Performance Analysis And Optimization On Heterogeneous Cellular Networks

Heterogeneous cellular networks (HetNets) are two-tier networks. This kind of networks introduces new communication methods into conventional cellular networks to enhance the coverage of dedicated area. They can greatly improve the spectral efficiency. Compared with the cell splitting method which is a conventional method to improve the spectral efficiency, HetNets is more flexible and cost-effective. There are a lot of types of HetNets with different kinds of enhance layer, such as relay, pico cell, femtocell, wireless local area networks (WLAN), Ad hoc networks and direct-connect-communication-mode (DCCM). In order to facilitate the description, the communication between macro base station and cellular users is denoted by cellular-communication, and both the base stations and cellular user equipment are cellular-communication-nodes (CCN). The communications of the enhance layer in HetNets is denoted by non-cellular-communication, and the communication nodes of the enhance layer including access nodes and user equipment are called non-cellular-communication-nodes (NCCN).The performance of HetNets is analyzed in this dissertation. On the one hand, the statistical characteristics of the interference in HetNets are derived based on modeling the spatial distribution of all the communication nodes. With the analytical results, the performance of DCCM enhanced HetNets and small scale base station (SSBS) enhanced HetNets are compared. The analytical results are also used to help designing the network parameters in this dissertation. On the other hand, the dynamic scheduling performance analysis method is studied. The analysis model is proposed and validated by analyzing the performance of three scheduling algorithms, i.e. round robin (RR), proportional fair (PF) and maximize carrier to interference ratio (Max-C/I).Firstly, the most serious interference scenario is considered, in which the CCNs and NCCNs fully reuse the time-frequency resource. The spatial Poisson point process (SPPP) is adopted to model the spatial distribution of users in a HetNet containing more than one macro cell. The mathematical expectation, probability density function (PDF) and cumulative distribution function (CDF) of the interference are analyzed. The CDF of the signal to interference ratio (SIR) of the NCCNs is also derived. Considering the different requirements of different types of HetNets, this dissertation configures different parameters for DCCM enhanced HetNets and SSBS enhanced HetNets, and compares the statistical characteristics of the interference in these two types of HetNets.Secondly, the network configuration and resource partition method are considered using the analytical results. On the one hand, assuming the CCNs and NCCNs fully reuse the time-frequency resource, the ranges of transmit power of NCCNs are given with limited increase in cellular outage probability. On the other hand, the analytical average interference and the receiver sensitivity degradation target of HetNets are taken into account. The area in which NCCNs must use the orthogonal resource of cellular users is calculated given a group of system parameters.Lastly, the dynamic scheduling performance analysis method of HetNets is studied based on Petri Nets (PN) model. Since the location of users is almost not changed in several seconds, the geometric signal to interference and noise ratios (SINR) is not changed. However, the fast fading coefficient changes during the time interval of several seconds. While modeling the wireless channel, the geometric SINR and the fast fading are modeled separately first. Then the two models are combined together. As a result, the wireless channel is modeled as a two-dimensional PN model. One dimension stands for the geometric SINR levels, and the other stands for the fast fading superimposing on the geometric SINR. The parameters of this two-dimensional PN model can be given based on the probability distribution of each geometric SINR level and the parameters of fast fading. The decomposition method of PN is applied in analysis. And finally, the proposed model is validated by analyzing the performance of three scheduling algorithms, i.e. RR, PF and Max-C/I.