To meet the ever-increasing traffic demand in modern cellular networks, deploying more BSs has imposed is a vital solution. Cellular network (re-)planning is crucial to mitigate the interference and improve the network performance whenever more base stations (BSs) are deployed or some BSs are removed. Realistic spatial modelling for the locations of BSs and accurate quantification for their spatial relationship are prerequisites for effective network planning.
In the first part of this thesis, we aim to describe the spatial structure of the BSs using two scalar measures: the density of the BSs and the amount of regularity. We investigate the use of three scalar metrics to measure the spatial relationships among the BSs in cellular networks. We propose a geometry-based scalar metric (the coefficient of variation (CoV) of the length of the corresponding edges of Delaunay triangulation) to quantify the spatial regularity of the repulsive wireless networks. This work develops new approaches for i) mapping between the internal parameters of different point processes commonly used to generate the BS locations, ii) approximating the performance of a repulsive network based on its amount of regularity, and iii) fitting point processes to the spatial deployment of BSs.
In the second part, we develop a novel stochastic geometry-based cellular network planning technique that relies on the spatial structure of the network to determine the best deployment or removal locations of the BSs. First, we apply this technique for cell deactivation during the low demand periods. More specifically, cells are switched off so that the remaining active cells are as far away as possible from each other, which maximizes the spatial regularity of the network. The results show significant energy saving and network performance enhancement. Second, we exploit this approach for the strategic densification with UAV-BSs in cellular networks: The deployment of multiple UAV-BSs in the presence of a terrestrial network where the UAV-BSs provide on-demand capacity to the end users. This study provides supply-side estimation for how many UAV-BSs are needed to support a terrestrial network so as to achieve a particular quality-of-service and also demonstrate where these UAV-BSs should hover.