Efficient COLREG-Compliant Collision Avoidance in Multi-Ship Encounter Situations

Ship collisions are major types of maritime accidents which may involve the loss of life and significant damage to property and the environment. Although many automatic ship collision avoidance algorithms have been suggested, most of them are only applicable to a single ship-to-ship encounter situation. Also, although there exist some studies on collision avoidance for multiple agent systems, maritime traffic rules have not been systematically incorporated in the algorithms which limit their practical applicability to real maritime traffic situations. In this study, we propose a rule-compliant automatic ship collision avoidance method that can be applied not only to single ship-to ship situations, but also to multiple-ship encounter situations with consideration of prediction uncertainty. In order to select appropriate evasive actions, a symmetric role-classification criterion is proposed by refining the current maritime traffic rules, and an efficient collision avoidance algorithm based on the probabilistic velocity obstacle method is applied. To verify and demonstrate the performance and practical utility of the proposed algorithm, Monte-Carlo simulations were conducted and the results are presented in this article.