Tensor-Based Parameter Estimation of Double Directional Massive MIMO Channel with Dual-Polarized Antennas

Cheng Qian, Xiao Fu, Nicholas D. Sidiropoulos, Ye Yang

Abstract

The 3GPP suggests combining dual-polarized (DP) antenna arrays with the double directional (DD) channel model for downlink channel estimation. This combination balances high-capacity communications with parsimonious channel modeling, enabling limited feedback schemes. However, most existing DD model channel estimation work has not considered DP arrays due to complex array manifolds and algorithm design difficulties. This paper reveals that the DD channel with DP arrays can be modeled as a low-rank four-way tensor, allowing for effective parameter estimation via tensor decomposition. To reduce complexity, the problem is recast as a four-snapshot three-dimensional harmonic retrieval problem solvable by efficient subspace methods. Theoretically, the DD channel with DP arrays is shown to be identifiable under mild conditions. Numerical simulations demonstrate the effectiveness of these methods.

Introduction

In the evolution toward 5G and beyond, Massive MIMO (Multiple-Input Multiple-Output) stands as a cornerstone technology. To enhance practicality, engineers employ Dual-Polarized (DP) antennas to increase capacity within compact spaces and use the Double Directional (DD) channel model to simplify complex radio environments into dominant paths defined by angles of arrival and departure.

However, integrating DP antennas with the DD model introduces significant mathematical complexity. The resulting array manifold is difficult to manage with traditional techniques. Existing methods often either ignore polarization or rely on Compressive Sensing (CS), which suffers from high computational costs due to the massive dictionaries required for high-resolution 2D arrays.

Proposed Tensor-Based Framework

The authors propose a shift from dictionary-based methods to tensor algebra, exploiting the inherent multi-dimensional structure of the signal.

  1. Tensor Modeling: The DD channel with DP antennas is represented as a low-rank four-way tensor. This captures the relationships between horizontal and vertical polarizations at both the transmitter and receiver.
  2. PARAFAC Decomposition: Parallel Factor Analysis (PARAFAC) is used to decompose the tensor and extract underlying parameters like angles and path losses directly.
  3. Improved Multidimensional Folding (IMDF): To enhance efficiency, the estimation is recast as a 3D harmonic retrieval problem. By leveraging the Vandermonde structure of uniform antenna arrays, subspace-based methods are applied, offering significant speed improvements over raw tensor decomposition.
  4. Identifiability: The paper proves that parameters can be uniquely identified even in "overloaded" scenarios where signal paths exceed the number of physical antennas.

Simulation Setup

The methods were evaluated through numerical simulations:

  • Scenario: A 3GPP-compliant setup with a $4 \times 8$ DP Uniform Rectangular Array at the Base Station and a 2-element DP Uniform Linear Array at the Mobile Station.
  • Channel Conditions: Multipath propagation gains followed a Rician distribution, with random angles simulating urban environments.
  • Benchmarks: Compared against standard Least Squares (LS) and Compressive Sensing (CS) approaches.

Key Results and Analysis

The simulation results highlight the advantages of the tensor-based approach over traditional benchmarks.

NMSE performance comparison of different estimation methods under known and unknown path counts (K).
  • Superior Accuracy: The PARAFAC method consistently outperformed LS and CS in terms of Normalized Mean Square Error (NMSE).
  • High SNR Robustness: At high Signal-to-Noise Ratios (above 14 dB), CS performance plateaued due to fixed dictionary resolution, while tensor methods (PARAFAC and IMDF) continued to improve.
  • Computational Efficiency: The IMDF algorithm emerged as a practical "sweet spot," providing accuracy comparable to PARAFAC but with significantly lower computational overhead, making it suitable for real-time applications.
  • Path Count Sensitivity: While performance remains strong, a slight dip in accuracy occurs when the number of signal paths ($K$) is unknown and must be estimated, as illustrated in the comparison between known and unknown path counts.

Conclusion and Future Work

By treating the channel as a tensor rather than a flat matrix, this research preserves the spatial and polarization structure of radio waves, allowing for high-resolution estimation that breaks the traditional antenna limit. Future work will involve testing these algorithms against real-world field data to further validate the simulation results and adjusting for potential array miscalibrations that might disrupt the assumed mathematical structures.

Summary updated

Abstract

The 3GPP suggests to combine dual polarized (DP) antenna arrays with the double directional (DD) channel model for downlink channel estimation. This combination strikes a good balance between high-capacity communications and parsimonious channel modeling, and also brings limited feedback schemes for downlink channel estimation within reach. However, most existing channel estimation work under the DD model has not considered DP arrays, perhaps because of the complex array manifold and the resulting difficulty in algorithm design. In this paper, we first reveal that the DD channel with DP arrays at the transmitter and receiver can be naturally modeled as a low-rank four-way tensor, and thus the parameters can be effectively estimated via tensor decomposition algorithms. To reduce computational complexity, we show that the problem can be recast as a four-snapshot three-dimensional harmonic retrieval problem, which can be solved using computationally efficient subspace methods. On the theory side, we show that the DD channel with DP arrays is identifiable under very mild conditions, leveraging identifiability of low-rank tensors. Numerical simulations are employed to showcase the effectiveness of our methods.
BibTeX
@inproceedings{qian2018tensor,
  author = {Cheng Qian and Xiao Fu and Nicholas D. Sidiropoulos and Ye Yang},
  title = {{Tensor-Based Parameter Estimation of Double Directional Massive MIMO Channel with Dual-Polarized Antennas}},
  booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing},
  year = {2018},
  doi = {10.1109/icassp.2018.8461557},
}
Figure