Multiuser Millimeter Wave Cloud Radio Access Networks With Hybrid Precoding

This paper investigates the performance of cloud radio access networks (CRANs) for a downlink multiuser millimeter wave (mmWave) system, where randomly distributed remote radio heads (<inline-formula><tex-math notation="LaTeX">${\rm RRH}$</tex-math></inline-formula>s) supported by a central baseband unit (<inline-formula><tex-math notation="LaTeX"> $\mathrm{BBU}$</tex-math></inline-formula>) communicate with multiple mobile equipment (<inline-formula> <tex-math notation="LaTeX">${\rm ME}$</tex-math></inline-formula>s). The fronthaul and access link transmissions are implemented with mmWave frequency bands. The <inline-formula><tex-math notation="LaTeX">${\rm RRH}$</tex-math> </inline-formula>s and <inline-formula><tex-math notation="LaTeX">${\rm ME}$</tex-math></inline-formula>s are modeled as independent poisson point processes. We characterize the outage probability, average latency, and throughput of this system under essential factors, such as blockages, <inline-formula><tex-math notation="LaTeX">${\rm RRH}$ </tex-math></inline-formula> density, and path loss. Two specific <inline-formula><tex-math notation="LaTeX">${\rm ME}$ </tex-math></inline-formula> association scenarios are considered: best channel participation (BCP) and nearest neighbor participation (NNP). We derived for both scenarios, closed-form expressions of outage probability in the noise-limited case, and upper and lower bounds of outage probability in the interference-limited case. Our results show blockages and path loss to have a positive effect of decreasing outage probability. Larger antenna arrays are shown to compensate for communication degradation (outage performance and latency) with higher <inline-formula> <tex-math notation="LaTeX">${\rm RRH}$</tex-math></inline-formula> deployment, which can be considered a tradeoff between intercluster interference and <inline-formula><tex-math notation="LaTeX">${\rm RRH}$</tex-math> </inline-formula> density. Finally, we show that for the <inline-formula><tex-math notation="LaTeX">${\rm ME}$ </tex-math></inline-formula> association process, BCP is the most viable for mmWave CRAN systems due to its outperforming NNP.