Managing dynamic requirements knowledge: An agent-based approach

Agent- and goal-based requirements engineering can be considered established in research for many years now. Also first successful applications to industrial practice have been reported. Agent- and goal-based approaches explicate the functional and non-functional goals as well as various kinds of dependencies of possibly conflicting stakeholders. Thereby, they provide enhanced means to support elicitation, analysis, documentation, as well as many other operations on requirements. The thesis strives to add to these advanced support facilities by addressing dynamic issues that are not yet considered in existing approaches. Several dynamic aspects of the requirements field have been targeted by various research groups. For example, use cases and scenarios have been introduced to capture the interactive features of a system to be developed. From an entirely different perspective, the dynamics of the requirements engineering process itself has been investigated, for example to learn how the volatility of requirements can be addressed. Inspired by two very different case studies - support for flexible inter-organisational networks of enterprises and the elicitation and analysis of control software requirements in small- and medium-sized enterprises (SMEs) - we address several new dynamic issues in a number of extensions to the i* requirements modelling framework proposed by Yu. First, the requirements modelling language is extended to capture the dynamic instantiation of roles by stakeholders in a concrete project. Furthermore, these roles can be related to each other in regard to evolutionary aspects. This allows to capture that the characteristics of a stakeholder can change over time. Secondly, the capture, processing, and analysis of individual project requirements is enhanced. The explicit representation of domain knowledge accelerates the capturing procedure. Model-based transformations improve the integration with later development stages. In addition, they are used as a bridge toward agent-based simulations. Simulation experiments and advanced analysis on top of these complement well existing formal model checking approaches. Thirdly, we consider the inter-project management of dynamic requirements knowledge. A requirements-based similarity search helps to identify related historic projects and thus to disclose potentially reusable solutions. We have also developed measures to keep up with the fast and project-driven evolution of domain knowledge at SMEs. A partially automated feedback loop integrates repeated, consolidated project experiences of an SME into the earlier mentioned domain knowledge based approach. In sum, a tailorable method with accompanying tool support is established that addresses the raised dynamic issues. The validations within the two case studies have shown that in particular the work within very innovative, flexible, and customer-oriented settings benefits from the proposed extensions and thus brings forward industrial acceptance of agent- and goal-based approaches in these fields