Analysing the Influence of Infrastructure and Power Control on Cellular and Cell-Free Massive MIMO Systems: Insights from Machine Learning

Abstract In large-scale multiple-input-multiple-output (mMIMO) networks, effective power control (PC) mechanisms play a pivotal role. Various algorithms, such as the weighted mean square error (WMMSE) algorithm, are utilized to estimate PCs, demanding considerable computational resources. This study examines the performance of PC in mMIMO systems, emphasizing the aggregate spectral efficiency (sum SE) and the cumulative distribution function (CDF) constrained by SE per user equipment (UE). This investigation explores the impact of different factors, including the number of UEs, access points/base stations (APs/BSs), and the implementation of deep neural network (DNN)-based PCs, within both cellular (CL) and cell-free (CF) architectures. Through empirical analysis, the study elucidates the influence of parameter 'g' on the DNN versus WMMSE comparison curve, underscoring the importance of accounting for the quantity of APs/BSs and antennas to attain optimal PC performance.