Neural domains and game strategies in collision risk mapping in algorithms for safe control in multi-autonomous ship situations

As maritime transport technology has shifted towards the greater use of autonomous ships, the safety requirements for their movement are growing. A gap in the scope of the direct representation of the collision risk in the autonomous object control algorithm exists. In this study, intelligent methods for autonomous control engineering of maritime objects were developed to assess the risk of collision in multi-object situations. These methods included crisis management before collisions to determine the optimal and safe ship trajectories. The value of the collision risk was determined from the object domains generated by the neural network or its three appropriate mathematical models. The basis for these considerations was neural dynamic control and risk game control. Simulations of the control algorithms performed on examples of real navigation situations under different environmental conditions of object movement enabled the assessment of their effectiveness in safe control. The most effective algorithm in good traffic conditions was the neural dynamic control algorithm, whereas the risk game control algorithm enabled effective noncooperative and cooperative control in restricted traffic conditions of autonomous objects. The results from this study provide greater awareness that the use of intelligent control methods can improve the safety of maritime navigation.