Bayesian artificial intelligence for tackling uncertainty in self-adaptive systems: The case of dynamic decision networks

Abstract

In recent years, there has been a growing interest towards the application of artificial intelligence approaches in software engineering (SE) processes. In the specific area of SE for self-adaptive systems (SASs) there is a growing research awareness about the synergy between SE and AI. However, just few significant results have been published. This paper briefly studies uncertainty in SASs and surveys techniques that have been developed to engineer SASs in order to tackle uncertainty. In particular, we highlight techniques that use AI concepts. We also report and discuss our own experience using Dynamic Decision Networks (DDNs) to model and support decision-making in SASs while explicitly taking into account uncertainty. We think that Bayesian inference, and specifically DDNs, provide a useful formalism to engineer systems that dynamically adapt themselves at runtime as more information about the environment and the execution context is discovered during execution. We also discuss partial results, challenges and future research avenues.