Fault-based analysis of cyber physical systems

Abstract

Cyber-physical systems (CPS) consist of computational networked and distributed components monitoring and controlling the physical environment. They are nowadays ubiquitous in many safety-critical applications including autonomous vehicles or smart medical devices. Therefore ensuring the correctness of CPS before their deployment is essential. The complexity and heterogeneity of CPS also makes them vulnerable to faults that are challenging to detect, understand and correct. These faults can propagate in time and space, producing observable misbehavior that is hard to trace to its origin because it is typically a result of complex combination of events. In this thesis, we develop methods for failure analysis of a CPS that help exposing and explaining faults. We develop theoretical, methodological and practical solutions to achieve this goal by focusing on the following topics: (1) methods to localize and explain failures, and (2) testing procedures that facilitate fault analysis. The scientific outcomes of this were implemented as prototypes and tools. The solutions developed in this thesis have been evaluated on publicly available benchmarks and industrial use cases from several domains to demonstrate the applicability of our research.