LMMSE Processing for Cell-free Massive MIMO with Radio Stripes and MRC Fronthaul

Cell-free massive multi-input multi-output (MIMO) provides ubiquitous connectivity for multiple users. Compared with co-located massive MIMO, the major cost includes fronthaul and access points (AP). This paper assumes an implementation using radio stripes. Maximum ratio combining (MRC) uses a low-loading fronthaul and low-cost APs, but the performance is bad. To solve these problems, this paper first proposes an SVD-based method named quasi-LMMSE (Q-LMMSE). Q-LMMSE is derived from standard linear minimum- mean-square-error (LMMSE) using only the MRC frounthaul signals. Employing the simple MRC fronthaul design, a novel processing method is derived from standard linear minimum- mean-square-error (LMMSE). Simulation results show that the proposed Q-LMMSE gains a higher SE with low fronthaul loading, AP complexity, and latency than existing distributed schemes and performs even better than centralized LMMSE in some cases.