Don't Know for Multi-valued Systems

This paper studies abstraction and refinement techniques in the setting of multi-valued model checking for the μ-calculus. Two dimensions of abstrace identified and studied: Abstraction by joining states of the underlying multi-valued Kripke structure as well as abstraction of truth values, for each following both an optimistic and pessimistic account. It is shown that our notion of abstraction is conservative in the following sense: The truth value in a concrete system is “between” the optimistic and pessimistic assessment. Moreover, ecking of abstracted systems is shown to be again a multi-valued model checking problem, allowing to reuse multi-valued model checking engines. Finally, whenever the optimistic and pessimistic model checking result differ, the r such an assessment is identified, allowing the abstraction to be refined to eventually yield a result for which both the optimistic and pessimistic assessment coincide.

• Type: Conference
• Year: 2009
• Where: ATVA
• Authors: Alarico Campetelli, Alexander Gruler, Martin Leucker, Daniel Thoma
• Keywords: Abstraction, ATVA 2009, Hardware, Model Checking, Multi-valued Model