Engineering Self-Adaptive Systems with the Role-Based Architecture of HELENA

When engineering self-adaptive systems, separating adaptation and application logic was proven beneficial to avoid interdependencies between adaptation strategy and standard behavior. Several engineering methods support this separation in different phases of the classical development process, but none addresses it consistently in all of them. We propose a holistic model-driven engineering process with systematic transitions between all phases to develop self-adaptive systems. Adaptation is achieved by changing the behavioral mode of a component in response to perceptions. We realize behavioral modes by roles which a component can dynamically adopt. For specification, we propose adaptation automata which allow to specify complex adaptation behavior by hierarchical structure and history of states. Furthermore, we propose the HELENA Adaptation Manager pattern to derive a role-based model from a specification. Due to its formal foundation, the model can be analyzed with Spin and executed with the Java framework jHELENA.