Enhancing Spectral Efficiency of 6G Downlink Beamforming via Cooperative Multi-Agent Deep Reinforcement Learning

Area of Science:

• Wireless Communication
• Artificial Intelligence
• Signal Processing

Background:

• Multi-User Multiple-Input Multiple-Output (MU-MIMO) systems are crucial for modern wireless networks.
• Efficient beamforming is essential for maximizing spectral efficiency and user experience.
• Existing methods face challenges in dynamic environments and complex interference scenarios.

Purpose of the Study:

• To develop and evaluate a novel beamforming algorithm for MU-MIMO systems.
• To leverage Multi-Agent Reinforcement Learning (MARL) for enhanced system performance.
• To improve Signal-to-Interference-Plus-Noise Ratio (SINR) and network throughput.

Main Methods:

• Implementation of a beamforming algorithm utilizing MARL.
• System configuration with two base stations, each employing Uniform Rectangular Array (URA) antennas.
• RL algorithms at each base station coordinate to optimize SINR and minimize interference.

Main Results:

• The proposed MARL beamforming algorithm demonstrated significant performance improvements.
• Achieved over a 2-fold increase in network throughput.
• Reported a substantial 5453% improvement in Signal-to-Interference-Plus-Noise Ratio (SINR).

Conclusions:

• The MARL-based beamforming algorithm effectively enhances MU-MIMO system performance.
• The dynamic adaptation of beam patterns maintains high SINR and improves resource utilization.
• This approach shows strong potential for future high-performance wireless communication systems.