A cellular network is usually modeled and analyzed as a Poisson point process (PPP). However, small cells are ordinarily clustered around hotspots in urban areas, which cannot be accurately modeled with a PPP. This letter presents an analysis of the downlink coverage of small cells that are modeled with a Matern cluster process (MCP). A new accurate formula for the probability generating function of an MCP is derived. We also propose a new distribution function for the desired distance (i.e., between the user and serving small cell) based on the spherical contact distribution function of the Boolean model. The above two formulas can be used to obtain a closed-form expression for the downlink average coverage probability of small cells. Simulation results confirm the accuracy of our derived formulas. Our derived MCP-based coverage is accurate and can be used to guide the small cell configuration of a cellular network.

We analyze the delay performance of small cell networks operating under random scheduling (RS) and round robin (RR) protocols. Based on stochastic geometry and queuing theory, we derive accurate and tractable expressions for the distribution of mean delay, which accounts for the impact of random traffic arrivals, queuing interactions, and failed packet retransmissions. Our analysis asserts that RR outperforms RS in terms of mean delay, regardless of traffic statistic. Moreover, the gain from RR is more pronounced in the presence of heavy traffic, which confirms the importance of accounting fairness in the design of scheduling policy. We also find that constrained on the same delay outage probability, RR is able to support more user equipments (UEs) than that of RS, demonstrating it as an appropriate candidate for the traffic scheduling policy of internet-of-things (IoT) network.

Recently, research attempts have been made to jointly model spatial and temporal randomness in cellular networks. However, realistic modeling and analysis of dynamic queuing interactions in heterogeneous networks (HetNets) still need to be investigated. In this work, we establish a realistic spatiotemporal model for a K-tier HetNet. Particularly, we model the locations of base stations (BSs) from each tier as an independent Poisson point process (PPP). Packet arrivals are modeled as independent Bernoulli processes. To reveal actual scenario, the effect of bias association is considered. We also allow users to have various packet arrival rates to account for different traffic loads. We capture the coupling relations between transmission success probability and cell active probability. Accurate expression for the mean packet throughput of the network is derived. Our analytical results provide the way to design optimal bias association factors to maximize the mean packet throughput of the HetNet, and also help design BS densities for certain user density in order to achieve a given mean packet throughput level.

Heterogeneous cellular networks (HCNs) are usually modelled as a Poisson point process (PPP) due to mathematical tractability. However, in urban areas, small cells are more likely to cluster around hotspots rather than be uniformly distributed. Meanwhile, in reality, small cell locations are not completely independent of the macro base station (MBS) locations. To capture both features, the authors propose a new modelling method considering inter‐tier dependence for a two‐tier HCN, where MBSs follow a PPP, small cells follow a Matern cluster process and are not allowed to be deployed inside the MBS exclusion regions (i.e. within a prescribed distance from each MBS). Approximate expressions of the average coverage probability and the area spectral efficiency (ASE) are obtained, respectively. Simulation results confirm the accuracy of the derivations. It is observed that the model can achieve a good tradeoff between coverage and ASE, and if the small cell density in this model is kept identical to the HCN without inter‐tier dependence, both coverage and ASE can benefit from involving the inter‐tier dependence in the HCN.