A Causal Analysis Method for Concurrent Hybrid Automata

Proceedings of the AAAI Conference on Artificial Intelligence, 21

Volume

Technical Papers

Reasoning about Plans and Actions

Modern artifacts are typically composed of many system components and exhibit a complex pattern of continuous/discrete behaviors. A concurrent hybrid automaton is a powerful modeling concept to capture such a system's behavior in terms a concurrent composition of hybrid automata for the individual system components. Because of the potentially large number of modes of the concurrent automaton model it is non-trivial to validate the composition such that every possible operational mode leads to a causally valid dynamic model for the overall system. This paper presents a novel model analysis method that validates the automata composition without the necessity to analyze a prohibitively large number of modes. We achieve this by formulating the exhaustive causal analysis of hybrid automata as a diagnosis problem. This provides causal specifications of the component automata and enables us to efficiently calculate the causal relationships for their concurrent composition and thus validate a concurrent automaton model.