Machine Learning Based Beam Selection for Maximizing Wireless Network Capacity

Abstract

In today's and future wireless communications, especially in 5G and 6G networks, machine learning (ML) methods are crucial. Potentially, these techniques bring many benefits such as increased data throughput, improved security, reduced latency, and, on the whole, enhanced network efficiency. Furthermore, to facilitate the processing of large amounts of data in real-time situations, machine learning is used for various functions in wireless networks. This article aims to explore the significance and application of machine learning, with a particular focus on classic reinforcement learning, in the context of predicting optimal beam configurations within wireless communications scenarios. Our goal is to minimize interference between transmitters by finding the optimal beamforming angles. For this, ray tracing techniques are deployed. We see this research as a step forward towards integrating digital twin (DT) technology in network management and control. In this article, different machine learning methods are used and their performance is compared. Firstly, the most effective angles for beamforming, maximizing channel capacity are identified. Then, by using these methods and after verifying their accuracy, the optimal antenna angles in scenarios with an increased number of transmitters and receivers is found and evaluated.