Learning-Based Symbolic Assume-Guarantee Reasoning with Automatic Decomposition

Compositional reasoning aims to improve scalability of verification tools by reducing the original verification task into subproblems. The simplification is typically based on the assume-guarantee reasoning principles, and requires decomposing the system into components as well as identifying adequate environment assumptions for components. One recent approach to automatic derivation of adequate assumptions is based on the L∗ algorithm for active learning of regular languages. In this paper, we present a fully automatic approach to compositional reasoning by automating the decomposition step using an algorithm for hypergraph partitioning for balanced clustering of variables. We also propose heuristic improvements to the assumption identification phase. We report on an implementation based on NuSMV, and experiments that study the effectiveness of automatic decomposition and the overall savings in the computational requirements of symbolic model checking.