Towards a systems framework for the assurance of maritime autonomous systems

ABSTRACT The burgeoning adoption of Robotic and Autonomous Systems with Artificial Intelligence (RAS-AI) necessitates standardized approaches for Systems Engineering (SE) and assurance of RAS-AI systems. Unlike crewed systems, RAS-AI systems lack real-time human control, posing unique challenges yet to be addressed by existing regulations and SE frameworks. This paper proposes the ”Autonomous Delta” approach for the Assurance of RAS-AI systems (As4AI), focusing on Maritime Autonomous Systems (MAS). It comprises seven parts, contextualizing RAS-AI assurance in an Australian maritime context, providing key definitions, a literature review, and outlining the research methodology. The core of the paper (Part 5) presents an As4AI framework, integrating concepts from SE and high-fidelity flight simulation. A prototype instantiation at the Australian Institute of Marine Science confirmed the framework's utility and identified areas for refinement. The paper concludes with a summary and suggestions for future research enabling operational deployment of MAS specifically and autonomous systems in general.