Deep Learning Approach to Channel Sensing and Hybrid Precoding for TDD Massive MIMO Systems

Abstract

This paper proposes a deep learning approach to channel sensing and downlink hybrid analog and digital beamforming for massive multiple-input multiple-output systems with a limited number of radio-frequency chains operating in the time-division duplex mode at millimeter frequency. The conventional downlink precoding design hinges on the two-step process of first estimating the high-dimensional channel based on the uplink pilots received through the channel sensing matrices, then designing the precoding matrices based on the estimated channel. This two-step process is, however, not necessarily optimal, especially when the pilot length is short. This paper shows that by designing the analog sensing and the downlink precoding matrices directly from the received pilots without the intermediate channel estimation step, the overall system performance can be significantly improved. Specifically, we propose a channel sensing and hybrid precoding methodology that divides the pilot phase into an analog and a digital training phase. A deep neural network is utilized in the first phase to design the uplink channel sensing and the downlink analog beamformer. Subsequently, we fix the analog beamformers and design the digital precoder based on the equivalent low-dimensional channel.A key feature of the proposed deep learning architecture is that it decomposes into parallel independent single-user DNNs so that the overall design is generalizable to systems with an arbitrary number of users. Numerical comparisons reveal that the proposed methodology requires significantly less training overhead than the channel recovery based counterparts, and can approach the performance of systems with full channel state information with relatively few pilots.

Model-Driven Learning for Generic MIMO Downlink Beamforming With Uplink Channel Information

“…We assume the downlink channel attenuation 𝛼 𝑑,𝑙 of the 𝑙-th path follows an independent Rayleigh distribution with zero mean and unit variance. Given the ULA channel model in (18), the nonlinear relation between the uplink channel h 𝐷 and dowlink channel h 𝐷 are characterized by:…”

“…A deep neural network using unsupervised training was proposed arXiv:2109.07819v1 [cs.IT] 16 Sep 2021 in [17] to map the received uplink pilots to the beamforming matrix at the BS for the intelligent reflecting surface assuming uplink-downlink channel reciprocity. A channel sensing and hybrid precoding design was proposed in [18], by using the received pilots without the intermediate channel estimation step for TDD massive MIMO systems.…”

Data-Driven Deep Learning Based Hybrid Beamforming for Aerial Massive MIMO-OFDM Systems with Implicit CSI

“…Specifically, the authors of [45] jointly modeled the pilot design, channel feedback, and digital beamforming as an end-to-end (E2E) neural network for narrowband FDD fully-digital MIMO systems. As for DL-based hybrid beamforming, the authors of [46], [47] proposed joint channel sensing and hybrid beamforming schemes based on DL for narrowband TDD MIMO systems, in which [47] considers the quantization of PSs' phase values. Furthermore, the authors of [48] proposed a DL scheme based on quantized received signal strength indicators to design hybrid beamforming for narrowband FDD MIMO systems.…”

“…However, these DL-based hybrid beamforming schemes with implicit CSI either only focus on the analog channel sensing and beamforming design in TDD systems without considering the joint design of analog and digital parts [46], [47], or only focus on the hybrid beamforming design in FDD systems without considering the design of pilot and CSI feedback [48], and they all only focus on narrowband MIMO systems. In general, the existing DL-based schemes can efficiently improve system performance through data-driven training, but there is still no unified E2E DL framework for both TDD and FDD multi-user broadband hybrid massive MIMO-OFDM systems.…”

“…Therefore, for aerial hybrid massive MIMO-OFDM systems, we propose to leverage the datadriven deep learning (DL) signal processing paradigm to jointly optimize the pilot design, channel estimation, CSI feedback, and hybrid beamforming for both TDD and FDD systems. On the other hand, since the intractable gradient backpropagation issue would be caused by the quantization constraint of the low-resolution PSs, the aforementioned DL-based analog beamforming schemes [42]- [44], [46], [48] assuming the ideal infinite-resolution PSs in training stage will inevitably suffer from the performance loss in the practical systems with low-resolution PSs. Besides, [47] uses a tanh-like function to approximate the PSs' phase values quantization in training stage, and uses an ideal quantizer in the test stage, which will still lead to the mismatch between training stage and test stage.…”