Interfaces and models for the diagnosis of cyber-physical ecosystems

When the multi-agent paradigm is applied to cyber-physical ecosystems, the diagnosis of physical failures and unexpected interference must be handled by the agents' internal adaptive task planning or external maintenance. In both cases, a profound knowledge of potential dependencies and their run-time manifestations is required. The open, evolutionary aspect of such systems adds to the complexity of failure interaction patterns. As a solution, we propose to adopt techniques from the systemlevel diagnosis of automation systems. By extending existing MAS metamodels with explicit resource and dependency models, an integrated diagnostic knowledge base can be established. The diagnostic viewpoint provides interfaces with enhanced semantics based on causality and symptom propagation and transformation. The formalization enables the checking of diagnostic consistency based on interface compatibility. Results are demonstrated within a smart airport transport scenario.