Deep Learning Toward Autonomous Ship Navigation and Possible COLREGs Failures

Abstract

Abstract A structured technology framework to address navigation considerations, including collision avoidance, of autonomous ships is the focus of this study. That consists of adequate maritime technologies to achieve the required level of navigation integrity in ocean autonomy. Since decision-making facilities in future autonomous vessels can play an important role under ocean autonomy, these technologies should consist of adequate system intelligence. Such system intelligence should consider localized decision-making modules to facilitate a distributed intelligence type strategy that supports distinct navigation situations in future vessels as agent-based systems. The main core of this agent consists of deep learning type technology that has presented promising results in other transportation systems, i.e., self-driving cars. Deep learning can capture helmsman behavior; therefore, such system intelligence can be used to navigate future autonomous vessels. Furthermore, an additional decision support layer should also be developed to facilitate deep learning-type technologies, where adequate solutions to distinct navigation situations can be facilitated. Collision avoidance under situation awareness, as one of such distinct navigation situations (i.e., a module of the decision support layer), is extensively discussed. Ship collision avoidance is regulated by the Convention on the International Regulations for Preventing Collisions at Sea (COLREGs) under open sea areas. Hence, a general overview of the COLREGs and its implementation challenges, i.e., possible regulatory failures, under situation awareness of autonomous ships is also presented with the possible solutions. Additional considerations, i.e., performance standards with the applicable limits of liability, terms, expectations, and conditions, toward evaluating ship behavior as an agent-based system in collision avoidance situations are also illustrated.