BeiDou Short-Message Satellite Resource Allocation Algorithm Based on Deep Reinforcement Learning.

Area of Science:

• Satellite Communication Systems
• Resource Allocation Algorithms
• Deep Reinforcement Learning

Background:

• The BeiDou-3 short-message communication system (SMSCS) faces challenges with scarce processing resources on short-message satellites.
• Efficient allocation and scheduling of these resources are crucial for maintaining adequate service quality and system efficiency, especially in multi-satellite coverage areas.

Purpose of the Study:

• To propose a novel deep reinforcement learning-based algorithm for short-message satellite resource allocation (DRL-SRA).
• To optimize resource utilization, minimize transmission path loss for terminals, and ensure satellite load balancing and quality of service.

Main Methods:

• Developed a multi-objective joint optimization satellite resource allocation model tailored for SMSCS characteristics.
• Implemented a region division strategy and feature extraction network to reduce input data dimensionality.
• Utilized a deep reinforcement learning algorithm within the deep deterministic policy gradient (DDPG) framework for continuous spatial state parameterization.

Main Results:

• The DRL-SRA algorithm effectively reduces the transmission path loss for short-message terminals.
• Demonstrated significant improvements in the quality of service provided by the satellite system.
• Achieved increased resource utilization efficiency for the short-message satellite system while maintaining load balance.

Conclusions:

• The proposed DRL-SRA algorithm offers an effective solution for optimizing resource allocation in satellite communication systems.
• The approach successfully balances competing objectives of minimizing loss, maximizing service quality, and ensuring efficient resource utilization.