Effect of dynamic safety distance of heterogeneous traffic flows on ship traffic efficiency: A prediction and simulation approach

Abstract

Compared to the heterogeneous traffic flow on roads, the heterogeneous characteristics of maritime traffic flow are more pronounced, due to disparities in the manoeuvrability, size, and safety requirements among different ships. These factors increase the complexity of quantifying traffic efficiency. This paper employs hierarchical clustering to categorise trajectories with similar speed variation features and utilises Long Short-Term Memory (LSTM) models to predict ship speeds, forming a speed control strategy based on time-series data. Moreover, extending from the principles of car-following, models are developed to compute dynamic safety distances for both manned and autonomous ships. Further, a port waterway simulation model based on cellular automaton (CA) is developed, integrating data-driven speed control strategies while maintaining dynamic safety distances, resulting in a comprehensive simulation model with dual speed control mechanisms. A case study of the Tianjin port shows the advantages of dynamic safety distances. Dynamic safety distances significantly improved ship navigation and overall channel traffic efficiency. The inclusion of autonomous ships further improves the efficiency and the benefits are directly related to the proportion of autonomous ships. This research provides new insights and methodologies for assessing transport efficiency and waterway capacity, which also offers support for research on autonomous maritime traffic flow.