Multi-level context adaptation in the Web of Things

Mehdi Terdjimi - SOC & TWEAK Team
LIRIS - Université Lyon 1
http://liris.cnrs.fr/~mterdjim/slides_presentation

At first, the Internet of Things


source: http://nwavetec.com/technology/the-internet-of-things/

Then, the Web of Things

  • Built upon the IoT
  • Relies on Web technologies and standards
  • Breaks through silos

Context in ubiquitous computing

  • high impact to the application's behavior
  • complex and highly dynamic


How to design generic context models...

  • to include what is needed by the application ?
  • to do the actual adaptation, specific to the application ?

The answer is...

Multi-level context models

Context of the work: ASAWoO

  • Exposing physical objects through a web interface: avatar (Mrissa et al., 2014)
  • Using context to deal with resource-constrained objects.

Challenges

  • Interoperable context data sources
  • Generic context models
  • Scalability in view of the high numbers of clients


Motivations

  • Semantic context models for the WoT (SotA)
  • Adaptation at several abstraction levels
  • Scalable adaptation engine (reasoning steps: modularization, distribution)

PLAN

  • State of the art
    • context modeling
    • mobile/client reasoning
  • Approach
  • Implementation
  • Evaluation
  • Discussion, future work

State of the art: context modeling

Former definitions

  • Location, identity of entities, changes to entities
    (Schilit and Theimer, 1994)
  • Physical and conceptual states of an entity
    (Pascoe, 1998)
  • Any information that can be used to characterize the situation of an entity
    (Dey et al., 1999)


What about specific applications ?

State of the art: context modeling

Physical World

Time, Location, Activity
(Schilit et al., 1993), (Dey et al., 1999), (Schmidt, 2003), (Zimmermann et al., 2007)
User, Device, Geospatial, Environment
Query recommendation and auto-complete (Arias, 2008)
Location, Time, Activity, Posture, Privacy
Context-aware web browser (Coppola et al., 2010)

State of the art: context modeling

Communication

Device computing resources, Network (status, type), Distances
(Gold and Mascolo, 2001), (Mascolo et al., 2002)
Context-aware adaptive routing (Musolesi and Mascolo, 2009)


User, Location, Network policy, Network capacity, Status (Wei et al., 2006)
Notion of static, dynamic network context (Raverdy et al., 2006)

State of the art: context modeling

Application architecture

Separation of the application core from the context information
(Gensel et al., 2008), (Chaari et al., 2005)
Device, Location, User, Social, Environment, Time, System, Application-specific Aspect-oriented approach (Munnely et al., 2007)
Location, Device, Application, Time (group calendar), Community (group, users, roles), Process Groupware systems (Kirsch-Pinheiro et al., 2004)

State of the art: context modeling

Social Computing

User sessions: click-trough & navigation, query reformulation & specialization (ranking)
(Cao et al., 2009), (Xiang et al., 2010)
Domain, User, Environment and Interaction provides context sharing between agents in MAS (Brézillon, 2003)
User, Item, Environment, Observer describes a situation in MAS (Bazire and Brézillon, 2005)

Approach

  • 1. Semantic context model
    based on the 4 abstractions levels and SotA dimensions
  • 2. Generate graphs from context models
    SPARQL querying on graphs
  • 3. Reasoning steps separation & modularization
    Distributed reasoning process
"Which communication protocols can be used?"
(Location: home, Security: Level_1, Time: Evening)

Perspectives

  • Improve/replace OwlReasoner
  • to study the impact of more complex queries
  • Semantic multi-level context model
  • to provide...
  • Adaptive reasoning process
  • based on context states (= graphs)
  • Rule engine
HyLAR: Hybrid Location-Agnostic Reasoning Mehdi Terdjimi, Lionel Médini, Michael Mrissa LIRIS - Université Lyon 1