Information-Theoretic Capacity of Clustered Random Networks

We analyze the capacity scaling laws of clustered ad hoc networks comprising significant inhomogeneities in the node spatial distribution over the area. In particular, we consider the class of networks in which nodes are distributed according to a doubly stochastic shot-noise Cox process, which allows to model a wide variety of inhomogeneous topologies. For this class of networks, we derive information theoretic upper-bounds to the capacity, identifying six operational regions. We also provide constructive lower bounds by devising, for each region, an optimal communication strategy to achieve the maximum network throughput. The performance of our communication schemes match, in terms of scaling exponent, the theoretical upper-bounds.