Unveiling New Frontiers of Downlink Training in User-Centric Cell-Free Massive MIMO

Abstract

Cell-free massive MIMO (CF-mMIMO) emerges as a pivotal technology in the landscape of beyond-5G and 6G wireless networks, addressing the ever-increasing demand for seamless connectivity and unprecedented data throughput. This paper undertakes a comprehensive exploration of scalable usercentric (UC) CF-mMIMO systems, focusing on critical aspects of downlink (DL) channel state information (CSI) acquisition and its intricate interactions with both distributed and centralized precoding strategies. The paper delves into the crucial role of DL CSI acquisition, particularly in scenarios of weak channel hardening arising from sparse subsets of access points (APs) serving specific mobile stations (MS) in UC strategies, and transmission over spatially correlated multiple keyhole Ricean fading channels. The main contributions of this research work include in-depth analyses of different detection schemes under varying precoding scenarios, offering valuable insights for practical deployment. The pivotal role of DL CSI acquisition in optimizing the performance of UC CF-mMIMO networks is fully assessed, dismissing the use of DL pilot-based detection approaches and advocating for either centralized precoding architectures with statistical CSI-based decoding strategies at the MSs or distributed precoding schemes with DL blind channel estimation-based decoders at the MSs.