An Enhanced AI-driven Beam Selection Method for D-MIMO

Beam selection is crucial for determining the optimal beam and transmission and reception points (TRPs) to serve users efficiently in Distributed Multiple-Input Multiple-Output (D-MIMO) networks. Measuring the downlink channel for all conceivable TRP-beam pairings to find the best TRP and beam pair for the user can be resource-intensive, particularly in the millimeter wave spectrum. Nowadays, Artificial Intelligence (AI) techniques are being explored to identify the optimal TRP and beam by sounding the channel for only a subset of all TRPs and beams. In this study, we propose an enhanced AI-driven beam selection method by incorporating the eigenvectors and autocorrelation matrix of received signal to the input of the AI model in addition to reference signal received power values to provide a better beam selection mechanism. By obtaining the simulation results, we show that our method provides better performance results compared to baseline methods.