Variational Optimization for Interference Constrained Wireless Signal Estimation

Abstract

Massive multiple-input multiple-output (MIMO) systems demand accurate signal estimation under increasingly severe interference, uncertainty, and imperfect channel state information and interference statistics. Existing variational Bayesian (VB) estimators enhance posterior interference by incorporating signal-recovery optimisation into interference control without explicit intervention. A robust uncertainty-aware variational Bayesian estimator is proposed in this paper for interference-constrained massive MIMO signal estimation. The posterior covariance and posterior-mean energy are incorporated into the interference metric in accordance with both estimated signal power and uncertainty-induced interference exposure. Further, a norm-bounded interference covariance model is introduced to provide robustness when compared with covariance-estimation errors. According to the constraint violations, the interference penalty is altered through the adaptive primal-dual optimisation mechanism during retention of the computational structure of the VB reference. Under limited conditions, the competitive estimation accuracy and detection reliability during the substantial reduction in interference leakage are proven in the simulation results. At 2 dB SNR and 12 dB INR, the normalized interference leakage is reduced by 32.3% compared with LMMSE, whereas the −2 dB SNR reduces 47.2% relative to robust-MMSE. The posterior uncertainty-aware interference regulation is used as an effective extension of conventional VB-based massive MIMO estimation, is established in the results.