Reflective self-adaptive systems

"We propose a software development framework where the system under construction can reason about what its components do and how they do it. In order to satisfy end-user needs and design constraints, the system determines needed changes in its configurationby reasoning with the components' what-and-how knowledge. System re-configuration is triggered by changes in user constraints or changes in the system environment that make the system's behavior violate these constraints. The system monitors its behaviorin order to signal constraint violations, diagnoses the cause of these violations, and triggers the corresponding repair adaptation. We proposed to use artificial intelligence model-based reasoning techniques for implementing these monitoring, diagnosisand re-configuration facilities. These techniques rely on domain knowledge provided by end-users and system designers. This knowledge includes repair heuristics, hierarchical component descriptions, probes and configuration transformations. An intelligentknowledge acquisition tool will guide end-users in order to define this domain knowledge without the assistance of programmers. The system's introspection facilities allow it to explain in user terms why its configuration is as it is, helping the user todetect weaknesses in the system domain knowledge. We will demonstrate the feasibility of our ideas through the development of a Phase I prototype. The proposed technology can be markete