Stochastic topology cognition in heterogeneous networks

To enable ultra-dense random deployments of fem-tocells toward the fifth generation (5G) cellular networks, effectively autonomous interference mitigation has been regarded as the key for providing unprecedented levels of network capacity and coverage. From recent research and engineering efforts, achieving this goal relies on two promising technologies of cognitive radio (CR) to fully utilize the time-frequency domain radio resources, and stochastic geometry to capture physical distances among nodes (and thus the power domain radio resources). To fully exploit radio resources in all domains of time, frequency, and power, existing schemes impose a strong assumption on the availability of node densities for autonomous resource allocations. To practically validate this assumption, femtocells need to estimate densities of femtocell base stations. However, subject to limited coverage ranges of femtocells, each femtocell can only acquire limited samples insufficient for an accurate estimation. As a result, femtocells need to cooperate with each other. One possible cooperation scheme is selecting a femtocell as fusion center to collect partial information of node densities from all femtocells. This scheme is, however, vulnerable to the fusion center failure and is costly. To provide a robust and efficient scheme, in this paper, we integrate technical merits of CR and stochastic geometry to propose a stochastic network topology cognition framework for femtocells. Our scheme adopts a local processes/estimations of partial information to achieve a competitive performance as compared with that of a unrealistic fusion center, to practically enable a full radio resource utilization in all domains.