Stochastic geometric modelling and simulation of cellular systems for coverage probability characterization

TL;DR

This paper advances stochastic geometry models for cellular networks by deriving new closed-form expressions, proposing an approximation for uplink, and revealing that coverage probability is independent of base station density, challenging prior assumptions.

Contribution

It provides a unified modeling approach for downlink and uplink, introduces novel closed-form solutions, and proves coverage probability independence from BS density under SG assumptions.

Findings

Derived two closed-form expressions for downlink coverage probability.

Developed an effective approximation for uplink coverage.

Proved coverage probability is independent of BS density.

Abstract

Stochastic geometry (SG) has been successfully used as a modelling tool for cellular networks to characterize the coverage probability in both the downlink (DL) and uplink (UL) systems, under the assumption that the base stations (BS) are deployed as a Poisson point process. In the present article, we extend this use and provide further results for interference limited and Rayleigh fading networks, culminating in a multifaceted contribution. First, we compactly model the two systems at once, allowing parallels to be drawn and contrast to be created. Also, for DL we manage to obtain two closed form expressions for two special cases. Moreover, for UL, notorious for being difficult, we develop a clever approximation that overcomes the difficulty, yielding excellent results. Additionally, we present two efficient Monte Carlo simulation algorithms, designed primarily to validate the models, but can be of great use for SG modelling of communications systems in general. Finally, we prove two theorems at odds with popular belief in cellular communications research. Specifically, we prove that under the SG model, the coverage probability in both DL and UL is independent of BS density. Based on this revelation, a plethora of results in the literature have to be re-examined to rid them of a parameter that has been proven superfluous.