Multi-Vehicular Beam Space Learning for Channel Estimation in 6G V2X Systems

Channel knowledge is a fundamental requirement for accurate analog-digital beamforming in hybrid MIMO systems. Conventional channel estimation techniques, e.g., unconstrained maximum likelihood, cannot be used to directly estimate the full MIMO channel in hybrid systems, as the analog beamforming limits the channel observation to the digital side only. We propose a two-stage channel estimation procedure for 6G Vehicle-to-Everything (V2X) systems, leveraging multiple recurrent vehicle passages over the same location, properly clustered in space, in both the analog beam training (stage 1) and the algebraic digital channel estimation (stage 2). We show that the proposed method practically matches the perfect channel knowledge performance on a wide range of Signal-to-Noise Ratio (SNR) values, discussing its behavior with respect to the cluster size.