A Benchmark for Component-based Hybrid Systems Safety Verification

Abstract

At scale, formal verification of hybrid systems is challenging, but a potential remedy is the observation that systems often come with a number of natural components with certain local responsibilities. Ideally, such a compartmentalization into more manageable components also translates to hybrid systems verification, so that safety properties about the whole system can be derived from local verification results. We propose a benchmark consisting of a sequence of three case studies, where components interact to achieve system safety. The baseline for the benchmark is the verification effort from a monolithic fashion (i.e., the entire system without splitting it into components). We describe how to split the system models used in these case studies into components with local responsibilities, and what is expected about their interaction to guarantee system safety. The benchmark can be used to assess the performance, automation, and verification features of component-based verification approaches.