A study on the avoidance timing of autonomous surface ships through machine learning

초록

The technologies for autonomous navigation are steadily being developed for the International Maritime Organization level 4 fully unmanned ships. Autonomous surface ships must comply with the Convention on the International Regulations for Preventing Collisions at Sea (COLREGs), which requires them to perceive and judge their situation and generate a collision avoidance path for safe navigation. While COLREGs are designed to prevent collisions between individual vessels, they do not provide clear criteria for when to initiate collision avoidance, which is why maritime collisions continue to occur. This study analyzed the collision avoidance timing of navigators using Automatic Identification System (AIS) data and applied the findings to the collision avoidance system of autonomous ships. A machine learning approach was employed using a decision tree model to classify collision avoidance timing rules, which were then integrated into the collision avoidance system of autonomous surface ships. By analyzing collision avoidance timing through a machine learning model, a system was developed to determine avoidance points in various scenarios. The effectiveness of the proposed system was validated through simulations conducted in diverse and complex environments.

키워드