A systems-theoretic approach to hazard identification of marine systems with dynamic autonomy

Autonomous marine systems may switch between various operational modes with different levels of autonomy (LoA), due to a rapidly changing environment and the complex nature of tasks. The dynamic autonomy brings an additional layer of complexity to ensuring safe marine operations, but this functionality is not sufficiently considered in current risk analysis methods. Hence, this paper proposes an approach to hazard identification based on the system theoretic process analysis (STPA) that includes unsafe transitions between different LoA in systems. A case study of a remotely operated vehicle (ROV) with four operational modes with different LoAs is used to illustrate the approach. The results show that the proposed approach contributes to: 1) communicating a shift of responsibilities among human operator and system controller in different operational modes by specifying how the allocation of the responsibility between human operators and the controller changes, and what updated process model of the operator and the controller are to ensure a successful transition; 2) refining safety constraints to be more concrete to improve system design, and operational procedures and 3) identifying triggering events for marine system modes' transitions to handle environmental interaction systematically and sufficiently.