Deep Learning Based Beam Training for Extremely Large-Scale Massive MIMO in Near-Field Domain

Abstract

Extremely large-scale multiple-input multipleoutput (XL-MIMO) systems are capable of improving spectral efficiency by employing far more antennas than conventional massive MIMO at the base station (BS). However, beam training in multiuser XL-MIMO systems is challenging. Firstly, new near-field channel models and near-field XL-MIMO transmit beamforming (TBF) codebooks have to be adopted due to the dramatic increase in the number of antennas, which results in an excessive pilot overhead for beam training. Secondly, when the user density is high, the wireless propagation environments of the adjacent users are similar and hence the pilot signals received by the BS from different users appear to be interrelated, which is potentially beneficial but difficult to exploit. Thirdly, different users might share the same beam-direction, which causes excessive inter-user interference. To tackle these issues, we conceive a three-phase graph neural network (GNN)-based beam training scheme for multiuser XL-MIMO systems. In the first phase, only far-field wide beams have to be tested for each user and the GNN is utilized to map the beamforming gain information of the far-field wide beams to the optimal near-field beam for each user. In addition, the proposed GNN-based scheme can exploit the position-correlation between adjacent users for further improvement of the accuracy of beam training. In the second phase, a beam allocation scheme based on the probability vectors produced at the outputs of GNNs is proposed to address the above beam-direction conflicts between users. In the third phase, the hybrid TBF is designed for further reducing the inter-user interference. Our simulation results show that the proposed scheme improves the beam training performance of the benchmarks based on traditional neural networks. Hence it is more suitable for multiuser XL-MIMO systems. Moreover, the performance of the proposed beam training scheme approaches that of an exhaustive search, despite requiring only about 7% of the pilot overhead.

Holographic MIMO Communications: Theoretical Foundations, Enabling Technologies, and Future Directions

“…An OMP cascaded convolutional autoencoder neural network was developed for hybrid near-and far-field channel estimation. [263] Uplink ULA NLOS (Near-field)…”

“…Indicatively, in [263], an XL-MIMO system was considered where both the transmitter and the receiver only have one RF chain, a deep learning framework was proposed for localization in the near-field region. The authors assumed that a predefined codebook is applied for beamforming, searching for the optimal codeword to align the beam to the LOS path and thus achieve the largest data rate.…”

Cross Far- and Near-field Wireless Communications in Terahertz Ultra-large Antenna Array Systems

“…For comparison, only angle information is considered in PWM-based designs. The additional distance domain information results in high complexity of channel estimation and beam training [8], [11].…”

“…In the far-field, the beams only process angular resolution, beam training usually searches the angular domain to find the best beam pair [10]. By contrast, beam training has to jointly search in the angular and distance domains, due to additional distance domain beam resolution [11]. As the searching beams are obtained from a pre-defined beam codebook, to meet the angular and joint angular and distance resolutions for far-and near-field, respectively, the codebook designs are different, both of which are not universally applicable for cross-field beam training.…”

“…Although UM-MIMO and IS systems have been recently exploited, such as channel modeling and analysis [6], [7], channel estimation [8], [9], beam training [10], [11], beamforming design [3] and multiple access [12], among other topics, the separate designs in either far-field or near-field are invalid as unified solutions of the cross-field communications. On one hand, the planar-wave model (PWM) where the wave propagation is approximated as a plane is adopted in the farfield.…”