ARAS: adaptation requirements for adaptive systems

Supriana, Iping,Surendro, Kridanto

DOI: https://doi.org/10.1007/s10515-022-00369-3

IF: 1.677

2022-11-30

Automated Software Engineering

Abstract:Uncertainty is a major issue for system designers in developing an adaptive system, especially when defining adaptation requirements. The design-time requirement becomes invalid at run-time in the event of unforeseen circumstances due to the effects of unpredictable contextual variability. It is because contextual requirements are run-time uncertainty and a feature of unforeseen evolution. Various efforts have been made to realize adaptation requirements for adaptive systems, and there are already many mature works. However, the approach to handling uncertainty based on contextual requirements has not received sufficient attention, especially if it is integrated with the Bayesian approach. The problem that can arise because of this uncertainty is when the system's knowledge of contextual requirements becomes incomplete or inconsistent at run-time, the system cannot determine the choice of adaptation action. This paper introduces an approach to adaptation requirements for adaptive systems through an expanded goal-based modeling language with control loop patterns and context inheritance hierarchies to define contextual requirements, and their mapping of Bayesian approach expansion to determine adaptation behaviors related to context uncertainty at run-time. The simulation results show that the proposed model has provided an alternative way of responding to changes that are influenced by uncertainty based on contextual (functional) and non-functional requirements, either caused by false assumptions or other factors related to uncertainty. The evaluation results show that the proposed model can provide design support for adaptive systems at the level of architecture adaptability index = 0.81, and can handle domain variability and requirements evolution at run-time.