A Software Engineering Framework for Context-Aware Service-Based Processes in Pervasive Environments - Advanced Design Approaches to Emerging Software Systems - page 112

A MODEL-DRIVEN FRAMEWORK
FOR MANAGING CONTEXT-
AWARE ADAPTIVE SERVICE-
BASED PROCESSES

plied. We use the Semantic Web Rule Language
(SWRL) on top of OWL for interpreting context
using domain specific rules and producing new
facts. However, the approach could be extended
to use other reasoner types.

Many different solutions have been proposed
by researchers to the problem of context-aware
adaptation during process development and provi-
sion. Indeed, process design and modeling must
be flexible enough to deal with constant changes.
The flexibility could be provided or addressed by
incorporating variabilities into a system (Koning
et al. 2009). Most of the approaches tackle process
adaptation on the process instance or definition
level by explicitly specifying some form of varia-
tion points. To date, a variety of different adapta-
tion approaches have been proposed for capturing
variabilities (e.g. Mietzner and Leymann 2008).
Common to all these approaches is that they cap-
ture the process variant as a monolithic structure
containing variation points to differentiate between
process family members. By making appropriate
choices to resolve the variation points, either at
design time or at runtime, a single process vari-
ant could be constructed. The problem is that, for
example, each task in the process is modeled as a
variation point in and of itself, each governed by
its own clause to determine inclusion or exclusion.
This is in contradiction with how the developer
or architect logically views the process variant
i.e. in terms of the features that determine the
difference between process variants in each usage
context. Moreover, managing and understanding
the process variants becomes more difficult when
the number of variabilities and their relationships
increase.
Motivated by these problems and directives
in mind, we propose an MDD-based framework
called Apto that introduces the evolution fragment
and evolution primitive constructs to capture the
variability in a more logical and independent form.

Discussion

The proposed approach can be seen from two
perspectives: (i) identifying context features and
giving them semantics by mapping context feature
models to OCM; and (ii) using feature models to
provide a representation of variability in context
models. This has several advantages.
Firstly, from the context modeler usability
perspective, the proposed approach is intuitive; it
allows her to think about the context information
from different perspectives and use the feature
model available tools. Indeed, it is possible to think
about the context information from different point
of views and design different feature models. For
example, the context modeler may choose to split
the context feature model into more than one sub
feature models each of which may be designed to
look at the context from a different view point.
Secondly, context feature model allows the
context modeler to devise context-specific fea-
tures that can be shared among all applications
that need to use this context. Moreover, retrieving
context information using general-purpose query
mechanisms remains possible by devising a special
context feature. Thirdly, unlike the reasoning on a
one monolithic context information, the proposed
approach gives the possibility to provide the con-
text information on arbitrary levels of abstraction
thanks to the arbitrary composition of context
primitives e.g. inference rules. Fourthly, the use of
context-specific features may improve the overall
performance of the system, since it might decrease
the number of network interactions between an
application and the context provider.