Addressing Operator Performance in Shore-based Control Centres for Autonomous Shipping

The maritime industry is investing in advanced technologies to reduce its environmental footprint, become more attractive to personnel, improve its safety record, and enhance its resilience against adverse conditions, whilst maintaining profitability. To contribute to achieving these goals, Artificial Intelligence (AI) is anticipated to play a central role in supporting navigators in critical decision making and possibly even allowing ships to sail without direct human involvement. Considering the safety-critical nature of maritime navigation, this means that AI-enabled systems need to demonstrate a high degree of reliability and robustness across a wide range of situations. However, given the limitations of such systems to operate in novel and complex situations, careful design, implementation, management and supervision is required when deploying these in real-world environments. Therefore, proposed autonomous ship concepts typically employ human operators to monitor, supervise, and potentially intervene in the system to ensure the required performance and safety levels are achieved. Decades of research has demonstrated that there are significant human performance challenges associated with assigning humans a supervisory role of highly automated systems. Recent research has suggested that by disclosing the system’s decisions, planned actions, and internal reasoning to the operator, i.e., by making the system “transparent”, some of these challenges may be alleviated. However, considering the novelty of the application of AI-enabled systems in safety-critical domains, there is limited experience with the effect of transparency in these settings. As such, there is an urgent need to generate new knowledge, methods, and tools with regards to how humans may successfully interact with these types of technologies. Therefore, this dissertation aims to explore the following overarching research question: How does agent transparency support human performance in supervisory control? A mix of quantitative and qualitative methods were deployed in this dissertation. The literature review found a promising effect of transparency on SA and task performance, without affecting mental workload, for studies where participants were responding to proposals or supervising automation. A Goal-Directed Task Analysis mapped and analysed the goals, decisions, and cognitive tasks associated with conventional- and supervised collision and grounding avoidance. A model for human information processing was adapted and repurposed to function as a model for transparency. Based on the structured information requirements, HMI concepts were developed for making an autonomous collision avoidance system transparent for its supervisor. Finally, the effect of transparency on SA, mental workload, and task performance was evaluated using a controlled experimental approach. The results demonstrated a promising effect of transparency on SA without affecting mental workload. However, the time to comprehend the provided information increased with increased levels of transparency. These results indicate the benefits of applying transparency principles to autonomous collision avoidance systems in terms of SA, but care should be taken in time-critical conditions where the added transparency information may affect timely decision making. Furthermore, considering the absence of the effect of transparency on mental workload, these results also indicate the value of applying a structured and systematic human-centred design process, as applied in this dissertation. To conclude, the results have implications for scientific research and for the application of transparency as a design principle for autonomous agents. In addition, this dissertation has made explicit the role-change that may be anticipated when introducing autonomous systems. With these new insights, meaningful human work may be created where the combined capabilities of human-agent teams can be optimised. Ultimately, this dissertation advocates the relevance of affording human operators with insight into the reasoning of autonomous systems and established transparency as an important prerequisite on the path towards safe and effective human-supervisory control.

This dissertation investigated the role of agent transparency in supervisory control and contributed with knowledge, methods, and tools regarding transparency in general and its application to the maritime domain specifically. As humans are foreseen to play a critical role in overseeing the functioning of AI-enabled systems, the operator’s ability to understand, predict, and evaluate agent behaviour becomes a critical aspect of the human’s supervisory task repertoire. Consequently, it is essential that humans are informed and supported in making accurate decisions to enable timely and appropriate control when needed. Therefore, the aim of this dissertation was to generate and advance the knowledge on how supervisory control can be supported through agent transparency. This dissertation has contributed to this aim by recognising the importance of transparency in safety critical domains in terms of human performance, understanding the impact of autonomy on the operator’s cognitive tasks, constructing a model for transparency, operationalising transparency for the maritime context, and assessing its effects in an experimental setting. The results have implications for scientific research and for the application of transparency as a design principle for autonomous agents. In addition, this dissertation has made explicit the role-change that may be anticipated when introducing autonomous systems. With these new insights, meaningful human work may be created where the combined capabilities of human-agent teams can be optimised. Ultimately, this dissertation advocates the relevance of affording human operators with insight into the reasoning of autonomous systems and established transparency as an important prerequisite on the path towards safe and effective human-supervisory control.

The maritime industry is constantly looking for ways to reduce cost, increase safety and improve operational efficiency. Given current trends in technology development, it is expected that, in the near future, at least some functions on vessels will be fully automated. This has consequences for the roles of seafarers. The primary role of seafarers, especially bridge crew, is to maintain safety during the vessel’s voyage at all times. Changing the roles of seafarers as a result of technological advancements (i.e. automation) has consequences for how safety is ensured and how a shipowner can prove adherence to safety requirements. Shore-based control centres (SCC) are proposed for meeting safety requirements for remote monitoring and control of autonomous vessels (AV). Therefore, it is essential to understand, evaluate and assess SCC operator functions, tasks and performance when monitoring and controlling autonomous vessels. The primary objective of the research is to understand, evaluate and assess the performance of shore control center (SCC) operators when monitoring and controlling autonomous vessels. Secondary objectives include mapping the distribution of functions between vessels and SCC, what tasks the SCC operators are expected to perform, and what information and interaction requirements can be defined to enable task performance, including proposals for design. Given the novelty of autonomy in the Maritime industry, there is a need to generate knowledge and understanding the role of SCCs, and the functions and tasks operators perform in achieving safe operation of autonomous vessels. The outcomes of the research, therefore, is relevant for developing knowledge and concrete methods for designing for SCC solutions, development of international maritime regulations regarding autonomous vessels, the development of class requirements such that the performance and safety of the system as a whole is ensured.