Information Technology Reference
In-Depth Information
extend this work in [17,18,19] by proposing a Continuous Adaptive RE (CARE) frame-
work for continuous online refinement of requirements at runtime by the system itself
involving the end-user. We proposed an architecture of an application that instantiate
CARE. We proposed a classification of adaptation at runtime by exploiting incremental
reasoning over adaptive requirements represented as runtime artifact. Similar ideas has
been proposed treating goals as fuzzy goals formalized using fuzzy logic representing
strategies for adaptation and operationalizing them as BPEL processes in [2]. Another
variation of this idea has been advocated in [22] as “Awareness Requirements”, as a
way to express constraints on requirements as meta requirements to deal with uncer-
tainty while developing SAS.
In goal-oriented modeling,
Tropos
has been extended to capture the contextual vari-
ability (mainly location) [1] by leveraging the concept of variation points [10] exploiting
the decomposition rules in a goal tree. Mainly, it helps in linking the alternative in the
goal model to the corresponding context (location) that helps in monitoring facts and
reasoning for adaptation in case of change in the context (location). Extended design ab-
stractions, including environment models, explicit goal types, and conditions for build-
ing adaptive agents have been proposed as an extension of
Tropos
, in Tropos4AS [13].
6
Discussion
It is worth to further discuss assumptions underlying the suggested problem formula-
tion, its generality as well as its practical impact. The problem formulation suggested
in this paper makes no assumptions and imposes no constraints on how the information
that is used and acquired. We thereby recognize that not all information can be col-
lected during requirements engineering, or at design time, but that this will depend on
the technologies used to implement the system. For example, the information about the
context, the formulas in
C
may - if the implementation technology allows - be obtained
by recognizing patterns in the data that arrives through sensors, then matching patterns
of data to templates of proposition or implications. We stayed in the propositional case,
since this was enough to define the main concepts and relations, and subsequently use
them to formulate the runtime requirements adaptation problem. The actual system will
operate using perhaps more elaborate, first-order formalisms to represent information,
so as to make that information useful for planning algorithms applied to identify can-
didate solutions. However, regardless of the formalism used, the system still needs to
be designed to ensure the general conditions and relations that the problem formulation
states: e.g., that the system needs an internal representation of information pertaining to
contexts, domain assumptions, tasks, goals, and so on, that goals and quality constraints
are satisfied through consistent combinations of
C
,
K
and
T
, among others.
Concerning generality of the proposed problem formulation and practical implica-
tion, our aim in this work was first to understand the general problem, and then focus
on developing particular requirements modeling languages to handle it. In this regards
we believe that recently proposed frameworks for engineering requirements for SAS
can be reconnected to our problem formulation. Consider, for example RELAX [23],
which proposes a formalism for the specification of requirements and a particular way
to relax them at runtime: if a requirement cannot be satisfied to the desired extent, then
