Study On Modeling And Optimization Problems In Cellular Networks

In the past three decades, the mobile communication system has achieved great development, cellular networks have developed from the first generation which can only support traditional voice service to the fourth generation which can support broadband data service in high speed mobile environment. The network structure has evolved from homogeneous networks consisting of only macro cells to heterogeneous networks consisting of macro cells, pico cells, femtocells and ect..Meanwhile, user terminals have developed form traditional mobile phones to smartphones. The mobile communication is facing a great many of theoretical and technical challenges because of these significant changes.For instance, the traditional Poisson distribution can model the voice traffic accurately but it is not accurate for the background traffic of smartphone applications; The impact of signaling traffic on early cellular networks is trivial, however, the nowadays cellular networks are facing great signaling traffic pressure arising from the background traffic of always-on applications. On the other hand, 3-dimensional scenarios, heterogenous networks and vertical beamforming are new to the cellular networks and they bring new optimization problems. Studying and resolving these problems is of significant importance for the cellular networks evolution.Therefore, the dissertation studies on the above problems and the main contributions of the dissertation are listed as follows:1. Since there is no accurate traffic arriving models for the background traffic, a new traffic arriving model is proposed in the dissertation. The traditional Poisson distribution can model the arriving of voice and packet switching traffic accurately, however, when it is applied to the background traffic there is an obvious error. Therefore, a modified Poisson distribution is proposed instead. Simulation results show that the modified Poisson distribution can be used to approximate the background traffic arrivals accurately. The main mathematical characteristics of the modified Poisson distribution is deduced in the dissertation and it is specified by three parameters. How to achieve these parameters is also provided in the dissertation when the number of users or the statistics of experimental data is given. In addition, the traditional Poisson distribution is a degenerate case of the modified Poisson distribution and the Gaussian distribution is the asymptotic distribution of the modified Poisson distribution. As a consequence, the proposed model is not only suitable for the background traffic of smartphones but also suitable for other traffic.2. Since the cellular network is facing with great signaling pressure, the dissertation studies and simulate the Radio Resource Control (RRC) state transition of smartphones in LTE networks. The formulas of the number of RRC state transition, the instantaneous number of smartphones in RRC_CONNECTED and the probability of a smartphone's being in RRC_CONNECTED are deduced. Theoretical and simulation results show that these parameters are directly affected by the inactive timer which also affect the power consumption of users. With the increase of the inactive timer, the number of RRC state transition decreases so does the signaling traffic, but the probability of a smartphone's being in RRC_CONNECTED increases so does the battery consumption. There is a trade-off between the signaling traffic and the power consumption of users. Simulation results also indicate that, compared with the Poisson distribution, the modified Poisson distribution can better fit the distribution of the number of smartphones being in RRC_CONNECTED.3. Since the high floor users suffer a poor coverage, the dissertation studies how to improve the performance of high floor users with picos equipped with uptilt antennas in 3 dimension (3D) scenarios. In addition,a fast pico antenna tilt optimization algorithm is proposed to further improve the network performance. The simulation is based on a real LTE network, the simulation results show that the mean SINR, 5%-tile SINR,mean capacity and the pico connected user ratio are significantly affected by both the antenna vertical half power beam width and tilt. Choosing appropriate parameters can significantly enhance the network performance.Moreover,the SINR's of users on different floors are analyzed in detail and the simulation results show that high floor users endure a poor coverage by Macros and the coverage of these users can greatly enhanced by picos with uptilt antennas. In addition,the base station (BS) downlink SINR, which is defined as the ratio of the serving users' received signals of the BS to the interfering users' received signals of the BS and the noise, is taken as the objective function, in this way the multiple variables non-convex and NP-hard problem is turned into multiple independent single variable problems and the algorithm complexity is therefore reduced significantly. The algorithm is further improved by adding an exponential factor on the serving users, received signals. Simulation results show that the exponential factor can impact on the average user capacity obviously. The iteration is further introduced into the algorithm, and it is shown that the algorithm can converge quickly and with the iteration the average user capacity can increase significantly. The fast pico antenna tilt optimization algorithm can dynamical improve the network performance by adjusting the tilts dynamically. It is also shown by the simulation results that compared with the 3D beam forming method, this algorithm can increase the user capacity significantly.