Mutual Information of Dynamic Metasurface Antenna-Enabled MIMO With Finite-Alphabet Inputs

Dynamic metasurface antennas (DMAs) are promising hardware architectures for supporting massive multiple-input multiple-output (MIMO) transmission. This paper investigates DMA-assisted MIMO systems with practical finite-alphabet signaling and proposes efficient precoding methods to maximize the mutual information (MI). For the single-stream case, an effective precoder approximation (EPA)-based algorithm is developed to jointly optimize the DMA coefficients and the digital precoder. The study is then extended to the more challenging multi-stream transmission scenario. Due to the intractable MI expression in multi-stream systems with finite-alphabet inputs, the cut-off rate (CR) is employed as a surrogate objective, and a penalty dual decomposition (PDD)-based optimization framework is proposed. To further reduce computational complexity, two low-complexity designs are proposed based on the low signal-to-noise ratio (SNR) asymptotic CR and the Gaussian signaling assumption, respectively. Numerical results demonstrate that the proposed algorithm in the single-stream transmission significantly outperforms benchmark methods, while the proposed multi-stream designs effectively operate across a wide range of regimes, including low- and high-SNR regimes as well as different numbers of data streams.