Deep Reinforcement Learning-Based Resource Management in Maritime Communication Systems.

Area of Science:

• Telecommunications Engineering
• Computer Science
• Maritime Technology

Background:

• The expanding maritime economy necessitates high-quality communication for maritime users.
• Current nearshore communication systems require dynamic resource allocation to improve data rates.
• Ensuring reliable communication is crucial for maritime operations and safety.

Purpose of the Study:

• To propose a novel beam allocation scheme using deep reinforcement learning (DRL) to maximize communication rates for maritime users.
• To address the complexity of resource management in maritime environments by discretizing the space.
• To ensure service quality for all areas, including those with poor channel conditions.

Main Methods:

• A virtual queue-based deep reinforcement learning (DRL) algorithm is developed for beam allocation.
• The maritime environment is discretized using a grid-based method to simplify resource management.
• The nearshore base station acts as a learning agent, optimizing beam allocation through interaction.
• A virtual queue method is implemented to service grids with poor channel state information.

Main Results:

• The proposed DRL beam allocation scheme effectively increases the communication rate for maritime users.
• The grid-based discretization reduces the complexity of the combinatorial optimization problem.
• The virtual queue mechanism ensures equitable service distribution across different maritime zones.
• Simulation results validate the performance improvements offered by the proposed scheme.

Conclusions:

• The virtual queue-based DRL beam allocation scheme significantly enhances maritime communication rates.
• The approach offers an effective solution for dynamic resource management in complex maritime environments.
• This method provides a foundation for future advancements in maritime communication quality and efficiency.