Addressing systemic risks in autonomous maritime navigation: A structured STPA and ODD-based methodology

Abstract

This paper presents a novel approach to managing unexpected risks for autonomous ships by addressing the unique challenges posed by their operational complexity and human–machine interactions. Autonomous ships operate in diverse environments, and their tasks span long-term voyage planning, short-term collision avoidance, and ship motion control, making traditional risk assessment methods insufficient. This paper proposes an enhanced methodology based on System-Theoretic Process Analysis (STPA), focusing on discrepancies in process models of several controllers to identify systemic risks in a structured manner. Additionally, the study introduces the importance of the Operational Design Domain (ODD) concept for autonomous ships, reflecting the specific conditions of maritime operations. It also proposes how to evaluate ODD and relevant loss scenarios extracted from STPA by applying several metrics for its safe control. Through a case study using Japan’s autonomous ship demonstration project, the methodology’s effectiveness is demonstrated by applying it to an autonomous container ship. Finally, the paper argues that by systematically refining ODD taxonomies and ensuring a holistic understanding of process models, autonomous ships can achieve safety assurance comparable to that of conventional vessels, despite the inherent challenges of maritime navigation.