Maritime shipping is essential to the global economy, while waterway transportation is recognized as a high-risk industry. Additionally, maritime accidents are frequently caused by human errors, and with the rapid improvement of science and technology, the improvement of autonomous ships has been technically feasible, which attracts the wide attention of researchers in academia and industry. However, the knowledge acquisition and representation methods are mainly based on knowledge-based research methods, while the existing research for automatically achieving the autonomous ships’ maneuvering decision-making by acquiring the seafarers’ operation characteristics is still scanty. In addition, it also lacks the appropriate theoretical methods to explore the problem of autonomous ship human-like maneuvering decision-making modeling. Therefore, the research on ship maneuvering decision-making methods still needs to be improved and further developed. This thesis focuses on the problem of modeling seafarers’ navigational decision-making in a typical scenario for autonomous ships’ safety. We propose the method to prioritize safety influencing factors of autonomous ships’ maneuvering decisions and a series of ship maneuvering knowledge learning models to give the autonomous ship the ability to make decisions like a human. The autonomous ship human-like maneuvering decision-making problem has been considered as a machine learning problem, and we translate the problem into learning the maneuvering decision characteristics of the officer on watch (OOW) using various decision tree algorithms. By constructing autonomous ship human-like decision-making maneuvering decision recognition models under multiple constraints in the specific scenarios, the decision-making mechanism of the OOW’s maneuvering behavior under specific water traffic safety influencing factors in the inbound scenario is analyzed, and the OOW’s decisionmaking knowledge is automatically acquired and represented...
|Award date||19 Jan 2022|
|Publication status||Published - 2022|