Iterative Hybrid Precoding and Combining for Partially-Connected Massive MIMO mmWave Systems

For partially-connected massive multiple-input multiple-output (MIMO) millimeter wave (mmWave) communications, an iterative precoding and combining algorithm is proposed to maximize the system spectrum efficiency. At the transmitter, the optimal unconstrained full digital precoder is firstly determined as a benchmark by singular value decomposition (SVD) of the channel matrix, and then resolved to achieve the hybrid analog and digital precoding in an iterative manner. Specifically, the initially formulated non-convex matrix decomposition problem is decomposed into a series of convex optimization sub-problems, with the amplitude of each element in the analog matrix restricted and the phase increment in adjacent iteration processes varies only in a small range. Moreover, with the unconstrained linear minimum mean square error (MMSE) combiner as the benchmark, similar iterations can be executed at the receiver. Numerical results show that the proposed algorithm acquires superior spectrum efficiency than the other classical scheme with partially-connected architecture. Furthermore, in spite of much lower hardware complexity and power consumption, the performance of our algorithm is almost identical to that of the full digital scheme as well as other mechanisms with fully-connected structure.