Autonomous ships are gaining in importance and are expected to shape the future of the global shipping industry. This evolutionary shift raises serious issues about compliance with the International Regulations for Preventing Collisions at Sea 1972 (COLREGs). This paper reviews the literature on autonomous ships from the perspective of the obligations of good seamanship imposed by COLREGs. The authors conclude that to facilitate the introduction of autonomous ships, the application barriers presented by COLREGs need to be analysed. With this goal, this paper presents a perspective from navigational practice. Four nautical scientists and two deck officers were invited to give their opinions. The analysis indicates that COLREGs require further elaboration and amendments to eliminate uncertainty of interpretation. In particular, the paper highlights the need to amend the ‘look-out’ rule (COLREGs Rule 5) to permit look-out by ‘computer vision’ alone while, at the same time, preserving the distinction between vessels navigating in restricted visibility and in sight of one another.

Forecasting vessel flows is important to the development of intelligent transportation systems in the maritime field, as real-time and accurate traffic information has favorable potential in helping a maritime authority to alleviate congestion, mitigate emission of GHG (greenhouse gases) and enhance public safety, as well as assisting individual vessel users to plan better routes and reduce additional costs due to delays. In this paper, we propose three deep learning-based solutions to forecast the inflow and outflow of vessels within a given region, including a convolutional neural network (CNN), a long short-term memory (LSTM) network, and the integration of a bidirectional LSTM network with a CNN (BDLSTM-CNN). To apply those solutions, we first divide the given maritime region into $M\times N$ grids, then we forecast the inflow and outflow for all the grids. Experimental results based on the real AIS (Automatic Identification System) data of marine vessels in Singapore demonstrate that the three deep learning-based solutions significantly outperform the conventional method in terms of mean absolute error and root mean square error, with the performance of the BDLSTM-CNN-based hybrid solution being the best.

The autonomous shipping industry is increasingly focusing on enhancing the safety and reliability of software-based systems. Conducting a risk assessment is a requirement for demonstrating the safety equivalence of autonomous ships based on such systems to conventional vessels. Traditional risk assessment models, however, primarily focus on hardware failures, often overlooking potential software-related failures and functional inadequacies. This study proposes a framework integrating Software Failure Mode and Effects Analysis (FMEA), System–Theoretic Process Analysis (STPA), and Bayesian Network (BN) for risk identification of autonomous ship software systems. The results of a case study reveal that the framework sufficiently addresses the multifaceted nature of risks related to software in autonomous ships. Based on the findings of this study, we suggest the need for standardization of software architecture development in the autonomous ship industry and highlight the necessity for an enhanced understanding of AI-specific risks and the development of tailored risk assessment methodologies.

Unsafe crew acts (UCAs) related to human errors are the main contributors to maritime accidents. The prediction of unsafe crew acts will provide an early warning for maritime accidents, which is significant to shipping companies. However, there exist gaps between the prediction models developed by researchers and those adopted by practitioners in human risk analysis (HRA) of the maritime industry. In addition, most research regarding human factors of maritime safety has concentrated on hazard identification or accident analysis, but not on early warning of UCAs. This paper proposes a Bayesian network (BN) version of the Standardized Plant Analysis Risk–Human Reliability Analysis (SPAR-H) method to predict the probability of seafarers’ unsafe acts. After the identification of performance-shaping factors (PSFs) that influence seafarers’ unsafe acts during navigation, the developed prediction model, which integrates the practicability of SPAR-H and the forward and backward inference functions of BN, is adopted to evaluate the probabilistic risk of unsafe acts and PSFs. The model can also be used when the available information is insufficient. Case studies demonstrate the practicability of the model in quantitatively predicting unsafe crew acts. The method allows evaluating whether a seafarer is capable of fulfilling their responsibility and providing an early warning for decision-makers, thereby avoiding human errors and sequentially preventing maritime accidents. The method can also be considered as a starting point for applying the efforts of HRA researchers to the real world for practitioners.