Abstract:
This thesis focuses on modelling future high capacity cellular networks operat-
ing in millimetre wave (mmWave) frequency bands. An analytical approach has
been proposed for network modelling rather than system-level simulations. The
increasing complexity of cellular networks due to their continuous evolution has
made conventional system-level simulations time-consuming. Recently stochas-
tic geometry has been recognised as a tractable and e cient analytical tool for
network modelling and quantifying key performance metrics.
The stochastic geometry approach involves representing base stations and user lo-
cations as a spatial point process. It allows us to investigate trends in performance
related to base station density and user distribution. Network performance char-
acteristics mainly depend on the nature of the point process model. Therefore it
is important that the chosen model should closely represent practical cellular net-
work deployments. The goal of this research is to adopt an accurate spatial point
process model to design cellular networks. Previously the Poisson Point Process
(PPP) model has been used as a way of representing these practical cellular net-
works. However, this model has some de ciencies, and the Binomial Point Process
(BPP) overcomes these de ciencies to some extent.
This research uses the BPP model to analyse the performance of high capacity
cellular networks operating at millimetre wave frequencies. By considering the
unique features of mmWave communications, such as di erent channel character-
istics for the line-of-sight (LOS) and the non-line-of-sight (NLOS) links and high
directional antenna gains provided by antenna arrays, we have developed stochas-
tic geometry frameworks to analyse both downlink and uplink coverage probability
for a reference receiver.
The BPP is de ned in a nite region with a xed number of nodes. Initially, a
simple nite disk region has been considered as the network's geographical region,
and then later, the analysis has been extended to a more general arbitrary region.The results indicate that the performance di erence between an arbitrary shape
and a disk shape is almost negligible for a centrally located user. Therefore a
simple disk region can be an acceptable shape to model nite cellular systems
for a centrally located user. Furthermore, we have also analysed the performance
trends representing the impact of antenna arrays, the base station (BS) density,
and LOS and NLOS links in the coverage analysis.