Re-use of Tests and Arguments for Assessing Dependable Mixed-criticality Systems

Abstract: 

The safety assessment of mixed-criticality systems (MCS) is a challenging activity due to system heterogeneity, design constraints and increasing complexity. The foundation for MCSs is the integrated architecture paradigm, where a compact hardware comprises multiple execution platforms and communication interfaces to implement concurrent functions with different safety requirements. Besides a computing platform providing adequate isolation and fault tolerance mechanism, the development of an MCS application shall also comply with the guidelines defined by the safety standards. A way to lower the overall MCS certification cost is to adopt a platform-based design (PBD) development approach. PBD is a model-based development (MBD) approach, where separate models of logic, hardware and deployment support the analysis of the resulting system properties and behaviour. The PBD development of MCSs benefits from a composition of modular safety properties (e.g., modular safety cases), which support the derivation of mixed-criticality product lines.

The validation and verification (V&V) activities claim a substantial effort during the development of programmable electronics for safety-critical applications. As for the MCS dependability assessment, the purpose of the V&V is to provide evidences supporting the safety claims. The model-based development of MCSs adds more V&V tasks, because additional analysis (e.g., simulations) would be carried out during the design phase. During the MCS integration phase, typically hardware-in-the-loop (HiL) plant simulators support the V&V campaigns, where test automation and fault-injection are the key to test repeatability and thorough exercise of the safety mechanisms.

This dissertation proposes several V&V artefacts re-use strategies to perform an early verification at system level for a distributed MCS, artefacts that later would be reused up to the final stages in the development process: a test code re-use to verify the fault-tolerance mechanisms on a functional model of the system combined with a non-intrusive software fault-injection, a model to X-in-the-loop (XiL) and code-to-XiL re-use to provide models of the plant and distributed embedded nodes suited to the HiL simulator, and finally, an argumentation framework to support the automated composition and staged completion of modular safety-cases for dependability assessment, in the context of the platform-based development of mixed-criticality systems relying on the DREAMS harmonized platform.