Open Multi-Scale Self-Explanations for Complex Cyber-Physical Systems

Cyber-Physical Systems (CPS), such as smart buildings or power grids, are increasingly dynamic, large-scale, and self-adaptive. As their complexity grows, so does the need for explanations that help users and developers understand unexpected system behaviors. However, generating such explanations is particularly difficult in CPS due to their open architectures, multiple levels of abstraction, and evolving behavior. This thesis addresses the challenge of producing multi-scale self-explanations, that is, runtime explanations across abstraction levels, generated by the system itself, in response to surprising or undesired events. Contributions include: (1) a roadmap for causal modeling in Complex Adaptive Systems (CAS), (2) a method to generate causal explanations in CPS, (3) an approximate algorithm for identifying actual causes, and (4) a mechanism for filtering causes based on user-relevant metrics. Future work will focus on integrating these components into a unified explanation framework for open, adaptive, multi-scale CPS.