responsible for communication : open ports, queries data

• ContextServer– Java-executable object– Storing all information about the contexts– Two-way communication : XML streams / serialized java

objects

• Adaptors and Applications– Simplify the communication

between ContextClient andContextServer

Impelemtation (cont.)

• Extending the framework– Covering new types of context

• Can be specialized by ContextObject

– Two abstract methods• toXML() : convert the values of object into an XML stream• fromXML() : parsing the stream and filling into object

• Exemplary Applications– Context-aware postbox

• Delivering images and intelligent business card (text) to mo-bile devices and local desktop computer

Open Issues

• Comprehensive context model– Integration of comprehensive context models– W3C supports CC/PP (composite capabilities / preference

profiles)

• XML protocol– ContextObjects can be converted into an XML representation

• Context can be processed by every application

• Context Sharing– Offering the opportunity to gain information about the con-

text of other context-aware devices– Exchanging complementary context information between

ContextServer by network– Security and reliability problem

Conclusion

• CASS– Server-based middleware– Context-aware application with high-level context abstrac-

tion– Context inferences separated from application– Configuration by users

• Hydrogen– Identifying the requirements for a context-aware architec-

ture– Three-layer framework for context-awareness system– Feasible context architecture with XML– Extensible framework to consider all kind of context in-

formation

Evaluation of papers

• Pros– Moving context sensors to separate computers and allow-

ing the sensors to communicate with the middleware al-lows any types of devices, with or without context sensing capabilities, to run context-aware applications

– Because devices acquires context data from middleware, context data can be shared among multiple devices

• Cons– Only getting context related data may limit the possible

application scenarios for context-aware services• An application will be more interested in local temperature

around the device, and not the temperature at some point where sensing node is deployed

And We Think……

• Similarity between CASS and Hydrogen– Requirements for context-aware system on mobile devices– Separation of context management module from devices and

applications• Middleware (CASS), Layer architecture (Hydrogen)

– Context modeling• Definition of context dimensions• High-level context abstraction• Extensibility and feasibility of context information

– Context sharing• To overcome the incompleteness of sensed data

• So, we should consider about……– The limitation of mobile device environment– Extensible definition of context dimensions– Separated and reusable modules in context-aware system– Proper algorithms of mechanisms for context-sharing

Appendix

• Gaia– A distributed middleware infrastructure that coordinates

software entities and heterogeneous networked devices contained in a physical space

– Exports services to query and utilize existing resources, to access and use current context, and provides a framework to develop user-centric, resource-aware, multi-device, context-sensitive, and mobile applications

– GAIA architecture

Appendix (cont.)

• GAIA services– event service– context service – presence service – space repository – context file system

• Discussion– It is good to think ubiquitous computing environment like

an operating system and give basic functionalities for ap-plications.

– But it is hard to determine what functionalities are needed for context aware applications. The authors suggest only basic functionality so that consideration about more func-tionalities is needed.