Cell-Free Massive MIMO-OFDM Over Spatially Correlated Doubly Selective Channels

This paper investigates the impact of doubly selective fading channels on the design and performance analysis of cell-free massive multiple-input multiple-output (MIMO)-orthogonal frequency division multiplexing (OFDM) networks. Time and frequency selectivity, as well as the intercarrier interference (ICI) effects, are fully characterized, and the spatial channel cross-covariance matrices associated to each mobile station (MS)/access point (AP) pair and for every OFDM symbol/subcarrier conforming a particular resource block (RB) are derived. Assuming that these cross-covariance matrices are known at the APs, optimal minimum mean square error (MMSE)-based and suboptimal projected MMSE (PMMSE)-based double selectivity-aware channel estimators/predictors are designed. Spatial cross-covariance matrices and channel estimates/predictions are then used to design scalable double selectivity-aware partial-MMSE (P-MMSE), averaged P-MMSE (AP-MMSE) and statistically averaged P-MMSE (SAP-MMSE) combiners and precoders to be used during the uplink (UL) and downlink (DL) payload data transmission phases. Extensive numerical simulations are conducted to assess the performance of the proposed double selectivity-aware scalable cell-free massive MIMO systems under scenarios showing different degrees of time and frequency selectivity. Simulation results clearly show the performance advantages of double selectivity-aware channel estimation/prediction schemes and combining/precoding strategies when working in high-mobility multipath scenarios of practical interest. Moreover, it is shown that, even though suboptimal PMMSE/AP-MMSE and PMMSE/SAP-MMSE strategies perform relatively well under poor frequency and time selectivity conditions, they suffer from very significant performance losses with respect to the MMSE/P-MMSE scheme as frequency and/or time selectivity levels increase.