Villarreal L. , PERALTA SAMANIEGO, FEDERICO, MILLÁN GATA, PABLO, BEJARANO PELLICER, GUILLERMO
No
Smart Water Quality Monitoring
Capítulo de un Libro
Científica
0
0
Internacional
01/01/2026
2-s2.0-105030217564
Designing controllers can be highly dependent on the knowledge of the underlying model, especially in aquatic environments where the hydrodynamics parameters of the vehicles are almost always unknown. This work leverages this difficulty by using a machine learning-based guidance subsystem that produces valid control outputs, not only for one unmanned surface vessel but also for a group of them so that they can efficiently move while maintaining a formation. The formation control network was trained using proximal policy optimization (PPO) and through a simulator considering the well-known Cybership-II ship. Results show that the actor-critic deep reinforcement learning network produces an efficient policy to control a fleet of unmanned surface vehicles (USVs) and maintains formation whilst moving towards a goal. © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.
Deep learning; Deep reinforcement learning; Distributed computer systems; Fleet operations; Intelligent systems; Optimization; Ships; Unmanned surface vehicles; Aquatic environments; Autonomous surface vehicles; Fleet coordination; Formation control; Hydrodynamic parameters; Policy optimization; Reinforcement learnings; Surface vehicles; Underactuated; Vehicle formations; Reinforcement learning