Runtime Safety Assurance for Adaptive Cyber-Physical Systems

Abstract

Cyber-Physical Systems (CPS) provide their functionality by the interaction of various subsystems. CPS usually operate in uncertain environments and are often safety-critical. The constituent systems are developed by different stakeholders, who – in most cases – cannot fully know the composing parts at development time. Furthermore, a CPS may reconfigure itself during runtime, for instance in order to adapt to current needs or to handle failures. The information needed for safety assurance is only available at composition or reconfiguration time. To tackle this assurance issue, the authors propose a set of contracts to describe components' safety attributes. The contracts are used to verify the safety robustness of the parts and build a safety case at runtime. The approach is applied to a use case in the automotive domain to illustrate the concepts. In particular, the authors demonstrate safety assurance at upgrade and reconfiguration on the example of ontology-based runtime reconfiguration (ORR). ORR substitutes a failed service by exploiting the implicit redundancy of a system.