SIR Coverage Analysis in Multi-Cell Downlink Systems With Spatially Correlated Queues

Xiaozhou Zhang, Howard H. Yang, Xiaofang Sun, Gang Zhu, Tony Q. S. Quek, Zhangdui ZhongIEEE Access 2020 · DOI

Introduction

Traditional wireless network modeling often relies on simplifying assumptions, such as "full-buffer" traffic or independent queues across base stations (BSs). However, these assumptions fail to capture the complex spatial dynamics of real-world networks. This paper introduces a novel framework that accounts for the spatial correlation of traffic queues, providing a more accurate prediction of Signal-to-Interference Ratio (SIR) coverage, especially in high-traffic scenarios.

The Challenge of Spatially Correlated Queues

In dense wireless environments, BSs are subject to a "vicious cycle" of spatial correlation. BSs located in crowded areas experience higher mutual interference, which reduces their service rates. Lower service rates lead to longer queues, keeping these BSs active for longer periods. This prolonged activity, in turn, maintains high interference levels for neighboring BSs.

The interaction between the MAC layer queues and the physical layer interference.

Ignoring this interaction leads to overly optimistic performance models. The mathematical complexity of modeling these interdependent queues has historically been a significant barrier. This research bridges the gap by combining stochastic geometry with queueing theory to create a tractable analytical framework.

Methodology: The Two-Set Model

The authors propose a "two-queue-length" model to approximate the steady-state behavior of the network:

  1. Stability Analysis: The system is analyzed for $\epsilon$-stability to ensure it remains functional without infinite queue growth.
  2. Spatial Modeling: BSs are categorized into two sets:
  3. Long-queue BSs: Modeled using a Neyman-Scott Process (NSP). This cluster process accurately represents the "hotspots" where high interference leads to persistent activity.
  4. Short-queue BSs: Modeled as a residual hole process, approximated by a standard Poisson Point Process (PPP) for mathematical simplicity.
  5. Iterative Approximation: An iterative algorithm calculates the probability of a BS being active. To maintain computational efficiency, the authors use a Beta distribution approximation for service rates, matching the first two moments (mean and variance).
Spatial distribution of long-queue and short-queue base stations, highlighting the clustering effect.

Key Insights and Results

The research yields several critical findings that challenge standard modeling practices:

  • Clustering is Real: Simulations confirm that active BSs naturally form clusters. Modeling these clusters with an NSP is significantly more accurate than the random distribution assumed by PPP models.
  • The Optimism Bias: Standard "Independent and Identically Distributed" (IID) queue models significantly overestimate coverage. For instance, at a decoding threshold of 12 dB, the IID model predicts a coverage probability nearly 1.8 times higher than what is observed in realistic simulations.
  • Model Accuracy: The proposed model achieves high fidelity, matching Monte Carlo simulations with a gap of less than 5.63%.
  • Efficiency: The iterative algorithm is highly convergent, typically reaching a solution within five iterations, making it a practical tool for network planning.

Conclusion and Future Directions

By moving beyond the full-buffer and independent queue assumptions, this work provides a more realistic lens for analyzing downlink performance. The study highlights that as traffic load increases, spatial correlation becomes a dominant factor that cannot be ignored.

Future research will focus on extending this framework to multi-user scenarios, ultra-dense networks, and heterogeneous deployments. Additionally, the impact of correlated queues on latency-critical applications, such as Vehicle-to-Everything (V2X) communications, remains a vital area for exploration.

Summary updated

BibTeX
@article{zhang2020sir,
  author = {Xiaozhou Zhang and Howard H. Yang and Xiaofang Sun and Gang Zhu and Tony Q. S. Quek and Zhangdui Zhong},
  title = {{SIR Coverage Analysis in Multi-Cell Downlink Systems With Spatially Correlated Queues}},
  journal = {IEEE Access},
  year = {2020},
  doi = {10.1109/access.2020.2990219},
}
Figure