Deep Reinforcement Learning-Based Adaptive Beamforming for Ultra-Reliable 6G Wireless Communication

Abstract

One of the most important features of the sixth-generation (6G) wireless systems is ultra-reliable and low-latency communication (URLLC), which allows supporting mission-critical applications like autonomous transportation, smart healthcare, and industrial automation. Nevertheless, traditional beamforming methods, such as Zero-Forcing (ZF) and Minimum Mean Square Error (MMSE) are highly dependent on the accuracy of channel state information (CSI) and are not very flexible in extremely dynamic wireless scenarios. To solve these issues, the present paper suggests a deep reinforcement learning (DRL)-based adaptive beamforming model to optimize beamforming vectors in real-time, under the conditions of uncertain and time-varying channel conditions. It is expressed as a Markov decision process, with a Deep Q-Network (DQN) to optimize the signal-to-interference-plus-noise ratio (SINR) by increasing the signal and reducing the interference and latency. A multi-user massive MIMO system in the millimeter-wave/terahertz band is used to assess the proposed approach. Simulation findings confirm that the DRL-based approach can reach up to 25% spectral efficiency gains and reliability improvement to 99.999% in comparison with the traditional beamforming algorithm and halves latency by a factor of about 30. These findings confirm the usefulness of the suggested model to support intelligent, flexible, and ultra-reliable communication in 6G wireless networks.