Reasoning about Requirements at Run-time Using AI Planning Techniques

Abstract

The emerging domain of Self-Adaptive Systems (SAS) has gained significant importance in software engineering community over the recent years. Self-adaptive application by definition should modify itself at run-time as a response to the changes in system environment or changes in system\user requirements [14]. Nowadays mobile software applications are being widely used. Such applications must ensure high customizability and at the same time effective reasoning to meet their objectives so that their end-user goals are met. Explicitly, they should be able to: (i) reason about their own requirements and refine and validate them at run-time by involving end-users [16] (ii) provide solutions for the refined or changed requirements at run-time, for instance by using available services [10]. In short, requirements engineering for adaptive mobile applications requires efficient techniques which is a major challenge. In this context, my thesis will focus on extending the reasoning capabilities in Continuous Adaptive RE framework i.e. [10, 16] using AI techniques. In order to understand the aim of this area of research, initially we focused on requirements problem i.e. requirement specification that entails the satisfaction of end user goals at run-time, which is considered as a planning problem. Therefore, methods that can support run-time reasoning of requirements are desirable. We focused on recently proposed techniques for automated reasoning with requirements goals and preference models to support run-time adaptations. These techniques are integrated into CARE (Continuous Adaptive RE), [10] which is a framework for requirements based engineering of self-adaptive system and continuous reappraisal of adaptive requirements at run-time. CARE framework has different components. In this thesis work, we are working on reasoning component of CARE framework, which provides effective decision making based on end-user preferences and goal models, supported by AI planning techniques subsequently providing reasoning about new solutions to the requirements problem. We envision that the utmost advantages of our approach are that the decision making process of self-adaptive system align directly with human accessible requirements models, facilitating thereby systematic engineering and accessibility both at design time and at run-time [10].