On the Performance of Downlink Channel Estimation-Based Cell-Free Massive MIMO Systems Under Correlated Rician Fading Channels

The prevailing research on cell-free massive multiple-input multiple-output (MIMO) systems posits that, due to channel reciprocity and channel hardening properties, user equipment (UEs) can decode downlink data solely based on uplink statistical channel state information (CSI). However, the phenomenon of channel hardening is not pronounced in cell-free massive MIMO systems, leading to unsatisfactory spectral efficiency (SE) performance with the aforementioned decoding approach. To address this limitation, this paper examines a downlink channel estimation (DL-CE) strategy for cell-free massive MIMO systems. We present a closed-form expression for the downlink achievable SE, considering a correlated Rician fading channel and imperfect CSI, enabling quantitative assessment of SE performance across diverse system configurations, such as varying numbers of access points, UEs, and antennas per AP. Subsequently, we formulate a joint optimization problem with respect to both uplink and downlink pilots to maximize the sum-SE. Given the intricate and nonconvex nature of the problem, we propose a genetic algorithm (GA)-based pilot assignment strategy to facilitate an effective resolution. Simulation outcomes indicate that the implementation of DL-CE markedly enhances SE performance relative to cell-free massive MIMO systems reliant on uplink channel estimation (UL-CE) alone. Moreover, the proposed joint pilot assignment scheme yields significant SE improvements over benchmark pilot assignment strategies.