Mitigating Unsafe Control Actions in Autonomous Navigation Systems: A SysML-Based Analysis for Enhanced Safety

The increasing complexity of autonomous navigation systems (ANS) poses significant challenges to ensuring safety and reliability, particularly in dynamic and high-risk environments. This paper presents a SysML-based STPA methodology that enhances hazard analysis efficiency, traceability, and integration within system design. Unlike traditional approaches that rely solely on STPA control structures, this method replaces them with SysML diagrams, providing a more structured and dynamic representation of system interactions over time. Sequence Diagrams are used to explicitly depict control actions and feedback, improving the identification of unsafe control actions (UCAs) and their causal factors, such as software errors, communication failures, and human errors. Additionally, this approach explores loss scenarios, which have not been addressed in previous studies. The proposed methodology is applied to an ANS operating in winter conditions in the Baltic Sea. This integration of SysML and STPA offers a unified framework for system and safety engineering, reducing analysis time while improving scalability and applicability to complex autonomous systems.