Joint Constellation Design for Noncoherent MIMO Multiple-Access Channels

Khac–Hoang Ngo, Sheng Yang, Maxime Guillaud, Alexis DecurningeIEEE Trans. Information Theory 2022 · DOI · arXiv

Abstract

We consider the joint constellation design problem for the noncoherent multiple-input multiple-output multiple-access channel (MAC). By analyzing the noncoherent maximum-likelihood detection error, we propose novel design criteria so as to minimize the error probability. As a baseline approach, we adapt several existing design criteria for the point-to-point channel to the MAC. Furthermore, we propose new design criteria. Our first proposed design metric is the dominating term in nonasymptotic lower and upper bounds on the pairwise error probability exponent. We give a geometric interpretation of the bound using Riemannian distance in the manifold of Hermitian positive definite matrices. From an analysis of this metric at high signal-to-noise ratio, we obtain further simplified metrics. For any given set of constellation sizes, the proposed metrics can be optimized over the set of constellation symbols. Motivated by the simplified metric, we propose a simple constellation construction consisting in partitioning a single-user constellation. We also provide a generalization of our previously proposed construction based on precoding individual constellations of lower dimensions. Numerical results show that our proposed metrics are meaningful, and can be used as objectives to generate constellations through numerical optimization that perform better, for the same transmission rate and power constraint, than a common pilot-based scheme and the constellations optimized with existing metrics.

Introduction

In traditional MIMO (Multiple-Input Multiple-Output) systems, the receiver typically requires knowledge of the channel state information (CSI), often acquired through pilot signals. However, in fast-fading environments or systems with high antenna counts, the overhead of these pilots becomes prohibitive. This motivates the noncoherent approach, where communication proceeds without prior CSI. While single-user noncoherent MIMO is well-established—often utilizing signal subspaces on a Grassmann manifold—the multi-user Multiple-Access Channel (MAC) case presents significant challenges. Designing symbols for independent users that remain distinguishable after being transformed by unknown channels requires novel joint constellation strategies that go beyond simple adaptations of single-user theory.

Methodology

The core of the proposed approach lies in analyzing the Pairwise Error Probability (PEP)—the likelihood that a receiver misidentifies one joint set of transmitted symbols for another.

  1. New Design Metrics: The authors introduce a metric, $b_{min}$, derived from the dominating terms of nonasymptotic upper and lower bounds of the PEP exponent. This is more robust for practical systems with finite antennas than traditional asymptotic metrics.
  2. Geometric Interpretation: Minimizing error is shown to be equivalent to maximizing the Riemannian distance between matrices in the manifold of Hermitian positive definite matrices. This provides a visual and mathematical framework for ensuring that the "footprints" of different symbol combinations are maximally separated on a curved mathematical surface.
Geometric interpretation of the distance between joint symbols on a manifold.
  1. Constellation Construction: Two practical construction methods are proposed: Partitioning, which splits a high-performing single-user constellation among users, and Precoding, which assigns unique subspace "signatures" to each user.
  2. Optimization: The researchers utilize the Manopt toolbox to perform conjugate gradient descent directly on the Grassmann manifold, optimizing the specific coordinates of the symbols.

Evaluation is conducted using synthetic Rayleigh block-fading channels and Monte Carlo simulations to empirically measure the Symbol Error Rate (SER) across thousands of channel realizations.

Key Insights

  • Nonasymptotic Robustness: By focusing on nonasymptotic bounds rather than limits at infinity, the $b_{min}$ metric provides better performance in realistic scenarios with limited antenna arrays.
  • The Manifold Connection: Framing noncoherent MAC as a separation problem on a Riemannian manifold allows for the application of advanced differential geometry tools to wireless communication.
  • Accounting for Interference: Unlike single-user systems where signal matrices can be purely unitary, MAC signals are sums of independent matrices. The proposed metrics explicitly account for this "sum-structure" interference, which is critical for multi-user performance.

Results and Performance

Numerical results demonstrate that constellations optimized with the proposed $d_{min}$ and $J_{1/2,min}$ metrics consistently achieve lower SER than standard pilot-based schemes. In a 2-user setup ($T=5, M=2, N=4$), these designs outperformed metrics adapted from point-to-point MIMO by several decibels at high SNR. Furthermore, the study highlights that optimizing per-user transmit power—rather than assuming uniform power—yields significant gains, particularly when users transmit at different rates.

Limitations and Future Work

While effective, the manifold optimization process is computationally intensive as constellation sizes or user counts grow. Additionally, the current model assumes a Maximum Likelihood (ML) detector, which has exponential complexity. Future research will focus on developing low-complexity receivers and addressing the challenges of time-synchronization in distributed multiple-access systems.

Summary updated

Abstract

We consider the joint constellation design problem for the noncoherent multiple-input multiple-output multiple-access channel (MAC). By analyzing the noncoherent maximum-likelihood detection error, we propose novel design criteria so as to minimize the error probability. As a baseline approach, we adapt several existing design criteria for the point-to-point channel to the MAC. Furthermore, we propose new design criteria. Our first proposed design metric is the dominating term in nonasymptotic lower and upper bounds on the pairwise error probability exponent. We give a geometric interpretation of the bound using Riemannian distance in the manifold of Hermitian positive definite matrices. From an analysis of this metric at high signal-to-noise ratio, we obtain further simplified metrics. For any given set of constellation sizes, the proposed metrics can be optimized over the set of constellation symbols. Motivated by the simplified metric, we propose a simple constellation construction consisting in <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">partitioning</i> a single-user constellation. We also provide a generalization of our previously proposed construction based on <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">precoding</i> individual constellations of lower dimensions. For a fixed joint constellation, the design metrics can be further optimized over the per-user transmit power, especially when the users transmit at different rates. Considering unitary space-time modulation, we investigate the option of building each individual constellation as a set of truncated unitary matrices scaled by the respective transmit power. Numerical results show that our proposed metrics are meaningful, and can be used as objectives to generate constellations through numerical optimization that perform better, for the same transmission rate and power constraint, than a common pilot-based scheme and the constellations optimized with existing metrics.
BibTeX
@article{ngo2022joint,
  author = {Khac–Hoang Ngo and Sheng Yang and Maxime Guillaud and Alexis Decurninge},
  title = {{Joint Constellation Design for Noncoherent MIMO Multiple-Access Channels}},
  journal = {IEEE Transactions on Information Theory},
  year = {2022},
  doi = {10.1109/tit.2022.3189254},
}
Figure