MIMO Systems with Arbitrary Antenna Array Architectures: Channel Modeling, Capacity and Low-Complexity Signaling

The focus of this work is on multi-antenna systems with arbitrary antenna array architectures. First, we propose a multi-antenna channel model that encompasses many of the existing models and study the capacity of such systems. We show that capacity is achieved by pre-nulling the input signals along the transmit array manifold with a transmit-SNR dependent rank and power control. The rank of the optimal signaling scheme is monotonically non-decreasing as SNR increases. Thus our results generalize many known results in this area. Further, we compute two explicit SNR values, rho_low below which rank-1 signaling and rho_high above which full-rank signaling are near-optimal, respectively. Finally, with a focus on low-complexity scalar signaling schemes, we propose a sub-optimal beamforming approach that minimizes the statistical feedback necessary to implement such schemes. While almost all works in this area assume a genie-aided statistics acquisition process, we show that the proposed scheme trades-off performance with statistical feedback overhead attractively.