Deep Learning for Joint Channel Estimation and Feedback in Massive MIMO Systems

TL;DR

The deep learning based joint channel estimation and feedback framework is proposed, which comprehensively realizes the estimation, compression, and reconstruction of downlink channels in FDD massive MIMO systems.

Abstract

The great potentials of massive Multiple-Input Multiple-Output (MIMO) in Frequency Division Duplex (FDD) mode can be fully exploited when the downlink Channel State Information (CSI) is available at base stations. However, the accurate CSI is difficult to obtain due to the large amount of feedback overhead caused by massive antennas. In this paper, we propose a deep learning based joint channel estimation and feedback framework, which comprehensively realizes the estimation, compression, and reconstruction of downlink channels in FDD massive MIMO systems. Two networks are constructed to perform estimation and feedback explicitly and implicitly. The explicit network adopts a multi-Signal-to-Noise-Ratios (SNRs) technique to obtain a single trained channel estimation subnet that works well with different SNRs and employs a deep residual network to reconstruct the channels, while the implicit network directly compresses pilots and sends them back to reduce network parameters. Quantization module is also designed to generate data-bearing bitstreams. Simulation results show that the two proposed networks exhibit excellent performance of reconstruction and are robust to different environments and quantization errors.

Figures (6)

• Figure 1: Illustration of CEFnet, including CE subnet and CF subnet with an encoder and a decoder. The CE subnet and encoder at the UE obtains CSI and compresses them to codewords which are then quantized into bitstreams and sent back through the uplink. The decoder at the BS recovers CSI from feedback bitstreams.
• Figure 2: The three-layer CE subnet, with each layer adopting a different-sized filter to perform the specific operation as illustrated.
• Figure 3: The CF subnet, with the FC layer at the UE compressing the estimated CSI and that at the BS decompressing it. The refinement block based on deep residual network refines it to obtain the final recovered CSI.
• Figure 4: Illustration of PFnet, including an encoder and an decoder. The encoder at the UE compresses input pilots to codewords and quantizes them into bitstream and sent back through uplink. The decoder at the BS recovers CSI from feedback bitstreams.
• Figure 5: The learning curve of two specific scenarios based on grid search. (a)The indoor scenario with $CR$=4 and $P$=16. (b)The outdoor scenario with $CR$=4 and $P$=32.