随着互联网、物联网和移动通信技术的发展，蜂窝通信网络承载的数据量呈指数增长的趋势，导致具有粗放性、静态性和局部性等特点的传统蜂窝网络提供的服务无法满足用户日益增长的QoS需求。针对上述问题，本文针对新的软件定义蜂窝网络(Software Defined Cellular Network， SDCN)架构，利用全局的网络视角和集中式的控制逻辑，通过实时、动态地获取网络状态信息，研究了不同网络部署中LTE与WiFi共存、地面基站与空中基站共存等场景下的资源管理机制，为未来蜂窝网络中的资源管理提供了有效的解决方案，具有重要意义。本文的研究工作如下： 首先，在LTE和WiFi共存的场景下，深入研究了SDCN中授权频段和非授权频段的资源管理机制，提出了一种面向SDCN网络的资源分配算法。该算法采用集成的毫微微-WiFi基站(Integrated Femto-WiFi Base Station，IFW fBSN)，允许智能终端设备可以同时使用授权频段(通过LTE接口)和非授权频段(通过WiFi接口)来缓解频谱资源的短缺。同时，算法考虑具有LTE接口和WiFi接口的智能设备(Smart Device，sDevice)、只具有Win接口的WiFi设备(WiFi Device，wDevice)，以及宏蜂窝设备(Macrocell Device，mDevice)等三种设备类型。利用SDCN控制器的全局化视图，本文通过最大化sDevice和wDevice的加权效用总和，来分配授权频段和非授权频段的资源，并使fBSN对mDevice的干扰低于阈值。进一步，采用并行的交替方向乘子法(Alternating Direction Method Of Multipliers，ADMM)和凸优化算法来求解建模的组合非凸问题。数值结果表明，本文提出的资源分配算法显著地提高了网络中所有设备的平均吞吐量和平均效用。通过新设计，sDevice和wDevice的平均吞吐量可以提高41.6%。 其次，在地面基站和无人机(Unmanned Aerial Vehicle，UAV)共存的场景下，考虑UAV辅助技术来提高应急场景中的用户性能，并利用SDCN的集中式控制逻辑提出了适应于未来网络的无人机部署与资源分配算法。首先，通过优化的三维(Three Dimensional，3D)UAV部署和用户关联来最大化无人机蜂窝用户的效用总和。其次，提出了优化的3D无人机部署和资源分配算法，该算法通过优化3D UAV的部署、用户的关联和无人机的传输功率，最大化了关联的用户数和传输功率的网络收益效用，同时保证用户的QoS需求大于阈值。经过数学分析，本文将混合整数的组合非凸问题降维为两阶段的子问题，并分别采用二分法和凸凹优化(Convex-Concave Procedure，CCCP)算法进行求解。仿真结果表明，与其他无人机部署方案相比，所提的3D UAV部署和用户关联算法提高了网络的吞吐量和效用，最大增益可以达到36.4%；3D UAV部署和资源分配算法提高了网络收益效用，最大增益可以达到35.08%。 最后，针对弹性部署的无人机受限于回程链路的问题，本文进一步研究了回程受限情况下的多维度地空资源分配。该算法利用SDCN控制器的全局化视角和集中式的控制逻辑，通过最大化用户的效用总和，联合优化了3D无人机的部署，用户的关联以及频谱资源的分配。进一步，本文提出了交替最大化算法，将混合整数的组合非凸问题分解为三个并行的子问题块，并分别采用连续凸优化(Successive Convex Optimization， SCO)技术和改进的ADMM进行求解。理论分析和仿真结果验证了该算法的收敛性。与传统的无人机部署方法相比，所提算法显著提高了用户的吞吐量和效用，其中，最大的吞吐量增益为74.9%。 关键词：软件定义蜂窝网 资源分配 用户关联 凸优化 无人机

 With the development of Internet, Internet of Things and mobile communications, the traffic carried by cellular networks shows an exponential growth trend, which results in that the services provided by traditional cellular systems with rough, static and local features cannot meet the QoS demands of users. To solve these problems, based on the software-defined cellular network (SDCN) architecture, which utilize the global view of the network, the thesis studies the resource management in the scenarios of LTE/WiFi coexistence and ground base station (GBS)/ Unmanned Aerial Vehicle (UAV) coexistence under different network deployments by using the real-time and dynamic network status information. The proposed methods provide an effective solution in the resource management for future cellular networks, which are of great significant. The contribution of this thesis are as follows: Firstly, in the scenario of LTE/WiFi coexistence, this thesis studied the resource management of licensed and unlicensed band in SDCN. A resource allocation algorithm for future network is proposed. The proposed algorithm utilizes the integrated femto-WiFi base station (IFW fBSN) to allow smart devices use both the licensed band (via LTE interface) and the unlicensed band (via WiFi interface) to alleviate the spectrum shortage. The proposed algorithm also considers the smart devices (sDevices) with both cellular and WiFi interfaces, the WiFi devices (wDevices) with WiFi-only interfaces, and the macrocell device (mDevice). Using the global view of the SDCN controller, a weighted utility maximization problem is proposed to allocate resources of licensed and unlicensed bands, and also keeps the interference from fBSN to mDevices below predefined thresholds. Furthermore, an alternating algorithm is used to solve the complex non-convex problem. Numerical results show that the proposed re- source allocation algorithm significantly improves the average throughput and average utility of all devices in the network. Throughput gains as large as 41.6% for the average of all sDevices and wDevices are achieved by using the new designs. Secondly, in scenario of GBS/UAV coexistence, this thesis considered the UAV-assisted technology to improve user performance in the emergency situations, and proposed the UAV deployment and resource allocation algorithm by utilizing the centralized control logic of SDCN. First, a drone cell users' aggregate utility maximization problem is proposed by optimizing the 3D UAV placement and the user association. Then, a 3D UAV placement and resource allocation algorithm is proposed by optimizing the 3D UAV placement, user association and UAV transmission power. The proposed algorithm maximizes the network revenue utility of the associated number of users and the transmitted power of UAV, and also keeps QoS demand of users greater than the threshold. After mathematical analysis, the mixed integer multidimensional non-convex problem was reduced to a two-stage optimization algorithm. The bisection search method and the Convex-Concave Procedure (CCCP) algorithm were used to solve the problem. Simulation results show that the proposed 3D UAV placement and user association algorithm improves the network throughput and utility compared with other traditional placement methods, and the maximum throughput gain can reach up to 35.4%; while the proposed 3D UAV placement and resource allocation algorithm improves the network revenue utility compared with other existed placement methods, and the maximum gain is 35.08%. Finally, in view of the problem that the performance of flexibly deployed UAV is limited by the backhaul link, this thesis further studied the multi-dimension air-ground resource allocation with the wireless backhaul. A utility maximization problem for multi-UAV enabled SDCN is proposed. The proposed problem jointly optimize the 3D UAV deployment, user scheduling and association, and the spectrum resources allocation. Furthermore, an alternating maximization problem is proposed to solve the mixed integer combined non-convex prob- lem by decoupling it into three alternating subproblem blocks which are solved via the successive convex optimization (SCO) technique and the modified alternating direction method of multipliers (ADMM) in the proposed algorithm. Theoretical analysis and simulation results verify the convergence of the algorithm, and extensive numerical results verify the superiority of the algorithm. Compared with the traditional UAV placement methods, this algorithm significantly improves the throughput and utility of the overall users, in which the maximum throughput gain is up to 74.9%. KEY WORDS: Software-defined cellular networks, Resource Allocation, User Association, Convex Optimization, UAV