A Fast Approach of Locating Complex System Design Errors

Abstract

Model checking technique can give a specific counterexample which explains how the system violates some assertion when model does not satisfy the specification. However, it is a tedious work to understand the long counterexamples. We propose a genetic algorithm to enhance the efficiency of understanding long counterexample by computing the minimal unsatisfiable subformula. Besides, we also propose a Craig interpolation computation-based method to understand counterexample. The causes which are responsible for model failure are extracted by deriving interpolation from the proof of the nonsatisfiability of the initial state and the weakest precondition of counterexample. Experimental results show that our methods improve the efficiency of understanding counterexamples and debugging significantly.