Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Integrated Near Field Sensing and Communications Using Unitary Approximate Message Passing Based Matrix Factorization

Zhengdao Yuan, Qinghua Guo, Yonina C. Eldar, Yonghui Li

TL;DR

This work tackles blind joint near-field localization and signal detection in uplink ISAC by formulating JNFLSD as a matrix-factorization problem with structured factors. It introduces UAMP-MF, a unified Bayesian solver that fuses variational inference with unitary AMP, incorporating whitening to enable efficient updates of the factor matrices under separable priors. The method derives matrix-Gaussian variational messages for $\boldsymbol{A}$ and $\boldsymbol{X}$, as well as hyperparameter updates, and applies a dynamic linearization to handle the nonlinearity of near-field steering, yielding accurate distance and angle estimates along with symbol detection without pilot signals. Extensive simulations demonstrate that UAMP-MF-based JNFLSD outperforms grid-based and near-field baselines in localization accuracy and BER/FER performance, closely approaching the CRLB in several cases. The approach provides a scalable, generalizable framework for ISAC systems operating in the near-field with structured MF challenges.

Abstract

Due to the utilization of large antenna arrays at base stations (BSs) and the operations of wireless communications in high frequency bands, mobile terminals often find themselves in the near-field of the array aperture. In this work, we address the signal processing challenges of integrated near-field localization and communication in uplink transmission of an integrated sensing and communication (ISAC) system, where the BS performs joint near-field localization and signal detection (JNFLSD). We show that JNFLSD can be formulated as a matrix factorization (MF) problem with proper structures imposed on the factor matrices. Then, leveraging the variational inference (VI) and unitary approximate message passing (UAMP), we develop a low complexity Bayesian approach to MF, called UAMP-MF, to handle a generic MF problem. We then apply the UAMP-MF algorithm to solve the JNFLSD problem, where the factor matrix structures are fully exploited. Extensive simulation results are provided to demonstrate the superior performance of the proposed method.

Integrated Near Field Sensing and Communications Using Unitary Approximate Message Passing Based Matrix Factorization

TL;DR

This work tackles blind joint near-field localization and signal detection in uplink ISAC by formulating JNFLSD as a matrix-factorization problem with structured factors. It introduces UAMP-MF, a unified Bayesian solver that fuses variational inference with unitary AMP, incorporating whitening to enable efficient updates of the factor matrices under separable priors. The method derives matrix-Gaussian variational messages for and , as well as hyperparameter updates, and applies a dynamic linearization to handle the nonlinearity of near-field steering, yielding accurate distance and angle estimates along with symbol detection without pilot signals. Extensive simulations demonstrate that UAMP-MF-based JNFLSD outperforms grid-based and near-field baselines in localization accuracy and BER/FER performance, closely approaching the CRLB in several cases. The approach provides a scalable, generalizable framework for ISAC systems operating in the near-field with structured MF challenges.

Abstract

Due to the utilization of large antenna arrays at base stations (BSs) and the operations of wireless communications in high frequency bands, mobile terminals often find themselves in the near-field of the array aperture. In this work, we address the signal processing challenges of integrated near-field localization and communication in uplink transmission of an integrated sensing and communication (ISAC) system, where the BS performs joint near-field localization and signal detection (JNFLSD). We show that JNFLSD can be formulated as a matrix factorization (MF) problem with proper structures imposed on the factor matrices. Then, leveraging the variational inference (VI) and unitary approximate message passing (UAMP), we develop a low complexity Bayesian approach to MF, called UAMP-MF, to handle a generic MF problem. We then apply the UAMP-MF algorithm to solve the JNFLSD problem, where the factor matrix structures are fully exploited. Extensive simulation results are provided to demonstrate the superior performance of the proposed method.
Paper Structure (17 sections, 77 equations, 10 figures, 3 algorithms)

This paper contains 17 sections, 77 equations, 10 figures, 3 algorithms.

Table of Contents

  1. Introduction
  2. Signal Model and Problem Formulation for Integrated Near Field Sensing and Communications
  3. Matrix Factorization Using Unitary Approximate Message Passing
  4. Variational Inference
  5. (U)AMP
  6. Design of UAMP-MF
  7. Update of $q(\boldsymbol{X})$
  8. Update of $q(\boldsymbol{A})$
  9. Update of $q(\lambda)$
  10. Discussion and Computational Complexity
  11. Hyper-Parameter Learning
  12. UAMP-MF for Integrated Near Field Localization and Communications
  13. Simulation Results
  14. Conclusions
  15. Proof of Proposition 1
  16. ...and 2 more sections

Figures (10)

  • Figure 1: Illustration of near-field angle-distance dependent signal model (only a single user is shown).
  • Figure 2: Factor graph of \ref{['eq:factor']}, where $f_A \triangleq p(\boldsymbol{A})$, $f_{Y} \triangleq p(\boldsymbol{Y}|\boldsymbol{X},\boldsymbol{A},\lambda)$, $f_X \triangleq p(\boldsymbol{X})$, and $f_{\lambda}\triangleq p(\lambda)$.
  • Figure 3: Factor graph for hyper-parameters learning, where $f_{x_{nl}}\triangleq p(x_{nl}|\gamma_{nl})$ and $f_{\gamma_{nl}}\triangleq p(\gamma_{nl})$.
  • Figure 4: Spatial power spectrum (normalized), the true locations (denoted by "$\mathrm{o}$") and estimated locations (denoted by "$\times$") of active users using the proposed algorithm.
  • Figure 5: (a) NMSE of distance estimation and (b) MSE of angle estimation versus grid size with $d_{max}$= 20m and SNR=-4dB and -6dB.
  • ...and 5 more figures

Theorems & Definitions (3)

  • proof
  • proof
  • proof