Characterizing spatial patterns of base stations in cellular networks

The topology of base stations (BSs) in cellular networks, serving as a basis of networking performance analysis, is considered to be obviously distinctive with the traditional hexagonal grid or square lattice model, thus stimulating a fundamental rethinking. Recently, stochastic geometry based models, especially the Poisson point process (PPP), have been attracting an ever-increasing popularity in modeling BS deployment of cellular networks due to its merits of tractability and capability for capturing non-uniformity. In this study, a detailed comparison between common stochastic models and real BS locations is performed. Results indicate that the PPP fails to precisely characterize either urban or rural BS deployment. Furthermore, the topology of real data in both regions are examined and distinguished by statistical methods according to the point interaction trends they exhibit. By comparing the corresponding real data with aggregative point process models as well as repulsive point process models, we verify that the capacity-centric BS deployment in urban areas can be modeled by typical aggregative processes such as the Matern cluster process, while the coverage-centric BS deployment in rural areas can be modeled by representative repulsive processes such as the Strauss hard-core process.