Rate-Splitting for CF Massive MIMO Systems With Channel Aging

Jiakang Zheng, Jiayi Zhang, Hongyang Du, Dusit Niyato, Dong In Kim, Bo Ai

TL;DR

This paper introduces Rate-Splitting (RS) technology to Cell-Free (CF) Massive MIMO systems to combat "channel aging"—the performance degradation caused by user movement. The authors develop a novel bisection-based precoding scheme and derive closed-form expressions for spectral efficiency, proving that RS significantly improves system robustness and interference management in high-mobility environments.

What Problem Does This Paper Address?

In the quest for 6G, Cell-Free (CF) Massive MIMO is a frontrunner. Unlike traditional cellular networks, CF systems use a massive number of distributed Access Points (APs) to serve users simultaneously, eliminating "cell edges" and providing uniform coverage.

However, these systems rely heavily on accurate Channel State Information (CSI). In the real world, users move. By the time an AP calculates a signal (precoding) based on measured CSI, the user has already moved, making the CSI "outdated." This phenomenon, known as channel aging, creates massive multi-user interference. While standard precoding techniques struggle to handle this "stale" data, this paper investigates whether Rate-Splitting (RS) can bridge the gap.

Methods

The authors model a downlink CF Massive MIMO system where $L$ APs serve $K$ users. They incorporate the Jakes’ model to simulate how channels evolve over time due to velocity.

To mitigate aging-induced interference, they employ Rate-Splitting (RS). In RS, a user's message is split into two parts:

  1. A Common Message: Combined with other users' common parts into one super-message that all users decode first.
  2. A Private Message: Decoded only by the intended user after removing the common message via Successive Interference Cancellation (SIC).

Key Technical Contributions: * Precoding Design: They compare three private precoding schemes: Maximum Ratio (MR), Local MMSE, and Centralized MMSE. * Bisection-Based Algorithm: They propose a novel algorithm to optimize the precoding of the common message. This is a max-min optimization problem designed to ensure all users can reliably decode the common part, even with outdated CSI. * Closed-Form Analysis: Using the "Use-and-Then-Forget" (UatF) bound, they derive a novel closed-form expression for the sum Spectral Efficiency (SE), allowing for performance prediction without exhaustive simulations.

Datasets

The paper relies on numerical simulations rather than a static dataset. The setup includes:

  • Environment: A $250\text{ m} \times 250\text{ m}$ square area with $L=20$ APs and $K=4$ users distributed uniformly.
  • Channel Model: Three-slope propagation model with Rayleigh fading.
  • Mobility: User velocities ranging from stationary ($0\text{ m/s}$) to high-speed ($30\text{ m/s}$ or $\sim108\text{ km/h}$).
  • Parameters: Carrier frequency of $2\text{ GHz}$, bandwidth of $20\text{ MHz}$, and a sampling time of $67\text{ }\mu\text{s}$.

Key Insights and Ideas

  • RS as an Interference Buffer: The "common message" in RS acts as a robust layer. Because it is designed to be decodable by everyone (even those with the worst channels), it effectively "soaks up" a portion of the interference that would otherwise crash a conventional system.
  • The "Saturation" Fix: In conventional systems, increasing transmit power eventually stops helping because interference grows just as fast as the signal (SE saturation). RS breaks this cycle, allowing SE to continue growing with power even when CSI is outdated.
  • Robustness to Velocity: The proposed bisection-based precoding for common messages is significantly more robust in complex mobile scenarios compared to simpler "superposition-based" or "random" precoding.
  • Centralized vs. Local: While Centralized MMSE provides the best performance, it requires massive backhaul data. RS provides a way to boost the performance of lower-complexity "Local" schemes to be more competitive.

Key Results

  • Superiority of Bisection Precoding: The proposed bisection-based scheme for common messages consistently outperformed superposition-based methods, especially as the channel aged (higher time indices).
  • Spectral Efficiency Gains: In high-mobility scenarios ($30\text{ m/s}$), the RS-assisted system maintained a significantly higher sum SE compared to conventional CF Massive MIMO without RS.
  • Power Splitting Optimization: The authors found that as CSI becomes more outdated, a larger "power-splitting factor" ($t$) is required. Essentially, the system needs to put more energy into the "safe" common message and less into the "risky" private messages as the channel becomes less predictable.

Limitations and Future Work

  • Uniform Power Splitting: The study assumes the power-splitting factor ($t$) is the same across all APs. In a real-world heterogeneous deployment, optimizing $t$ per AP could yield even higher gains.
  • Perfect Initial CSI: The authors assumed CSI is perfectly known at time $t=0$ and only ages thereafter. In reality, initial channel estimation is also imperfect.
  • Future Direction: The authors suggest that future research should explore dynamic power-splitting factors and the impact of RS on the uplink (user-to-AP) in the presence of channel aging.

Summary updated

Abstract

The cell-free (CF) massive multiple-input multiple-output (MIMO) system is considered a cutting-edge technology in next-generation mobile communication due to its ability to provide high-performance coverage seamlessly and uniformly. This paper aims to mitigate the negative impact of channel aging due to the movement of user equipment in CF massive MIMO systems by utilizing rate-splitting (RS) technology. Taking into account the outdated channel state information, we obtain the achievable sum spectral efficiency (SE) of the RS-assisted CF massive MIMO system, where the private messages can directly adopt conventional maximum ratio, local minimum mean square error (MMSE), and centralized MMSE precoding schemes. Moreover, we propose a bisection-based precoding scheme that maximizes the minimum SE of common messages, which outperforms the superposition-based and random precoding schemes and exhibits strong robustness in complex mobile scenarios. Furthermore, we derive a novel closed-form sum SE expression for the considered system. The results demonstrate that RS technology can mitigate interference in mobile CF massive MIMO systems, improving overall system performance.
BibTeX
@article{zheng2024rate,
  author = {Jiakang Zheng and Jiayi Zhang and Hongyang Du and Dusit Niyato and Dong In Kim and Bo Ai},
  title = {{Rate-Splitting for CF Massive MIMO Systems With Channel Aging}},
  journal = {IEEE Trans. Vehicular Technology},
  year = {2024},
  url = {https://openreview.net/forum?id=DI2DSVIFKj},
}

Loading related content...

Figure