Architecture Optimisation of Embedded Systems under Uncertainty in Probabilistic Reliability Evaluation Model Parameters

Software plays a vital role in most of the embedded systems including safety and mission-critical systems in avionics, automotive, nuclear and medical applications. Along with the functional complexity of software, the quality of software-intensive systems has become a crucial concern. Numerous techniques are being developed to evaluate the quality of a software system from its architecture in terms of quality attributes such as reliability, safety and performance, and to automate the search for alternative designs which provides good trade-offs with respect to those quality attributes of interest. However, the results of these quantitative architecture evalu-ations depend on design-time estimates for a series of model parameters, which may not be accurate and can change at run-time. Conventional approaches use numeri-cal values (point estimates) as design-time estimates, where the uncertainty in the parameter estimation is not part of the evaluation. As a result, architecture-based quality evaluations at design-time can be inaccurate and thus, sub-optimal design decisions may be taken