Dependability analysis in the Ambient Assisted Living Domain: An exploratory case study

Ambient Assisted Living (AAL) investigates the development of systems involving the use of different types of sensors, which monitor activities and vital signs of lonely elderly people in order to detect emergency situations or deviations from desirable medical patterns. Instead of requiring the elderly person to manually push a button to request assistance, state-of-the-art AAL solutions automate the process by 'perceiving' lonely elderly people in their home environment through various sensors and performing appropriate actions under the control of the underlying software. Dependability in the AAL domain is a critical requirement, since poor system availability, reliability, safety, or integrity may cause inappropriate emergency assistance to potentially have fatal consequences. Nevertheless, contemporary research has not focused on assessing dependability in this domain. This work attempts to fill this gap presenting an approach which relies on modern quantitative and qualitative dependability analysis techniques based on software architecture. The analysis method presented in this paper consists of conversion patterns from Unified Modeling Language (UML) behavior models of the AAL software architecture into a formal executable specification, based on a probabilistic process algebra description language, which enables a sound quantitative and qualitative analysis. The UML models specify system component interactions and are annotated with component failure probabilities and system usage profile information. The resulting formal specification is executed on PRISM, a model checking tool adequate for the purpose of our analysis in order to identify a set of domain-specific dependability properties expressed declaratively in Probabilistic Computational Tree Logic (PCTL). The benefits of using these techniques are twofold. Firstly, they allow us to seamlessly integrate the analysis during subsequent software lifecycle stages in critical scenarios. Secondly, we identify the components which have the highest impact on software system dependability, and therefore, be able to address software architecture and individual software component problems prior to implementation and the occurrence of critical errors.