Channel Estimation for Downlink Massive MIMO Exploiting UE-side Channel Information Database

Sparse Bayesian learning (SBL) is a promising algorithm for downlink multi-user massive multiple-input multiple-output (MIMO) channel estimation (CE). However, existing SBL-based CE methods typically suffer from two critical limitations: excessive pilot overhead and slow convergence rates. To address these challenges, we propose an environment-aware SBL framework, leveraging the user equipment (UE) location information, and the historical channel information queried from a UE-side channel information database (UECID). Specifically, our algorithm incorporates spatial prior information such as location-dependent multipath angle-of-departure (AoD) into the conventional SBL framework, achieving dual benefits: significantly accelerated convergence speed and substantially reduced pilot overhead.