The Optimal DoF Region for the Two-User Non-Coherent SIMO\n Multiple-Access Channel

Khac–Hoang Ngo, Yang Sheng, Maxime GuillaudarXiv 2018 · DOI · arXiv

Abstract

The optimal degree-of-freedom (DoF) region of the non-coherent multiple-access channels is still unknown in general. In this paper, we make some progress by deriving the entire optimal DoF region in the case of the two-user single-input multiple-output (SIMO) generic block fading channels. The achievability is based on a simple training-based scheme. The novelty of our result lies in the converse using a genie-aided bound and the duality upper bound. As a by-product, our result generalizes previous proofs for the single-user Rayleigh block fading channels.

Introduction

In wireless communications, "Coherent" systems operate under the assumption that the receiver possesses perfect Channel State Information (CSI). However, in high-mobility or high-frequency scenarios, the channel changes rapidly, making CSI estimation a resource-intensive task. These are known as "Non-Coherent" systems. A fundamental question in information theory is determining the high-speed limit of such systems, represented by the Degrees of Freedom (DoF)—the number of independent signaling "pipes" available at high signal-to-noise ratios.

While the DoF for a single-user non-coherent channel is well-established, the DoF region for multiple users sharing a receiver (the Multiple-Access Channel or MAC) has remained an open problem. This paper provides a definitive answer for the two-user single-input multiple-output (SIMO) case, where each transmitter has one antenna and the receiver has $N$ antennas, over a coherence block of time $T$.

Methodology

The paper establishes the optimal DoF region through two primary theoretical pillars:

1. Achievability (The Lower Bound)

The authors demonstrate that a straightforward training-based scheme is sufficient to achieve the optimal region. In a block of $T$ symbols, the two users dedicate the first two slots to sending orthogonal "pilot" signals. This allows the receiver to estimate the channel coefficients. The remaining $T-2$ slots are then utilized for simultaneous data transmission. This result is significant because it suggests that sophisticated non-coherent space-time codes do not offer a DoF advantage over intuitive training methods in this SIMO setup.

2. Converse (The Upper Bound)

The primary technical innovation is the proof that no scheme can exceed the performance of the training-based approach. This is achieved using a duality-based upper bound combined with a genie-aided argument. By introducing a "genie" that provides the receiver with minimal auxiliary information and carefully selecting an auxiliary output distribution, the authors derive a tight outer bound. Notably, this proof is robust; it does not rely on the standard Rayleigh fading assumption and holds for any generic fading distribution.

Key Results and Insights

The research quantifies the "training tax" required for non-coherence. To effectively utilize $N$ receive antennas for two users, a minimum of two time slots must be sacrificed for training. The shape of the optimal DoF region $(d_1, d_2)$ depends heavily on the antenna count $N$ and coherence time $T$:

  • Single-Antenna Case ($N=1$ or $T \le 2$): The sum DoF is capped at $1 - 1/T$. The region is a simple triangle defined by $d_1 + d_2 \le 1 - 1/T$, indicating that Time Division Multiplexing (TDMA) is optimal.
  • Multi-Antenna Case ($N > 1$ and $T \ge 3$): The region expands into a polygon. While individual users are still capped at $1 - 1/T$, they can simultaneously achieve $(1 - 2/T, 1 - 2/T)$. The boundary is defined by the intersection of two linear constraints: $$\frac{d_1}{T-2} + d_2 \le 1 - \frac{1}{T} \quad \text{and} \quad d_1 + \frac{d_2}{T-2} \le 1 - \frac{1}{T}$$
Optimal DoF regions for varying coherence times $T$ and antenna counts $N$.

As illustrated in the figure above, as the coherence time $T$ increases, the penalty for learning the channel diminishes. When $T$ approaches infinity, the region converges to the "coherent" DoF region (the red square), where both users can simultaneously achieve a DoF of 1.

Conclusion

This work settles the DoF region for the two-user non-coherent SIMO MAC, proving the optimality of simple training-based schemes for generic fading. Future research remains to extend these techniques to the general MIMO case, where transmitters possess multiple antennas, and to scale the converse proofs for more than two users.

Summary updated

Abstract

The optimal degree-of-freedom (DoF) region of the non-coherent\nmultiple-access channels is still unknown in general. In this paper, we make\nsome progress by deriving the entire optimal DoF region in the case of the\ntwo-user single-input multiple-output (SIMO) generic block fading channels. The\nachievability is based on a simple training-based scheme. The novelty of our\nresult lies in the converse using a genie-aided bound and the duality upper\nbound. As a by-product, our result generalizes previous proofs for the\nsingle-user Rayleigh block fading channels.\n
BibTeX
@misc{ngo2018optimal,
  author = {Khac–Hoang Ngo and Yang Sheng and Maxime Guillaud},
  title = {{The Optimal DoF Region for the Two-User Non-Coherent SIMO\textbackslash\{\}n Multiple-Access Channel}},
  year = {2018},
  eprint = {1806.04102},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1806.04102},
}
Figure