Modeling and analysis of HetNet interference using Poisson Cluster Processes

Abstract

Future mobile networks are converging towards being heterogeneous, owing to the co-existence of multi-tier networks within the same geographical area, including macro, pico- and femto-cells. The deployment of such networks is generally based on user demand, which is irregular and random, implying that the deployment of base stations (BSs) is random as well. As a result, analyzing the communication protocols over heterogeneous networks (HetNets) is very challenging. A popular approach is to use stochastic geometry and treat the location of the BSs as points distributed according to a spatial Point Process. Most of the related work on the interference modeling normally assumes homogeneous Poisson point process (PPP). This assumption holds when the nodes are uniformly distributed in space, such as sensor networks or ad-hoc networks. Due to geographical factors, it may be the case for mobile users to cluster around highly populated cities and the PPP assumption does not provide an accurate model for the interference in these conditions. This motivates us to find better ways to characterize the aggregate interference when the transmitting nodes are clustered following a Poisson Cluster Process (PCP). Furthermore, the BSs belonging to different tiers may differ in terms of the transmit power, the node densities, and their link reliabilities. To this end, we consider K-tier HetNets, where, by using the Laplace transform approach, we characterize the aggregate interference at a given destination as a heavy-tailed distribution. Using the derived distribution, we investigate the probability of outage and coverage for such networks. Due to some difficulty in obtaining closed-form expressions for these measures, we derive tight bounds and verify that through numerical examples. We also compare the performance of HetNets when the nodes are clustered and otherwise. We observe that using the PPP results in larger success probability, but using the clustered process results in a larger coverage probability. We also observe that there is an optimal intensity, i.e., number of nodes, that achieves the maximum coverage probability for the given SINR (signal-to-interference-plus-noise ratio) threshold.