Enhancing WMMSE With EVM Feedback Through Bayesian Optimization

This paper presents an enhancement to the classical weighted minimum mean square error (WMMSE) approach which assumes ideal Gaussian-distributed transmitted symbols, addressing its performance decline for multi-user precoding in practice. By incorporating scaling coefficients that bridge the gap between theoretical communication rates and actual spectral efficiency based on the modulation and coding schemes (MCS) of each user, the proposed method maximizes real communication rates by optimizing precoding vectors and estimating scaling coefficients jointly. With fixed scaling coefficients, we first design an efficient algorithm for optimal precoders, which is surprisingly almost the same as the pure WMMSE. Due to the implicitness of the scaling coefficients, we then estimate these coefficients by minimizing the error vector magnitude (EVM) of all users, which is solved by Bayesian optimization due to the black box nature of the mapping between these coefficients and EVM. Simulation results demonstrate the superiority of the proposed method over existing benchmarks.