Big Data Analytics for Assistance Operation Identification in Ice-Covered Waters

Abstract In winter, shipping activities in the Baltic Sea area involve risks because of the harsh navigational environment. To ensure maritime safety and the efficiency of the transport system, icebreaker assistance is needed when the ship is not able to navigate independently in the prevailing ice condition. The existing ship operation mode determination practices are primarily based on empirical knowledge. To better understand when the assistance operation is needed for various ships operating in real ice conditions from a quantitative perspective, this paper proposes the use of a data-driven assistance operation identification model. The model combines different data types, including Automatic Identification System (AIS) data and ice data that portray the real operational conditions in the Baltic Sea. The aim is to measure temporal-spatial similarities among trajectories of icebreakers and potentially assisted ship(s) using clustering algorithms, automatically identifying assistance scenarios in ice-covered waters. Both the static voyage features (e.g., ship types, departures /destinations, voyage length) and the dynamic navigation features (e.g., speed, course, and ship trajectory spatial distance) are considered in the model. The outcomes of the model would assist in better understanding of ship behaviors and coupled kinematics of ships linking with the following distance and speed under icebreaker assistance in various ice conditions. They would also contribute to data-driven ship operation mode determination, therefore supporting the development of an intelligent decision support system in winter navigation.