Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Partial reciprocity-based precoding matrix prediction in FDD massive MIMO with mobility

Ziao Qin, Haifan Yin

TL;DR

This work addresses timely downlink precoding in FDD massive MIMO under mobility by proposing partial reciprocity-based, closed-form eigenvector interpolation methods. It introduces two schemes, EGVP-WCM and EGVP-CGM, to predict precoding matrices using interpolated channel weights derived from UL/DL partial reciprocity, with CGM reducing EVD complexity. Theoretical results establish a closed-form eigenvector prediction model and a correlation between wideband and Gram-based eigenvectors, while complexity analysis and simulations show substantial reductions in computation and robust performance across speeds up to 500 km/h. The proposed approach enables timely, accurate DL precoding in mobility scenarios, reducing latency and feedback burden.

Abstract

The timely precoding of frequency division duplex (FDD) massive multiple-input multiple-output (MIMO) systems is a substantial challenge in practice, especially in mobile environments. In order to improve the precoding performance and reduce the precoding complexity, we propose a partial reciprocity-based precoding matrix prediction scheme and further reduce its complexity by exploiting the channel gram matrix. We prove that the precoders can be predicted through a closed-form eigenvector interpolation which was based on the periodic eigenvector samples. Numerical results validate the performance improvements of our schemes over the conventional schemes from 30 km/h to 500 km/h of moving speed.

Partial reciprocity-based precoding matrix prediction in FDD massive MIMO with mobility

TL;DR

This work addresses timely downlink precoding in FDD massive MIMO under mobility by proposing partial reciprocity-based, closed-form eigenvector interpolation methods. It introduces two schemes, EGVP-WCM and EGVP-CGM, to predict precoding matrices using interpolated channel weights derived from UL/DL partial reciprocity, with CGM reducing EVD complexity. Theoretical results establish a closed-form eigenvector prediction model and a correlation between wideband and Gram-based eigenvectors, while complexity analysis and simulations show substantial reductions in computation and robust performance across speeds up to 500 km/h. The proposed approach enables timely, accurate DL precoding in mobility scenarios, reducing latency and feedback burden.

Abstract

The timely precoding of frequency division duplex (FDD) massive multiple-input multiple-output (MIMO) systems is a substantial challenge in practice, especially in mobile environments. In order to improve the precoding performance and reduce the precoding complexity, we propose a partial reciprocity-based precoding matrix prediction scheme and further reduce its complexity by exploiting the channel gram matrix. We prove that the precoders can be predicted through a closed-form eigenvector interpolation which was based on the periodic eigenvector samples. Numerical results validate the performance improvements of our schemes over the conventional schemes from 30 km/h to 500 km/h of moving speed.
Paper Structure (11 sections, 2 theorems, 40 equations, 4 figures, 1 table)

This paper contains 11 sections, 2 theorems, 40 equations, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. DL Channel estimation
  3. Precoding matrix prediction scheme
  4. EGVP-WCM prediction scheme
  5. EGVP-CGM prediction scheme
  6. Performance analysis
  7. Complexity analysis
  8. Numerical results and analysis
  9. Conclusion
  10. Proof of Theorem \ref{['Theorem 1']}
  11. Proof of Theorem \ref{['Theorem 2']}

Key Result

Theorem 1

Given ${{N_t},{N_f} \to \infty }$, the eigenvectors obtained from the estimated DL channels based on JADD can be estimated by the following model. where prediction error satisfies $\mathbb{E}{\left\{ {\frac{{\left\| {{{\mathbf{u}}_r^d\left( t \right)}-\tilde{\bf{u}}_r^d\left( t \right)} \right\|_2^2}}{{\left\| {{\bf{u}}_r^d\left( t \right)} \right\|_2^2}}} \right\}_{{N_L}}} \le 1- \eta$. The coef

Figures (4)

  • Figure 1: Framework of two EGVP based precoding matrix precoding schemes.
  • Figure 2: SE performances under different mobile environments with noise-free channel samples, $\kappa=\frac{1}{{8}}$.
  • Figure 3: Eigenvector PE performances under different EVD cycle lengths, $v=30$ km/h, $\kappa=\frac{1}{{16}}$.
  • Figure 4: SE performances under different channel sampling noises, $v=120$ km/h, $\kappa=\frac{1}{{8}}$.

Theorems & Definitions (4)

  • Theorem 1
  • proof
  • Theorem 2
  • proof