Unmanned surface vehicle (USV) swarms are increasingly utilized for complex maritime tasks, such as continuous environmental monitoring, island patrols, and strategic surveillance. However, existing methods still face significant challenges in online autonomous task allocation, dynamic replanning, and real-time adaptation to unexpected events in open and uncertain environments. This study aims to improve the decision-making autonomy, robustness, and task continuity of USV swarms during long-duration area coverage operations. To achieve this, we propose a real-time decision-making and cooperative control framework powered by large language models (LLMs).

The proposed method introduces a hierarchical decomposition framework that integrates semantic interpretation with structured task planning. By incorporating LLMs, natural language mission instructions are dynamically translated into executable objectives, enabling USV swarms to autonomously generate efficient, conflict-free navigation paths. A hybrid architecture is adopted, where a centralized planner handles global task allocation, while distributed planners on individual USVs compute local collision-free trajectories in parallel. To manage uncertainty, a dual closed-loop feedback mechanism is implemented: (i) at the task level, dynamic prompt generation enables reallocation when resources change or failures occur; (ii) at the action level, incremental trajectory correction through geometric collision detection ensures safe adaptation without the need for full replanning. This multilayered structure enhances scalability, responsiveness, and resilience for long-term multi-agent operations.

The proposed framework was validated in a simulated maritime island patrol scenario, where multiple USVs collaboratively encircled randomly generated islands and returned to base. Comparative experiments with classical methods, such as A*, genetic algorithms, and ant colony optimization, demonstrate that our approach achieves comparable efficiency in terms of path length and travel time, while uniquely ensuring strict adherence to natural language mission goals. Cross-model evaluation with phi-4, Qwen-3, and DeepSeek R1 further shows that lightweight, reasoning-oriented models strike the optimal balance between planning accuracy and computational efficiency. Experimental results indicate that the framework achieves an average task success rate of over 80% in complex, long-term, multi-task scenarios, while maintaining a success rate above 70% for real-time responses and handling of abnormal events (e.g., random USV failures). Ablation studies confirm that hierarchical task decomposition and multi-level feedback significantly contribute to the overall performance improvement.

This study demonstrates that LLMs can effectively support decision-making for USV swarms in dynamic maritime environments. By integrating semantic instruction understanding, hybrid centralized–distributed planning, and a dual closed-loop feedback mechanism, the proposed framework ensures task continuity, adaptability, and safety. These findings present a promising approach for deploying USV swarms in reliable, long-endurance missions and provide a foundation for extending this method to large-scale, real-world multi-agent systems.

To address the limitations of current health state estimation methods for naval vessel energy equipment, such as insufficient cross-device generalization ability and high model adaptation complexity, this study proposes a new energy equipment health state estimation method based on industrial large models. The aim is to quickly develop dedicated estimation models using a unified framework and parameter-efficient fine-tuning strategies, thereby achieving high-precision health state estimation.

An industrial large model is constructed using a pre-trained Transformer architecture, incorporating attention mechanisms and low-rank adaptation (LoRA) techniques. This approach dynamically adjusts key parameters to meet the needs of different tasks. It incorporates a sliding window attention mechanism to capture local dynamic characteristics, a conditional attention mechanism to account for environmental influences, and low-rank adaptation fine-tuning modules to rapidly adapt to new equipment tasks.

Verification experiments were conducted on two typical types of new energy equipment: lithium batteries and fuel cells. The results showed that in the lithium battery dataset, the mean absolute error was no greater than 0.0216, and the minimum determination coefficient was no lower than 0.971 3. In the fuel cell dataset, the mean absolute error was no greater than 0.003 3, and the minimum determination coefficient was no lower than 0.9197. The model demonstrated excellent estimation accuracy and generalization ability.

The method based on industrial large models effectively improves the accuracy and reliability of health state estimation for new energy equipment, offering a novel technical approach for equipment health management. Future work can focus on further enhancing model performance through model optimization, cross-device adaptation, and data processing improvements.

With the continuous advancement of high-energy weapon technology, energy storage systems are becoming increasingly crucial for maintaining stable energy supply on naval platforms. Therefore, this paper simulates thermal runaway in lithium battery energy storage systems on naval platforms.

First, using Fluent and Simulink software, simulation models for submerged and liquid-cooled plate energy storage systems are constructed. Based on the actual working conditions of a specific vessel, thermal runaway simulation experiments were conducted.

The results show that, compared to the liquid-cooled plate method, the submerged cooling method exhibits superior performance in managing thermal runaway, is better at controlling temperature fluctuations, and maintains system stability, which makes it more promising for naval platform applications.

This research provides reference for the safety control technologies of energy storage systems used on naval platforms.

The standardization of ship engine room operations is a critical component of ship safety management. Therefore, the practical examination for crew members includes the disassembly and assembly of ship equipment as a key assessment item. To enhance the digitalization and intelligence of crew practical examinations, a computer vision-based automated recognition method for assessing the standardization of ship equipment disassembly and assembly processes is proposed.

First, the backbone network of the ship equipment detection model is constructed using YOLOv8n, and the shuffle-attention (SA) mechanism is introduced to improve the model's feature extraction capability and training efficiency. Subsequently, a reparameterized generalized feature pyramid network (GFPN) fusion structure is incorporated into the neck network to enhance the model's ability to fuse multi-scale features. Finally, the original CIoU loss function is replaced with the wise intersection over union (WIoU) loss function to improve the model's accuracy.

Experimental results on a self-constructed dataset demonstrate that, compared to YOLOv8n, the improved object detection algorithm achieves a 0.15 increase in mean average precision and a 0.6 frames-per-second improvement in real-time detection, enabling accurate recognition of the disassembly and assembly processes of gear pumps.

The improved algorithm exhibits superior recognition capabilities and is better suited for identifying the standardization of ship equipment disassembly and assembly processes.

Artificial intelligent technologies have become an important approach to improving the safety of shipping and reducing the operating costs of shipping companies. In order to further improve the level of ship intelligence and break down the data barriers between different shipping companies, an efficient privacy-preserving federated learning method (EPFL) is proposed in this paper.

Federated learning is adopted to organize multiple ship participants to collaboratively train a global fault diagnosis model, and cryptography technologies are used to protect their local data information. Considering Internet of Ships (IoS) scenarios, this paper introduces sparsification technology to compress the model parameters uploaded by shipping participants and reduce their number.

Theoretical analysis and the experimental results show that the proposed EPFL method can effectively reduce the resource consumption of cryptographic computation and data communication while protecting the local data information of ship participants.

The proposed EPFL method can provide references for the establishment of intelligent ship systems.