Today’s complex operating environments require systems to capture and document context together with requirements. Such contextual requirements are used by adaptive systems. During the lifetime of an adaptive system contextual requirements might become outdated if the operational environment changes or might be affected by changes in the monitoring infrastructure (e.g., sensor failure). Such uncertainty might affect the satisfaction of contextual requirements. Overcoming this limitation in adaptive systems calls for runtime support on the adaptation of contextual requirements that are affected by runtime uncertainty.
Alessia Knauss, Daniela Damian, and Angela Rook from our lab, in collaboration with Xavier Franch, Hausi Müller, and Alex Thomo developed ACon, an approach that supports the adaptation of contextual requirements at runtime. ACon uses a feedback loop to detect contextual requirements affected by uncertainty and integrates data mining techniques to determine the current measurable context conditions in which the contextual requirement is valid.
We evaluated ACon based on our OAR Northwest project in an unpredictable environment, the Atlantic Ocean, with lots of uncertainty involved. We mined contextual data of 46 sensors for five contextual requirements. Our evaluation showed promising results and leaves room for further investigations in this direction.