Abstract
In 5G Ultra-Dense Networks (UDNs), traditional signal strength metrics like Reference Signal Received Power (RSRP) often fail to capture the nuanced quality of radio links required for advanced services like high-precision localization. This paper proposes "X-line-of-sight" (XLOS) service probability as a new key performance indicator (KPI), utilizing the Rician K-factor to quantify link visibility. We derive a closed-form mathematical framework for XLOS service probability using stochastic geometry and the multivariate Fox H-function. Our results, validated through extensive Monte-Carlo simulations and empirical data, demonstrate how network density and neighbor monitoring influence service availability. A GPU-accelerated software tool is also provided to facilitate the practical application of these complex analytical models in network optimization.
Introduction: Beyond Binary Visibility
As 5G deployments move toward ultra-dense configurations, managing radio resources becomes increasingly complex. A critical factor in network performance is the "visibility" of the radio link—whether it is direct Line-of-Sight (LOS), obstructed (OLOS), or Non-Line-of-Sight (NLOS). Historically, visibility has been treated as a binary or categorical state. However, in reality, visibility exists on a spectrum.
This paper introduces the concept of XLOS (X-line-of-sight) to describe this gradual visibility. By using the Rician K-factor—the ratio of direct path power to scattered power—as a metric, the authors provide a way to quantify the probability that at least one base station in a user's vicinity provides a sufficiently "visible" link. This is particularly vital for tasks like switching between sub-6GHz and millimeter-wave (mmWave) frequencies, where a high-quality LOS path is a prerequisite for performance.
Methodology: Mathematical Rigor and Computational Efficiency
The study employs a multi-faceted approach to model 5G environments:
- Stochastic Geometry: Base station distributions are modeled as a Poisson Point Process (PPP), representing the random yet uniform nature of real-world deployments.
- Lognormal K-factor Model: A distance-based model incorporates environmental shadowing, reflecting how the Rician K-factor degrades with distance and obstacles.
- Multivariate Fox H-Function: To achieve an exact "closed-form" solution for service probability, the researchers utilized the Fox H-function. This generalized function allows for the representation of complex probability distributions across multiple tiers of a network.
- GPU Acceleration: Recognizing the computational intensity of Fox H-functions, the authors developed a GPU-enabled MATLAB routine using linear optimization to automate the definition of the Mellin-Barnes contour, a necessary step for calculating these complex integrals.
The model was calibrated using empirical parameters (antenna heights, beamwidths, path-loss exponents) from a measurement campaign in New Jersey, ensuring that the theoretical framework aligns with physical realities.
Results and Validation
The primary contribution of this work is the derivation of the first closed-form expression for XLOS service probability in multi-tier networks. The analytical formulas were tested against Monte-Carlo simulations across three network scenarios: Low-Density Networks (LDN), Heterogeneous Networks (HetNet), and Ultra-Dense Networks (UDN).
As shown in Figure 1, the analytical results perfectly match the simulation data. The figure also highlights the "Power of Neighbors": by increasing the size of the monitoring set ($M$), the probability of maintaining a high-quality LOS link significantly improves, even as the required visibility threshold ($K_{th}$) increases.
Asymptotic Analysis and Practical Implications
The researchers identified two key asymptotic regimes that simplify the complex general model for specific network conditions. These regimes provide engineers with "rules of thumb" for network planning based on the ratio of the visibility threshold to the network density parameters.
In a true UDN environment, the probability of having a high-quality LOS path approaches unity, making it an ideal candidate for mmWave services. Conversely, in LDNs, the probability drops sharply, indicating that such networks may struggle with services requiring high link visibility.
Conclusion
This research establishes the Rician K-factor-based XLOS service probability as a robust KPI for 5G and beyond. By moving beyond simple signal power, network operators can better trigger handovers and optimize localization services. Future work will focus on extending this framework to high-speed vehicular (V2X) scenarios, where rapid mobility introduces dynamic changes to the K-factor that require further modeling.
Summary updated