Skill-Based Verification of Cyber-Physical Systems

Abstract

Cyber-physical systems are ubiquitous nowadays. However, as automation increases, modeling and verifying them becomes increasingly difficult due to the inherently complex physical environment. Skill graphs are a means to model complex cyber-physical systems (e.g., vehicle automation systems) by distributing complex behaviors among skills with interfaces between them. We identified that skill graphs have a high potential to be amenable to scalable verification approaches in the early software development process. In this work, we suggest combining skill graphs with hybrid programs. Hybrid programs constitute a program notation for hybrid systems enabling the verification of cyber-physical systems. We provide the first formalization of skill graphs including a notion of compositionality and propose S k e d i t o r, an integrated framework for modeling and verifying them. S k e d i t o r is coupled with the theorem prover K e Y m a e r a X, which is specialized in the verification of hybrid programs. In an experiment exhibiting the follow mode of a vehicle, we evaluate our skill-based methodology with respect to savings in verification effort and potential to find modeling defects at design time. Compared to non-compositional verification, the initial verification effort needed is reduced by more than 53%.

Runtime Verification of Correct-by-Construction Driving Maneuvers

“…The concrete models are contained as examples in the tool suite Skeditor [21], which integrates this work as part of its maneuver-centric development approach for cyber-physical systems. 2…”