Overview of current algorithms for autonomous vessels route optimization

An overview of multiple current algorithms for optimizing the routes of autonomous surface vessels is provided in the paper. One of the effective methods for route optimization is the implementation of algorithms and software based on graph theory to prevent collisions. Key algorithms include Dijkstra's algorithm, A*, artificial potential fields, the "dynamic window" method, and the velocity obstacle method. Collision prevention using radar of the maritime autonomous navigation vessel, geometric vessel factors, genetic algorithms, neural network training is also separately discussed. While most algorithms are only considered theoretically, some works describe practical observations: neural networks using deep learning, Markov decision processes, Q-learning; developed autonomous collision avoidance system; heuristic search for optimal ship routes using the A2015 algorithm. Overall research analysis shows that many authors have made significant progress in their work, but the topic is not fully explored. Some works do not consider divergence with multiple vessels, while others do not utilize maneuvers involving speed changes. Certain works face challenges in parameter tuning for algorithm efficiency. The optimality criterion for multiple maneuvering, considering factors beyond minimizing the closest point of approach to another vessel, such as responsibility distribution for maneuver execution, is not fully developed. It is noted that all necessary conditions for creating a universal algorithm in the future already exist due to modern technologies and the research topic relevance.