Abstract
In this paper, we propose a property-specific sequential invariant extraction algorithm to improve the performance of the SAT-based Unbounded Modeling Checkers (UMCs). By analyzing the property-related predicates and their corresponding high-level design constructs such as FSMs and counters, we can quickly identify the sequential invariants that are useful in improving the property proving capabilities. We utilize these sequential invariants to refine the inductive hypothesis in induction-based UMCs, and to improve the accuracy of reachable state approximation in interpolation-based UMCs. The experimental results show that our tool can outperform a state-of-the-art UMC in most cases, especially for the difficult true properties.