Modelling Human-Machine Interaction Risks in Remote Maritime Autonomous Surface Ship Operations Using System Theoretic Process Analysis and Bayesian Networks

The safe operation of next-generation autonomous ships relies on effective Human-Machine Interactions (HMIs), particularly in remote operations where operators supervise from Remote Operation Centres (ROCs). The increasing interaction complexity among operators, automation, and stakeholders within Maritime Autonomous Surface Ships (MASS) ecosystem introduces emerging risks that remain challenging to capture using conventional risk assessment methods. The aim of this paper is to demonstrate a framework for analysing and quantifying HMI-related risks in remote ship operations for MASS degree of autonomy (DoA) level 3 by combining System-Theoretic Process Analysis (STPA) with Bayesian Network (BN) modelling. A concept of operations (ConOps) based case study of a remotely controlled ship was developed to capture critical interaction scenarios between ROC operators, autonomous ship systems, and operational stakeholders within the MASS ecosystem such as vessel traffic services (VTS), tugboat operators and port authorities. STPA was applied to systematically identify hazards, unsafe control actions, and interaction-related causal pathways in the remote operation. The causal factors identified by STPA were then organised into a set of Human-Machine Interaction Risk Influence Factors (HMI-RIFs), which formed the basis for constructing the Bayesian Networks. Model outputs are evaluated using a Normalized Risk Index (NRI) which converts probabilistic outcomes into a scalar metric for comparing operational scenarios. The results demonstrate that the proposed approach provides a structured and transparent representation of HMI-related risk. Sensitivity analysis indicates that Information Quality and Operator Workload are the dominant contributors to HMI-related risk. The novelty of this work lies in demonstrating a STPA-BN framework for analysing human-machine interaction risks in remote MASS operations, enabling a transparent transition from qualitative hazard analysis to quantitative probabilistic modelling and scenario-based risk comparison.

Keywords: Maritime Autonomous Surface Ships, Human-Machine Interactions, System Theoretic Process Analysis, Bayesian Networks, Risk Modelling.