Run-time goal-model artifacts represent a notable approach to communicate requirements to the system and open new directions for dealing with self-adaptation.

This work presents a theoretical framework and a general architecture for system evolution, self-configuration and self-healing. The novelty is that of breaking design-time constraints between system goals and tasks. The user may inject, at run-time, goal-models that do not contain tasks, i.e. the description of how to address them. Therefore, the architecture is responsible to configure its components as the result of deductions made at the knowledge level. The strength of this architecture is to promote reusability and domain independence.

Finally, the proposed implementation of the architecture has been evaluated in the context of self-configuration and self-healing through the execution of a set of randomized stress tests.