Adaptive Spatio-Temporal Voxel-Based Trajectory Planning and Optimization for Close-Quarters Ships Collision Avoidance

Close-quarters encounter scenarios, characterized by their potentially severe consequences, represent one of the most formidable challenges in ship collision avoidance decision-making and trajectory planning. An efficient collision avoidance framework must simultaneously satisfy critical requirements for real-time decision-making, accurate trajectory planning, and effective vessel maneuvering, all while maintaining rigorous compliance with the International Regulations for Preventing Collisions at Sea (COLREGs). This framework serves as the critical safeguard for the navigation of maritime autonomous surface ships. To design an efficient framework, this study proposes an innovative method for real-time partitioning of collision-free navigable space, termed “adaptive spatio-temporal voxels”, which accounts for the ship’s maneuverability by incorporating the velocity obstacle concept. Furthermore, a spatio-temporal graph, incorporating COLREGs compliance, yields an optimal collision avoidance decision represented as a sequence of voxels. These voxel sequences subsequently serve as constraints within a model predictive control (MPC) framework to optimize and plan a safe, navigable trajectory. This framework has undergone rigorous testing in a variety of close-quarters encounter scenarios. Its computational performance, with both decision-making and trajectory optimization completed within 1s, effectively meets the demands of real-time maritime collision avoidance. The results demonstrate its ability to generate efficient collision avoidance trajectories in complex environments, significantly enhancing overall collision avoidance performance.

IEEE Transactions on Intelligent Transportation Systems

IEEE Transactions on Intelligent Transportation Systems

26

1558-0016

N/A

12

16