On the Design of Large Scale Wireless Systems

Rohit Aggarwal, Can Emre Koksal, Philip Schniter

Abstract

As wireless networks evolve toward massive, dense deployments (5G/6G), traditional theoretical models often fail due to mathematical singularities in path-loss assumptions. This paper presents a rigorous framework for analyzing large-scale wireless systems using a truncated path-loss model and Extreme Value Theory (EVT). We derive sum-rate bounds and scaling laws across various fading environments, establishing design principles for maintaining per-user throughput and network profitability.

Introduction

The rapid expansion of wireless infrastructure—comprising base stations, femtocells, and a burgeoning number of mobile devices—demands a robust understanding of network scalability. Traditional "far-field" path-loss models often predict capacity scaling as $\Theta(\log K)$, where $K$ is the number of users. However, these models contain a singularity where signal strength approaches infinity as distance goes to zero, leading to unrealistic performance predictions.

This research addresses these discrepancies by introducing a realistic, non-asymptotic framework for large-scale Orthogonal Frequency Division Multiple Access (OFDMA) systems. By eliminating the zero-distance singularity, the study provides a blueprint for service providers to optimize resource allocation and infrastructure density.

Methodology

The study models a downlink network with $B$ transmitters, $K$ users, and $N$ resource blocks. Key methodological innovations include:

  • Truncated Path-Loss Model: A minimum distance $r_0$ is introduced to ensure the model aligns with physical reality and avoids mathematical singularities.
  • Extreme Value Theory (EVT): EVT is utilized to analyze the "expected maximum" channel gain, recognizing that system performance in dense networks is primarily driven by the strongest available links.
  • Mathematical Bounding: The authors derive upper bounds by ignoring interference and lower bounds by assuming equal-power allocation across resource blocks.
  • Distributed Resource Allocation: A decentralized scheme is proposed where users calculate their own Signal-to-Interference-plus-Noise Ratio (SINR), enabling optimal scaling without the overhead of a centralized controller.

The framework is evaluated across several statistical fading environments, including Rayleigh, Nakagami-m, Weibull, and LogNormal fading, to ensure applicability across diverse urban and indoor propagation scenarios.

Scaling Laws and Network Design

A pivotal finding of this work is the shift in capacity scaling. Under a realistic path-loss model, the capacity scaling law transitions from $\Theta(\log K)$ to $\Theta(\log \log K)$. This suggests that "multi-user diversity" provides diminishing returns more quickly than previously thought.

To ensure a network remains "scalable"—defined as maintaining non-vanishing per-user throughput as the user base grows—the total resources ($BN$) must scale at least as $\Omega(K / \log \log K)$. Furthermore, the study demonstrates that in peer-to-peer networks, linear sum-rate growth with the number of nodes $B$ is only achievable if $B$ grows very slowly relative to $K$ (specifically $B = O(\frac{\log K}{\log \log K})$).

Key Results

The paper establishes rigorous sum-rate bounds for various fading models. For instance, in Rayleigh fading environments, the expected sum-rate $C^*$ scales as $O(BN \log \log K)$. The following table summarizes these results for other common fading distributions, showing how the scaling laws adapt to different statistical environments.

Sum-rate scaling laws for various fading distributions.

These results lead to four specific "Design Principles" for regulators and providers, balancing network stability with Return on Investment (ROI). Notably, the proposed distributed resource allocation scheme achieves the same sum-rate scaling as the theoretical upper bound, proving that high performance is attainable without expensive centralized coordination.

Conclusion and Future Directions

While this framework provides a significant step toward realistic network modeling, some limitations remain. The current bounds assume perfect Channel State Information (CSI) at the transmitters, and the focus is primarily on downlink performance. Future research will explore extending these proofs to uplink scenarios and investigating the impact of "constant factors" that scaling laws typically omit, which may be critical for fine-tuning smaller-scale practical deployments.

Summary updated

BibTeX
@article{aggarwal2013design,
  author = {Rohit Aggarwal and Can Emre Koksal and Philip Schniter},
  title = {{On the Design of Large Scale Wireless Systems}},
  journal = {IEEE J. Sel. Areas Commun.},
  year = {2013},
  url = {https://openreview.net/forum?id=CeBw5YHD3K1S},
}
Figure