Crash Stop Maneuvering Performance Prediction: a Data-Driven Solution for Safety and Collision Avoidance

Abstract

The continuous increase of marine traffic and the entry of autonomous ships into the market is urging an improvement in safety measures to guarantee avoidance of collisions between moving objects at sea. This rise in automated maneuverability requires gaining further insight in the vessel’s behavior. The ship design has to ensure that the vessel is controllable and capable of maneuvering securely, even at critical operating conditions. Crash stop maneuvering performance is one of the key indicators of the vessel’s safety properties for designers and shipbuilders. Many factors affect this performance, from the hull design to the environmental conditions; hence, it is non-trivial to assess them accurately during the preliminary design stages. In this paper, the authors focus on predicting accurately and with minimal computational effort the crash stop characteristics of a vessel in the design stage, for the preliminary assessment of safety requirements imposed by the classification societies. The crash stop prediction model of the said vessel can be utilized in combination with collision avoidance algorithms. The authors propose a new data-driven method, based on the popular Random Forests learning algorithm, for predicting the crash stop maneuvering performance. Results from full-scale measured data show the effectiveness of the proposed method.