An ontology to aid the goal-oriented requirements elicitation and specification for self-adaptive systems

Abstract

Self-Adaptive Systems (SAS) can adapt their own behavior in response to context information or changes in the environment and also in response to their own behavior. The interest in requirements engineering for SAS has grown in recent years, but despite this, some works involving requirements specification of these systems do not guide requirements elicitation. Goal-oriented requirements engineering (GORE) modeling languages are widely used to specify requirements for SAS. There are GORE modeling languages specifically proposed for the SAS domain and each of them presents a fixed and small set of concepts. Ontologies can be used to overcome the limitation of concepts, since they can help in the representation of concepts within a domain, as well as in the communication and specification of requirements. The purpose of this thesis is to provide a richer set of SAS concepts to guide the elicitation and specification of requirements for such systems. An ontology for SAS is proposed, as well as a process to guide the use of the ontology for eliciting and specifying requirements for SAS. The unique core ontology for requirements for SAS in literature does not cover all main concepts that SAS involves, like the modeling dimensions and a feedback loop. In order to achieve the objective, first, two systematic literature reviews (SLRs) were performed to analyze the work involving knowledge representation for SAS and context-aware systems. A total of twenty-three studies were selected in both. Then, three GORE modeling languages for SAS were analyzed - Tropos4AS, AdaptiveRML and Design Goal Model - to identify the concepts that these languages are able to represent. It was observed that the analyzed languages do not represent most of the concepts involved in the SAS domain. With the results of both SLRs and the analysis of the GORE modeling languages, an ontology was proposed to aid the requirements engineer to perform the elicitation and specification of SAS. To create the ontology, three methodologies were used: Uschold and Gruninger’s, METHONTOLOGY and SABiO. The proposed ontology covers the main concepts of self-adaptive systems, such as the feedback loop concepts, context, the modeling dimensions for SAS, and goal-oriented requirements. It was also proposed a process for the use of the ontology. The ontology evaluation was based on six criteria: comprehensiveness, verification, validation, utility, easiness of use and accordance. Our ontology is embracing in comparison to the related works selected in both SLRs. The ontology was verified and validated by instantiating a multimedia news system. The usefulness and easiness of use of both the ontology and the process were evaluated by case study and a survey, where requirements engineers used the process to instantiate an ambulance dispatch system. This evaluation found that the ontology is useful, although the process is not so easy to use. Another survey was answered by SAS specialists to evaluate the accordance of the ontology, who agreed with the concepts of the ontology.