The Generation and Evolution of Adaptation Rules in Requirements Driven Self-adaptive Systems

One of the challenges in self-adaptive software systems is to make adaptation plans in response to possible changes. A good plan mechanism shall have the capability of: 1) selecting the most appropriate adaptation actions in response to changes both in the environment and requirements; 2) making adaptation decisions efficiently to react timely to arising situations at run-time. In existing approaches for plan process, rulebased adaptation provides an efficient offline planning method. However, it can react neither to changeable requirements nor to unexpected environment changes. On the contrary, goalbased and utility-based approaches provide online planning mechanisms, which can well handle a highly uncertain environment with dynamically changing requirements and environment. However, online adaptation decision making is often computationally expensive and may encounter less-efficiency problems. The aim of our research is to improve the planning process in requirements driven self-adaptive systems, i.e., enabling the self-adaptive system to efficiently make adaptation plans to cope with the dynamic environment and changeable requirements. To achieve such advantages, we propose a solution to enhance the traditional rule-based adaptation with a rule generation and a rule evolution process, so that the proposed approach can maintain the advantages of efficient planning process while being enhanced with the capability of dealing with runtime uncertainty.