Compositional Realizability Checking within FRET

Abstract

A set of requirements for a reactive system is realizable if, for any sequence of inputs that satisfy the assumptions on the environment, the guarantees always hold. Realizability checking is essential to ensure that an implementation can be constructed that satisfies the requirements. We propose a framework that supports users in the non-trivial task of developing realizable requirements. Our framework uses architectural information to automatically de-compose a set of requirements into subsets that can be analyzed separately, and therefore more efficiently. It then integrates existing algorithms in order to detect unrealizability, identify minimal sets of conflicting requirements, and compute counterexamples. The capability to focus on minimal conflict sets is key for localizing and correcting the sources of unrealizability. Our approach supports this process by enabling users to interactively visualize and explore the produced conflict sets and counterexamples. We have implemented our framework in the open-source Formal Requirements Elicitation Tool (FRET), and have used it on a variety of industrial-level case studies, showcasing the strengths of our approach in terms of raw performance, as well as diagnostic potential.