Reference
Abstract
A SAT-based incremental, inductive algorithm for model checking CTL properties is proposed. As in classic CTL model checking, the parse graph of the property shapes the analysis. However, in the proposed algorithm, called IICTL, the analysis is directed by task states that are pushed down the parse tree. To each node is associated over- and under-approximations to the set of states satisfying that node’s property; these approximations are refined until a proof that the property does or does not hold is obtained. Each CTL operator corresponds naturally to an incremental sub-query: given a task state, an EX node executes a SAT query; an EU node applies IC3; and an EG node applies FAIR. In each case, the query result provides more general information than necessary to satisfy the task. When a query is satisfiable, the returned trace is generalized using forall-exists reasoning, during which IC3 is applied to obtain new reachability information that enables greater generalization. When a query is unsatisfiable, the proof provides the generalization. In this way, property-directed abstraction is achieved.
BibTeX
@string{CAV = "International Conference on Computer Aided Verification (CAV)"}
@inproceedings{iictl-cav12,
author = {Zyad Hassan and Aaron R. Bradley and Fabio Somenzi},
title = {Incremental Inductive CTL Model Checking},
booktitle = CAV,
year = {2012},
}