The Reliability of Embedded Software-Based Systems: Novel Ways to Enhance and Predict

Embedded systems have become increasingly popular over the past years, both in civil and military applications. The probability of an embedded system to perform the required operations during the specified period of time, i.e., the reliability, depends on both the underlying hardware reliability and software reliability. The traditional approach to predict the software reliability on the basis of the programmatic errors implies specially designed testing and predictive models. The main problem of this approach is that it usually considers a parametric model with the parameters to be estimated during the testing, whereas there is no physical or mathematical law underlying the parametric model itself. The proposed method originates from a Bayesian approach so that its dependence on the initial underlying parametric model is relatively weak. At the same time, it accounts for the previous/expert data in a most effective way. To enhance the reliability essential in safety-critical embedded systems, we will extend our method of conversion the informal design requirements into the reference specifications via the EFSM (Extended Finite State Machines).