Interference Coordination in Heterogeneous Networks: Stochastic Geometry Based Modelling and Performance Analysis

Recently data traffic has experienced explosive increase with the proliferation of wireless devices and the popularity of media-based free services. The academic and industry of mobile communications have predicted an estimated $1000$x increase in traffic volume for the forthcoming 5G networks. This traffic explosion stimulates the deployment of heterogeneous networks (HetNets) with small cells (SCs) underlying in the traditional macrocells, which has been considered as a promising technique to contribute to the $1000$x traffic capacity gain. Initially, licensed spectrum bands are expected to be used in SCs, thus the SC deployment introduces the cross-tier interference between SCs and macrocells, which degrades the downlink signal to interference plus noise ratio (SINR) of user equipments (UEs) severely, especially for the edge UEs in a ultra-densely deployed scenario. To alleviate this cross-tier interference between SCs and macrocells, unlicensed spectrum bands are advocated to be used in SCs. Specifically, with the aid of carrier aggregation, the $5$ gigahertz (GHz) unlicensed band has become an option for SCs in the Long Term Evolution (LTE)-Unlicensed (LTE-U) scheme, but the $5$ Ghz unlicensed band has already been used by WiFi networks. Thus downlink cross-tier interference also occurs between LTE-U and WiFi networks. Accordingly, downlink cross-tier interference is inevitable no matter licensed or unlicensed spectrum band (i.e., 5 GHz) is used in SCs, and interference coordination schemes, such as further enhanced inter-cell interference coordination (FeICIC) for macrocells and SCs, and Licensed Assisted Access (LAA) for WiFi networks and LTE-U networks, have been proposed to mitigate these cross-tier interferences. In this dissertation, we mainly focus on the modelling and performance analysis of HetNets with the aforementioned two interference coordination schemes (i.e., FeICIC and LTE-LAA) under the stochastic geometry framework. Firstly, as the configuration of reduced power subframe (RPS)-related parameters was not well investigated in a two-tier HetNet adopting RPSs and cell range expansion (CRE), we derive the analytical expressions of the downlink coverage probability and rate coverage probability in such a HetNet. The optimal settings for the area of macrocell center regions, the area of SC range expansion regions, and the transmit power of RPSs for maximizing the rate coverage probability are analysed. As compared with the rate coverage probability in the two-tier HetNet with almost blank subframes (ABSs), which is proposed in the previous version of FeICIC, i.e., the enhanced inter-cell interference coordination (eICIC), the results show that ABSs outperform RPSs in terms of the rate coverage probability in the two-tier HetNet with the optimal range expansion bias, but lead to a heavier burden on the SC backhaul. However, with static typical range expansion biases, RPSs provide better rate coverage probability than ABSs in the two-tier HetNet. Secondly, the conventional FeICIC scheme ignores the potential of RPSs being adopted in both tiers of a two-tier HetNet without CRE, which is envisioned to improve the SINR level of edge UEs in both tiers. Accordingly, we study the downlink coverage probability and rate coverage probability of a two-tier HetNet applying with our proposed scheme. The results reveal that adopting RPSs in both tiers not only improves the coverage probabilities of edge UEs, but also increases the rate coverage probability of the whole two-tier HetNet. Thirdly, in both previous works, strict subframe alignment (SA) was assumed throughout the whole network, which is difficult to maintain between neighbouring cells in reality. Consequently, we propose a novel subframe misalignment (SM) model for a two-tier HetNet adopting RPSs with SM offsets restricted within a subframe duration, and analyse the coverage probability under the effects of RPSs and SM. The numerical results indicate that the strict SA requirement can be relaxed by up to $20\%$ of the subframe duration with a loss of below $5\%$ in terms of the downlink coverage probability. Lastly, since stochastic-geometry-based analysis of the coexisting LTE-LAA and WiFi networks, which adopt the carrier-sense multiple access with collision avoidance (CSMA/CA) as the medium access control (MAC) scheme and share multiple unlicensed channels (UCs), was missing, we analyse the downlink throughput and spectral efficiency (SE) of the coexisting LTE-LAA and WiFi networks versus the network density and the number of UCs based on the Matern hard core process. The throughput and SE are obtained as functions of the downlink successful transmission probability (STP), of which analytical expressions are derived for both LTE-LAA and WiFi UEs. The results show that the throughput and SE of the whole coexisting LTE-LAA and WiFi networks can be improved significantly with an increasing number of accessible UCs. Based on the numerical results, insights into the trade-off between the throughput and SE against the number of accessible UCs are provided. All the derived results have been validated by Monte Carlo simulation in Matlab, and the conclusions observed from the results can provide guidelines for the future deployments of the FeICIC and LTE-LAA interference coordination schemes in HetNets