Statistical Beamforming for FDD Downlink Massive MIMO via Spatial Information Extraction and Beam Selection

Xiaojun Yuan, Hang Liu, Xiaojun Yuan 0002, Ying Jun ZhangIEEE Trans. Wireless Communications 2020 · DOI

TL;DR

This paper tackles the challenge of efficient downlink beamforming in FDD massive MIMO systems where the base station lacks instantaneous channel state information. They propose a novel method to extract crucial spatial information (angles of departure and large-scale fading) from the uplink channel, enabling effective beamforming without downlink training. This approach significantly reduces overhead and improves performance compared to existing methods.

What Problem Does This Paper Address?

Massive MIMO (Multiple-Input Multiple-Output) systems, with their large antenna arrays, promise significant improvements in wireless communication throughput. However, achieving these gains, especially in downlink beamforming, critically relies on accurate channel state information (CSI) at the base station (BS). In Frequency Division Duplexing (FDD) systems, obtaining this downlink CSI is particularly challenging. Unlike Time Division Duplexing (TDD) systems where uplink and downlink channels are often assumed to be reciprocal, FDD systems operate on different frequency bands, making direct reciprocity unreliable. Traditional methods for acquiring downlink CSI in FDD systems involve extensive pilot transmissions and feedback, leading to substantial training and feedback overhead, which can be infeasible for massive MIMO systems with a large number of antennas. This paper addresses this fundamental challenge by proposing a method to perform effective downlink beamforming without requiring instantaneous CSI at the BS.

Methods

The core of this paper lies in its novel approach to spatial information extraction and subsequent beamforming.

  • Spatial Information Extraction: The authors leverage the concept of "spatial reciprocity," which posits that the directional properties (angles of departure and power angular spectrum) of the uplink and downlink channels are congruent, even if the channels themselves are not identical. They propose a method to extract these spatial parameters from the uplink channel estimation procedure. This is achieved by using a blind channel-and-signal estimation algorithm on the uplink signals, which provides estimates of the angles of departure (AoDs) and large-scale fading coefficients. This bypasses the need for explicit downlink training.
  • Beamforming Design (BS-SBF): Based on the extracted spatial information, the paper proposes a "beam-selection-based statistical beamforming" (BS-SBF) algorithm. Instead of using all available antennas, the BS selects a subset of "angular beams" for each user. This approach exploits the inherent sparsity of the physical channel in the angular domain.
  • Beam-Time Block Coding (BTBC): To address potential phase mismatch issues when multiple beams are assigned to a single user, they introduce a beam-time block coding (BTBC) scheme, which is an angular-domain adaptation of space-time block coding (STBC). This aims to provide diversity gain.
  • Beam Selection Algorithms: To efficiently select the optimal set of angular beams for each user, two low-complexity iterative algorithms are developed:
  • Forward Stepwise (FS) Search: This algorithm iteratively selects beams for users, aiming to maximize the sum-rate while considering interference.
  • Gibbs Sampling: This algorithm is employed to mitigate interference, especially when the number of users is large, by probabilistically updating beam assignments.
  • Sum-Rate Analysis: The paper derives an exact sum-rate expression for the proposed system and provides asymptotic analysis for low and high SNR regimes. Approximations of the sum-rate are also derived to reduce computational complexity for the beam selection algorithms.

Datasets

Not discussed in the paper. The paper focuses on theoretical analysis and simulation-based evaluation using synthetic channel models.

Key Insights and Ideas

  • Spatial Reciprocity as a Game Changer: The paper demonstrates that by exploiting spatial reciprocity, it's possible to infer crucial downlink channel characteristics (AoDs and large-scale fading) from uplink measurements. This is a significant departure from traditional methods that require direct downlink CSI.
  • Angular Domain Sparsity for Efficiency: The core idea of using angular domain sparsity to select a subset of beams for each user is key. This reduces the complexity of beamforming and allows for more efficient resource allocation.
  • Beam-Time Block Coding for Robustness: The introduction of BTBC as an extension of STBC to the angular domain is a clever way to combat phase mismatch issues and gain diversity, especially when multiple beams are assigned to a user.
  • Low-Complexity Beam Selection is Crucial: The development of FS and Gibbs sampling algorithms highlights the practical need for efficient beam selection methods, as exhaustive search becomes computationally intractable for realistic system sizes.
  • Performance Trade-offs with Γ: The analysis of the parameter Γ (number of beams per user) reveals a trade-off between diversity gain and interference. A smaller Γ is preferred for mitigating interference in dense user scenarios, while a larger Γ can offer diversity but might increase interference.

Key Results

  • Significant Performance Improvement: Numerical results show that the proposed BS-SBF algorithm with FS and Gibbs sampling beam selection significantly outperforms state-of-the-art statistical beamforming (SBF) and training-based beamforming algorithms in terms of sum-rate.
  • Robustness to Imperfect Spatial Information: The proposed algorithms show a degree of robustness to angle mismatch, with performance degradation that is manageable, especially when the number of users is reduced to increase angular separation.
  • Low-SNR and High-SNR Behavior: The analysis confirms that the sum-rate scales linearly with SNR in the low-SNR regime and logarithmically with SNR in the high-SNR regime, consistent with theoretical expectations.
  • Computational Efficiency: The proposed beam selection schemes (FS and Gibbs sampling) are shown to be orders of magnitude faster than exhaustive search, making them practical for real-world deployment.

Limitations and Future Work

  • Perfect Uplink CSI Assumption: The proposed method relies on accurate estimation of uplink channel parameters. Errors in uplink CSI estimation would directly impact the performance of the downlink beamforming.
  • Angle Mismatch Impact: While the algorithms show some robustness, performance degradation due to angle mismatch is observed, particularly at higher transmission powers. The paper suggests reducing the number of users to mitigate this.
  • Future Work: The paper mentions that studies on the angle mismatch phenomenon are postponed. Further investigation into more advanced techniques to handle angle mismatch and explore its impact on different system configurations would be valuable. The paper also suggests that the choice of Γ is closely related to system size and user density, implying that adaptive selection of Γ could be a direction for future research.

Summary updated

BibTeX
@article{yuan2020statistical,
  author = {Xiaojun Yuan and Hang Liu and Xiaojun Yuan and Ying Jun Zhang},
  title = {{Statistical Beamforming for FDD Downlink Massive MIMO via Spatial Information Extraction and Beam Selection}},
  journal = {IEEE Trans. Wireless Communications},
  year = {2020},
  doi = {10.1109/twc.2020.2985686},
}
Figure