Decision Support for Marine Traffic Control based on Route Clustering

Current paper is about navigation safety of marine traffic at sea areas. In addition to traditional approach of danger situation detection based on vessels approaching, the current paper introduces another metrics de-rived from kinematic parameters of the vessel itself, to identify whether it follows patterns (rules) of the traffic at a certain sea area. Authors focused their efforts on analyzing existing traffic schemas in order to identify its danger level in general rather than scrutinizing on individual cases. Along with the traditional approach of sea traffic schema identifications, proposed original method of automated identification of sea traffic schemes based on clustering of movement parameters using historical AIS data. The kinematic model with additional clustering of the objects at sea area is used. Clustering is built up on identification of “typical” objects with splitting maritime area into subsets of arias with similar movement patterns. For the clustering decomposition Mountain clustering and Subtraction clustering algorithms were considered. The model under research is built depending on the heading parameter and is applicable for both unidirectional and bidirectional traffic. The historical AIS data of sea traffic at Tsugaru strait is used for identifying traffic schema with the model designed under presented research. The proposed model shows correct identification of the traffic schema, accurately identifies straight-line traffic clustering areas crossing likes traffic clustering areas and active maneuvering areas. Based on model output results, the algorithm is proposed to help with navigation decision making for safe ship traffic in certain sea areas. The calculation of deviation between identified clusters and real ship traffic shows that about 15% of the vessels are not following sea area specific traffic schema. As a part of the work there are analyses of relevant research with emphasizes on superior differences such as automated identification of vessel movement types and quantitative measure of overall danger for the traffic scheme based on clustering analyses. Study of proposed approach for safety measures on real navigation area proves its potential for future implementation in real applications to provide navigation schema for vessel navigators.