Rician ${K}$ -Factor-Based Analysis of XLOS Service Probability in 5G Outdoor Ultra-Dense Networks

Hatim Chergui, Mustapha Benjillali, Mohamed-Slim Alouini

TL;DR

This paper introduces X-line-of-sight (XLOS) service probability as a novel performance metric for 5G ultra-dense networks (UDNs), moving beyond binary line-of-sight (LOS) models to a "gradual visibility" approach. The authors derive a rigorous closed-form expression for this probability using the multivariate Fox H-function and provide a GPU-accelerated MATLAB tool to make these complex calculations practical for network optimization.

What Problem Does This Paper Address?

In the transition to 5G and beyond, networks are becoming "ultra-dense," relying on technologies like massive MIMO and millimeter-wave (mmWave) frequencies. These technologies are highly sensitive to the visibility conditions between the base station (gNB) and the user equipment (UE).

Traditionally, researchers have treated visibility as a binary state: either you have Line-of-Sight (LOS) or you don't (NLOS). However, real-world environments involve a spectrum of "gradual visibility"—which the authors call XLOS—encompassing LOS, obstructed-LOS (OLOS), and NLOS.

The challenge is that network operators need to know the probability of achieving a certain visibility threshold (defined by the Rician $K$-factor) to make critical decisions, such as:

  • Algorithm Switching: Deciding when to use high-complexity massive MIMO localization vs. simpler triangulation.
  • Traffic Offloading: Determining when to move a user from a sub-6GHz "anchor" carrier to a high-capacity mmWave carrier based on link quality.

Prior to this work, there was no unified mathematical framework to calculate this XLOS service probability in a multi-tier, heterogeneous network environment.

Methods

The authors employ stochastic geometry and advanced special functions to solve this problem:

  • System Modeling: They model the network as an $N$-tier outdoor environment where base station locations follow a Poisson Point Process (PPP). They incorporate lognormal shadowing and a distance-based lognormal $K$-factor model.
  • XLOS Definition: They define XLOS service probability as the likelihood that at least one cell in a user's "monitoring set" (the $M$ strongest cells) exceeds a specific $K$-factor threshold ($K_{th}$).
  • The Multivariate Fox H-Function: The core mathematical contribution is deriving the probability expression in terms of the multivariate Fox H-function. Because this function is notoriously difficult to compute, the authors developed a GPU-enabled MATLAB routine.
  • Automated Contour Generation: They used linear optimization to automate the definition of the Mellin-Barnes complex contours required to evaluate the H-function, significantly reducing the manual effort usually required for such complex math.
  • Residue Theory: To make the results more intuitive, they applied residue theory to find "asymptotic" behaviors—simplified formulas that describe how the network performs at very high or very low visibility thresholds.

Datasets

The paper does not use a machine learning training dataset. Instead, it uses empirical parameters from a known measurement campaign conducted in New Jersey (referenced as [6] in the paper) to calibrate the model. These parameters include:

  • Standard deviations for shadowing ($\sigma_n$) and $K$-factor ($\sigma_K = 8 \text{ dB}$).
  • Antenna heights ($h_0 = 3 \text{ m}$) and beamwidths ($\theta_0 = 17^\circ$).
  • Path-loss exponents ($\nu = 3$).

The analytical model is then validated against extensive Monte-Carlo simulations generated by the authors to ensure the mathematical derivations match simulated network behavior.

Key Insights and Ideas

  • The $K$-factor as a Resource Trigger: The paper argues that the Rician $K$-factor is a more useful metric for 5G mobility management than simple signal strength (RSRP), as it directly correlates with the effectiveness of beamforming and mmWave links.
  • Unified Lognormal CDF Expression: The authors present a new theorem (Theorem 1) that provides a unified expression for the product of multiple independent lognormal Cumulative Distribution Functions (CDFs). This is a significant mathematical tool that could be applied to many other fields in communications and statistics.
  • Density Matters: The results show that in Ultra-Dense Networks (UDN), the "low ratio" regime dominates, meaning users almost always have excellent visibility conditions (probability $\approx 1$). In contrast, Low-Density Networks (LDN) frequently fall into NLOS conditions where XLOS probability drops sharply.
  • Multi-Cell Monitoring: By considering the top 2 or 3 strongest cells (rather than just the primary serving cell), the XLOS probability increases substantially, highlighting the importance of multi-connectivity in 5G.

Key Results

  • Mathematical Accuracy: The derived closed-form expressions perfectly match the Monte-Carlo simulation results across various scenarios (UDN, HetNet, and LDN).
  • Visibility Thresholds: For a UDN, a high visibility condition (e.g., $K_{th} = 12 \text{ dB}$) is achieved with nearly 100% probability. However, in an LDN, the probability of even a modest $K$-factor (e.g., $K = -5 \text{ dB}$) can drop to 0.5, indicating frequent NLOS issues.
  • Computational Efficiency: The introduction of the GPU-oriented MATLAB routine allows for the evaluation of these complex multivariate functions in a timeframe practical for researchers.

Limitations and Future Work

  • Static/Low Mobility: The current model is optimized for "outdoor fixed and/or sporadically moving UEs" (like IoT sensors or fixed wireless access).
  • The Vehicular Challenge: The authors explicitly note that extending this XLOS model to high-speed vehicular cases (where channel dynamics change much faster) is "left for future works."
  • Environment Specificity: The analysis is focused on outdoor 2 GHz OFDMA networks; applying this to indoor environments or much higher mmWave bands might require adjustments to the underlying path-loss and vegetation models.

Summary updated

Abstract

In this letter, we introduce the concept of Rician K-factor-based radio resource and mobility management for fifth generation (5G) ultra-dense networks (UDN), where the information on the gradual visibility between the new radio node B and the user equipment (UE)-dubbed X-line-of-sight (XLOS)-would be required. We therefore start by presenting the XLOS service probability as a new performance indicator; taking into account both the UE serving and neighbor cells. By relying on a lognormal K-factor model, a parametric expression of the XLOS service probability in a 5G outdoor UDN is derived, where the link between network parameters and the availability of an XLOS condition is established. The obtained formula is given in terms of the multivariate Fox H-function, wherefore we develop a fast graphical processing unit-enabled MATLAB code. Residue theory is then applied to infer the relevant asymptotic behavior and show its practical implications. Finally, numerical results are provided for various network configurations, and underpinned by extensive Monte-Carlo simulations.
BibTeX
@article{chergui2019rician,
  author = {Hatim Chergui and Mustapha Benjillali and Mohamed-Slim Alouini},
  title = {{Rician $\{K\}$ -Factor-Based Analysis of XLOS Service Probability in 5G Outdoor Ultra-Dense Networks}},
  journal = {IEEE Wireless Communications Letters},
  year = {2019},
  url = {https://openreview.net/forum?id=cV96mCQE2Ro},
}
Figure