Automated assume-guarantee reasoning for omega-regular systems and specifications

Sagar Chaki, Arie Gurfinkel
2011 Innovations in Systems and Software Engineering  
We develop a learning-based automated Assume-Guarantee (AG) reasoning framework for verifying ω-regular properties of concurrent systems. We study the applicability of non-circular (AG-NC) and circular (AG-C) AG proof rules in the context of systems with infinite behaviors. In particular, we show that AG-NC is incomplete when assumptions are restricted to strictly infinite behaviors, while AG-C remains complete. We present a general formalization, called LAG, of the learning based automated AG
more » ... aradigm. We show how existing approaches for automated AG reasoning are special instances of LAG. We develop two learning algorithms for a class of systems, called ∞-regular systems, that combine finite and infinite behaviors. We show that for ∞-regular systems, both AG-NC and AG-C are sound and complete. Finally, we show how to instantiate LAG to do automated AG reasoning for ∞-regular, and ω-regular, systems using both AG-NC and AG-C as proof rules.
doi:10.1007/s11334-011-0148-1 fatcat:mozrp3cw3jgxpluvj2622bdhze