Configuration Optimization and Channel Estimation in Hybrid Beamforming mmWave Systems With Channel Support Side Information

Abstract

Hybrid beamforming (HBF) is a promising technique in millimeter-wave (mmWave) systems with a limited number of radio frequency (RF) chains. Acquiring channel state information is essential for HBF design. Various compressive-sensing-based channel estimation (CE) schemes that exploit the sparse nature of mmWave channels have been proposed. The design of an analog beamforming matrix plays a key role in providing high CE quality with low pilot overhead. In this paper, a novel and flexible HBF architecture called adaptive sum product network (ASPN) is proposed, which can realize more flexible analog beamforming with less hardwares. Based on the ASPN architecture, a novel variational-Bayesian-inference-based compressive CE algorithm is proposed, which can fully exploit the channel support side information (CSSI) at the base station under a general uncertain measurement matrix induced by the off-grid parameters for power leakage elimination. To optimally configure the degrees of freedom in the ASPN according to the CSSI, a dynamic configuration optimization algorithm is proposed by formulating the CE performance metric in terms of the configuration parameters of ASPN. The importance and effectiveness of the flexibility introduced in ASPN HBF architecture, the compressive CE algorithm as well as the configuration algorithm are illustrated by the superb performance of the proposed scheme compared to the representative state-of-the-art techniques through extensive simulation results.