CTL Model Checking Partially Specified Systems

Eshita Zaman, Christopher Johannsen, Andrew S. Miner, Gianfranco Ciardo, Samik Basu

Abstract: The behavioral specifications of a system are often partial at early stages of development due to pending design decisions. We view such specifications as being parameterized by these decisions and consider the problem of identifying a possible concretization, i.e., taking a subset of these decisions, to ensure that the system conforms to the desired requirements. We capture such a partially specified system as a partially-labeled Kripke structure (\(\texttt{plKS} \)), where certain propositions labeling the states of the Kripke structure correspond to pending design decisions, thus may be unknown. We then reduce the verification problem to model checking different \(\texttt{plKS} \) instances, each corresponding to a specific set of design decisions taken, inducing a semi-lattice on the instances. Central to our solution strategy is the effective and efficient exploration of this semi-lattice and the application of model checking techniques to \(\texttt{plKS} \)s with 3-valued semantics of temporal properties, to take into account unknown state labels. We also address the problem of identifying an optimal instance of a partially-specified system that satisfies a desired property, where the cost of design decisions drive the optimality criterion. We use a prototype implementation of our strategy to validate its viability on a multi-objective path planning problem.

External IDs:dblp:conf/ifm/ZamanJMCB25