Learning-Aided Intelligent Cooperative Collision Avoidance Mechanism in Dynamic Vessel Networks

With the increasing traffic density and the highly maritime safety requirements, it is a great challenge to make the optimal motion decision for avoiding the collision in the complex vessel networks. However, the current collision avoidance mechanisms are only considering unilateral static information, ignoring the multilateral dynamic information, which is quite important. In this paper, we investigate the information transmission and the collision avoidance problem in the complex vessel networks, in which the devices are equipped with the capability of Artificial Intelligence (AI). With the empowered learning ability in the networks, we propose a novel two-step-smooth-turn cooperative collision avoidance mechanism for the dynamic vessel networks, considering the motion states of the vessels. In particular, the AI-powered vessels network motion model is designed, which can learn the motion information to predict the safety motion states by automatic collection of the information. Then, a K-Means algorithm combining with genetic algorithm is proposed by expanding the operation of cross genetic and genetic mutations, which can achieve the optimal multi-vessels coordination motion policy for the collision avoidance. Simulation results show that the proposed approach could improve the stability of genetic algorithm and weaken the early convergence of genetic algorithm efficiently. In addition, it shows that the proposed mechanism can achieve an optimal safety route plan for cooperative collision avoidance.