Beamforming for Fully Connected Dynamic Metasurface Antennas

The dynamic metasurface antenna (DMA), with its unique characteristics of subwavelength spacing and real-time reconfigurability of elements, is considered a potential technique for future wireless communication systems. With these characteristics, DMAs can offer a high-capacity, low-cost, and energy-efficient solution for high spectral efficiency demands. Despite these advantages, in conventional DMA architectures, the number of microstrips increases with the number of radio frequency (RF) chains. This paper presents a novel fully connected DMA-based architecture for reducing the number of RF chains in downlink multiple-input single-output systems. In this architecture, we combine the signals from the RF chains (fewer than the number of microstrips) and map the resulting composite signals to the input port of each microstrip via an analog precoding unit. We develop a gradient projection-manifold optimization algorithm to maximize the signal-to-noise ratio (SNR) for the single-user (SU) scenario and propose an alternating optimization algorithm to maximize the weighted sum rate (WSR) of the system for the multi-user (MU) scenario. The simulation results show comparable WSR and SNR performances of the existing and proposed architectures, whereas the proposed architecture achieves higher energy efficiency in both the SU and MU scenarios.