Decision-making of Vessel Collision Avoidance Based on Support Vector Regression

Abstract

Regardless of numerous collision avoidance regulations to prevent collisions of vessels, accidents still happen. The collision avoidance decision system is an important part of intelligent ship applications, as it can provide decision support to avoid collision accidents. In this paper, using the encounter samples extracted from Automatic Identification System (AIS) data, a vessel collision avoidance decision-making model is developed by the Support Vector Regression (SVR) approach. During the model training and validation tests, the SVR model has high prediction accuracy and solves the nonlinear problem with the multiple motion parameters and vessel collision avoidance behavior in different encounter situations. However, due to the sensitivity of the model to the magnitude of collision avoidance behavior, prediction errors are inevitable. These findings can improve the real-time performance of the collision avoidance decision-making and illustrate the necessity of collision avoidance behavior in real situations. It provides a reference to collision avoidance action and decision guidance of vessel autonomous driving systems.