There is a need for simple, yet realistic models for telecommunication network architectures capturing the spatial distribution of the elements of the network and having a limited number of parameters.
This can be obtained via probabilistic models based on point processes and stochastic geometry. The probabilistic setting reflects the network variability in time and space, this is particularly relevant for mobile communications. The modeling approach consists in representing the components of a network (nodes, communication links, service zones) as a family of random objects (point patterns, graphs, tessellations), i.e. as realizations of stochastic processes. Several network characteristics can then be expressed as functionals of these processes and thus will depend only on their distribution parameters.
This approach often allows for an explicit evaluation of network characteristics as capacity, throughput, cost, etc., and for direct parametric optimization.