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

This paper presents a new beamforming algorithm for Multi-User Multiple-Input Multiple-Output (MU-MIMO) systems using Multi-Agent Reinforcement Learning (MARL). The proposed approach is shown to significantly enhance the efficiency and performance of future wireless communication systems. The system comprises two base stations, each equipped with a Uniform Rectangular Array (URA) of directional antennas. Each base station has RL algorithms that use beamforming to provide the optimal Signal-to-Interference-Plus-Noise Ratio (SINR) for each user. These algorithms also work with the other base stations to prevent user interference and ensure efficient resource use. Simulation results demonstrate that the potential of the proposed method has the potential for dynamically adapting beam patterns and maintaining high SINR across the network, resulting in more than a 2-fold improvement in throughput and a 5453% improvement in SINR.