Edelkamp, Stefan and Lluch-Lafuente, Alberto
Abstraction in directed model checking.
In: Workshop on Connecting Planning Theory with Practice, International Conference on Automated Planning and Scheduling, June 3-7 2004, Whistler, Canada
Abstraction is one of the most important issues to cope with large and infinite state spaces in model checking and to reduce the verification efforts. The abstract system is smaller than the original one and if the abstract system satisfies a correctness specification, so does the concrete one. However, abstractions may introduce a behavior violating the specification that is not present in the original system.
This paper bypasses this problem by proposing the combination of abstraction with heuristic search to improve error detection. The abstract system is explored in order to create a database that stores the exact distances from abstract states to the set of abstract error states. To check, whether or not the abstract behavior is present in the original system, effcient exploration algorithms exploit the database as a guidance.
Actions (login required)