Towards a Formal Model of Context
Awareness
Mikkel Baun Kjærgaard
Jonathan Bunde-Pedersen
Department of Computer Science
University of Aarhus
Motivation
Pervasive computing vision:
– Move computation from the desktop computer to a number of devices
embedded in the environment of the user.
Context-awareness
– However the task of handling all these devices should not overwhelm
the human user.
– Therefore the devices and applications need to become aware of the
context they are situated in.
– The context of interest could here both be the physical, social and
computational environment.
Modeling Context Information
– Complex
• A lot of different factors where some of them can only be described
probabilistically.
– Interwoven
• The different types of information depend and relate to each other in many
ways.
May 2006
Mikkel Baun Kjærgaard
2
Example
Pervasive Healthcare
– Context information types
•
•
•
•
•
May 2006
Personal Status
Activities
Location
Staff
Patients
Mikkel Baun Kjærgaard
3
Motivation
Context-aware applications
Contextual information
May 2006
Mikkel Baun Kjærgaard
4
Motivation
Application areas
– Healthcare, ...
– A lot of the applications are critical in some way
– At build time / online validation?
• For interesting properties: consistency, privacy, …
Formal foundation
– Should be expressive enough to model complex and
interwoven context-information.
Goal
– How can the Ambient Calculus be extended to allow
us to model complex and interwoven sets of context
information?
May 2006
Mikkel Baun Kjærgaard
5
Ambient Calculus
The Ambient Calculus
– deals with ambients which are tree-structured
recursively defined entities.
– An ambient may be situated in another ambient and
itself contain ambients, and
– Information between ambients may only flow between
nearby ambients, i.e. parents or siblings.
Modeling Context using Ambients
– If context-information should be made available for an
ambient it must be positioned near the ambient in the
tree.
May 2006
Mikkel Baun Kjærgaard
6
Idea
Multiple Contexts
– Ambients which can be present in one or more trees.
Two types of ambients
– Context ambients vs. Reference ambients
New capabilities which enables the reference ambients
to navigate in multiple contexts.
May 2006
Mikkel Baun Kjærgaard
7
Example
May 2006
Mikkel Baun Kjærgaard
8
Example
May 2006
Mikkel Baun Kjærgaard
9
Example
1: out{*}
2: out{Person&Status&Entity}
3: in{Person} Doctor
4: in{Status} NotBusy
5: in{Person&Status&Location&~Entity} ?
May 2006
Mikkel Baun Kjærgaard
10
Extensions ambients
– Context ambients
• Context information is represented in trees.
• Names of context trees are unmutable.
• Context ambients have limited capabilities.
– Reference ambients
• Only one reference in a single context
• Reference ambients must not contain references to other
reference ambients.
• Reference ambients are unique in the sense that all
references to ambient n points to the same n.
• Reference ambients cannot remove themselves directly from
a context.
May 2006
Mikkel Baun Kjærgaard
11
Extensions capabilities
Navigation
– Multiple contexts
– Simple Boolean expressions
– Wildcard name
Observability
– Coenter / Coexit
May 2006
Mikkel Baun Kjærgaard
12
Defining semantics
Ambient like
– Example of a simple in-rule
May 2006
Mikkel Baun Kjærgaard
13
Discussion
Limitations
– Restricts context information to a number of trees
which limits the expressiveness but to a lesser degree
than prior research.
– Does not take probabilistic context information into
account. It is important for a calculi to be able to
handle such information because it is the typical
output of most sensors of context information.
– Is the above needed in practice?
May 2006
Mikkel Baun Kjærgaard
14
Discussion
The present CONAWA calculus does not bridge the gap
to a combined theory and systems building approach
– base for simulation, verification and software prototyping
– (i) Building a simulator which can interpret CONAWA
descriptions
– (ii) Defining a logic which make it possible to state properties
which could be verified
• “This awarephone must never reveal my location to other people
than the ones I have explicitly granted this priviledge”?
– (iii) Building a code generation facility that makes it possible to
build skeleton code from the models.
May 2006
Mikkel Baun Kjærgaard
15
Conclusion
– Expressiveness of formal models of context
awareness.
– Here the models expressiveness should be sufficient
to model complex and interwoven sets of contextinformation.
– The CONAWA calculus was given as an example of a
calculus that is a step in the direction of making such
modeling possible.
May 2006
Mikkel Baun Kjærgaard
16
Descargar

Architecture Description Languages (ADL)