Resource Allocation for Massive MIMO HetNets with Quantize-Forward Relaying

Ahmad Abu Al Haija, Min Dong, Ben Liang, Gary BoudreauarXiv 2021 · arXiv

Abstract

In the evolution toward 5G and beyond, Heterogeneous Networks (HetNets) are pivotal for enhancing coverage and capacity. This paper addresses the challenge of uplink transmission in HetNets by combining massive MIMO technology with Quantize-Forward (QF) relaying. We introduce a novel "QF-WZTD" scheme that employs Wyner-Ziv binning and time-division transmission. This approach achieves the same data rates as complex joint decoding methods but with significantly reduced, linear computational complexity, making it suitable for ultra-dense network deployments.

Introduction

Heterogeneous Networks (HetNets), characterized by the deployment of small cells within larger macro cells, are essential for meeting the increasing demand for data capacity. In the uplink, users far from the Macro-Cell Base Station (MCBS) but near a Small-Cell Base Station (SCBS) can utilize the SCBS as a relay.

Heterogeneous Network (HetNet) architecture showing the interaction between Macro-Cell (MCBS) and Small-Cell (SCBS) base stations.

The primary challenge lies in the relay's processing strategy. While "Decode-and-Forward" (DF) requires full decoding at the relay—often impossible over weak links—"Quantize-Forward" (QF) offers more flexibility by digitizing the received signal. However, traditional QF requires Joint Decoding (JD) at the macro cell, which suffers from exponential computational complexity as the number of users increases.

System Model and Proposed Methods

The authors leverage the channel hardening and signal orthogonalization properties of Massive MIMO to simplify the network model and reduce interference.

Mathematical model of the multi-user relay channel illustrating direct links ($h_{d1}, h_{d2}$) and relay links ($h_{r1}, h_{r2}, h_{dr}$).

QF-WZTD Scheme

The core contribution is the QF-WZTD scheme, which combines Wyner-Ziv (WZ) binning with Time-Division (TD) transmission:

  • Wyner-Ziv Binning: This source coding technique allows the MCBS to decode quantized signals from the SCBS using the direct user-to-MCBS signal as "side information."
  • Time-Division Transmission: The SCBS transmits quantized data for different users in distinct time slots, avoiding the need for complex joint decoding.

Optimization and Detection

A novel water-filling algorithm, based on Descartes’ rule of signs, was developed to optimize quantization levels. Unlike traditional methods, this algorithm relies on large-scale fading (pathloss) rather than instantaneous channel states, reducing feedback overhead. Both small and macro cells utilize Zero-Forcing (ZF) detection to efficiently separate user streams.

Key Insights

  • Complexity Reduction: The QF-WZTD scheme scales linearly ($O(K)$) with the number of users, a massive improvement over the exponential ($O(2^K)$) scaling of standard QF-JD.
  • Performance Equivalence: Mathematically, QF-WZTD is proven to achieve the same rate region as the more complex joint decoding approach.
  • Adaptive Quantization: The optimal strategy dictates that the SCBS should use finer quantization for users with stronger links to the small cell relative to their direct macro cell link.
  • Ultra-Lean Design: By basing resource allocation on slowly changing large-scale fading, the system minimizes the frequency of feedback required from the macro cell.

Results and Discussion

Numerical simulations modeled a 5G environment with 500 antennas at the MCBS and 50 at the SCBS. Key findings include:

  • Scalability: For a 10-user scenario, the codebook size for QF-WZTD is proportional to $10 \cdot 2^{0.5n}$, compared to $2^{10n}$ for QF-JD.
  • Throughput Gains: The proposed QF schemes consistently outperform direct transmission and standard LTE-A dual-hop relaying, particularly for cell-edge users.
  • Algorithm Accuracy: The water-filling algorithm effectively identifies which users benefit from relaying and which should rely on direct links.

Limitations and Future Work

The current study assumes full-duplex operation at the SCBS; future research should investigate the impact of residual self-interference in practical half-duplex or imperfect full-duplex systems. Additionally, while ZF detection is effective for massive MIMO, exploring non-linear detectors in smaller antenna configurations and considering high-mobility scenarios with rapidly changing large-scale fading remain open areas for investigation.

Summary updated

Abstract

We investigate how massive MIMO impacts the uplink transmission design in a heterogeneous network (HetNet) where multiple users communicate with a macro-cell base station (MCBS) with the help of a small-cell BS (SCBS) with zero-forcing (ZF) detection at each BS. We first analyze the quantize-forward (QF) relaying scheme with joint decoding (JD) at the MCBS. To maximize the rate region, we optimize the quantization of all user data streams at the SCBS by developing a novel water-filling algorithm that is based on the Descartes' rule of signs. Our result shows that as a user link to the SCBS becomes stronger than that to the MCBS, the SCBS deploys finer quantization to that user data stream. We further propose a new simplified scheme through Wyner-Ziv (WZ) binning and time-division (TD) transmission at the SCBS, which allows not only sequential but also separate decoding of each user message at the MCBS. For this new QF-WZTD scheme, the optimal quantization parameters are identical to that of the QF-JD scheme while the phase durations are conveniently optimized as functions of the quantization parameters. Despite its simplicity, the QF-WZTD scheme achieves the same rate performance of the QF-JD scheme, making it an attractive option for future HetNets.
BibTeX
@misc{haija2021resource,
  author = {Ahmad Abu Al Haija and Min Dong and Ben Liang and Gary Boudreau},
  title = {{Resource Allocation for Massive MIMO HetNets with Quantize-Forward Relaying}},
  year = {2021},
  eprint = {2105.11581},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2105.11581},
}
Figure