Precoding Design of Reconfigurable Massive MIMO in the Electromagnetic Domain

Reconfigurable massive multiple-input multiple-output (RmMIMO) is expected to unlock the unexplored degrees of freedom in the electromagnetic (EM) domain, offering more flexibility to the design of communication systems. Specifically, optimizing the delicate EM properties such as radiation pattern depends on detailed domain knowledge. Traditional spatial domain channel state information (sCSI) has obscured the underlying EM domain information of the channel, thereby hindering the base station's efforts in exploring the EM properties of the channel for precoding optimization. This paper exploits the detailed EM domain channel state information (eCSI) for radiation pattern design at the base station. Using the orthogonal decomposition method on the spherical harmonics bases, the radiation pattern is decomposed into the linear combination of these orthogonal bases. We then formulate the pattern design problem as a problem of optimizing projection coefficients on these bases. Specifically, we propose a manifold optimization-based method that can alternatively optimize the radiation pattern design and the digital precoder. Simulation results demonstrate that, in comparison to traditional mMIMO systems with fixed antenna radiation patterns, RmMIMO architecture induces significant throughput gain for downlink transmission.