A Formal Methodology for Engineering Heterogeneous Railway Signalling Systems

Abstract

Over the last few decades, the safety assurance of cyber-physical systems has become one of the biggest challenges in the field of model-based system engineering. The challenge arises from an immense complexity of cyber-physical systems which have deeply intertwined physical, software and network system aspects. With significant improvements in a wireless communication and microprocessor technologies, the railway domain has become one of the frontiers for deploying cyber-physical signalling systems. However, because of the safety-critical nature of railway signalling systems, the highest level of safety assurance is essential. This study attempts to address the challenge of guaranteeing the safety of cyber-physical railway signalling systems by proposing a development methodology based on formal methods. In particular, this study is concerned with the safety assurance of heterogeneous cyber-physical railway signalling systems, which have emerged by gradually replacing outdated signalling systems and integrating mainline with urban signalling systems. The main contribution of this work is a formal development methodology of railway signalling systems. The methodology is based on the Event-B modelling language, which provides an expressive modelling language, a stepwise model development and a proof-based model verification. At the core of the methodology is a generic communication-based railway signalling Event-B model, which can be further refined to capture specific heterogeneous or homogeneous railway signalling configurations. In order to make signalling modelling more systematic we developed communication and hybrid railway signalling modelling patterns. The proposed methodology and modelling patterns have been evaluated on two case studies. The evaluation shows that the methodology does provide a system-level railway signalling modelling and verification method. This is crucial for verifying the safety of cyber-physical systems, as safety is dependent on interactions between different subsystems. However, the study has also shown that automatic formal verification of hybrid systems is still a major challenge and must be addressed in the future work in order to make this methodology more practical.