随着社会和经济的快速发展，移动用户的数据业务呈爆炸式增长，宽带移动通信网络为了实现低时延、高连接数、高流量密度、高移动性等需求与愿景，引入超密集组网(Ultra-Dense Networks，UDN)、全双工(Full Duplex，FD)、D2D (Device-to-Device)通信等技术。新技术的引入能够提高系统容量、增加频谱效率，但也带来了严重的干扰问题。如何有效地消除自干扰，降低用户间、小区间干扰是宽带移动通信网络的关键技术问题。本文从干扰对象角度出发，针对用户自干扰、用户间干扰和小区间干扰，研究宽带移动通信干扰管理问题，对干扰进行由点到线、再到面的优化管理。 首先针对全双工系统同时同频收发信号时会受到自身信号干扰的问题，对全双工系统中自干扰消除技术进行深入研究。利用谱估计算法对带内自干扰信道进行线性化建模，使用遗传算法优化自适应滤波器的参数，设计实现基于最小均方误差(Least Mean Square，LMS)滤波的改进算法，并对算法进行仿真试验，结果表明，本文提出的算法相比于基于训练序列信道估计的信道估计算法具有更强的抗干扰能力和更好的收敛性。 其次针对D2D通信系统中大量用户在高速移动蜂窝边缘同时切换将导致拥堵和模式接入的问题，对D2D通信系统中干扰协调技术进行深入研究。在高密度和高移动性的D2D场景中，提出了一种基于D2D分簇的速度感知切换算法，利用D2D多播系统和D2D簇解决大量UE在小区边缘同时切换的问题，该方案能够有效缓解广播信令风暴的压力，降低中断概率和切换延迟，提高用户QoS。另外针对D2D设备建立连接后的模式选择问题，研究了三种D2D模式，即传统蜂窝模式和直接通信模式，以及D2D设备中继模式。对于大量D2D UE群体，采用基于复制动态的算法来设计分布式D2D模式选择算法。仿真结果表明所提出的方法相比于max-SINR贪婪算法、基于距离选择算法和随机选择算法方案，实现了最高效用。 最后在UDN组网中，针对复杂网络结构的不同层干扰管理问题，开展超密集组网下干扰避免技术的研究。一方面在构造基于图论的超密集组网系统模型的基础上，提出了一种基于冲突因子图的资源分配算法。该算法分解了全局资源冲突图，利用因子图降低复杂度，解决了PRBs资源块分配和功率分配的多目标问题。仿真结果表明，所提算法不仅提升了边缘吞吐量，也提高了整网平均吞吐量，并在提高频谱效率的同时降低了用户间干扰。另一方面提出了一种基于图信号处理的强化学习功率分配算法，利用图信号处理工具对网络干扰进行分析，得到全网的干扰分析结果作为改进强化学习的状态设计，指导功率分配。仿真结果表明本文所提算法在保证峰值速率的同时提高了边缘用户的吞吐量，从而提高了整网吞吐量。 今后进一步研究方向可从结合器件的非线性特性进一步优化自干扰消除算法，以及在复杂密集的网络系统中对移动性加以分析。 关键词：干扰管理；干扰消除；干扰协调

 With the rapid development of society and economy, the data service of mobile users has been exploded. To realize the demands and vision for low delay, high connection number, high density and high mobility, we will introduce some new technologies such as ultra-dense networks (UDN), full duplex (FD), D2D (Device-to-Device) communications, etc. The introduction of these advanced technologies could increase system capacity and spectrum utilization, but also brought some serious and complex interference problems. How to effectively eliminate self-interference and reduce inter-user and inter-cell interference are key technical issues in broadband mobile communications. In this paper, the interference management problems of user self-interference, inter-user and inter-cell interference are studied, and the local-to-global optimized interference management are carried out. Fist, for the problem of self-interference that the FD transceiver will be interfered by its own signal when receive interferce from When transmits and receives signals at the same frequency simultaneously. We study the self-interference cancellation technology in the full-duplex system. The spectral estimation algorithm is used to linearize the in-band self-interference channel. The genetic algorithm is used to optimize the parameters of the adaptive filter. The improved algorithm based on Least Mean Square (LMS) filtering is designed and implemented. Simulation results show that the improved algorithm proposed in this paper has a stronger anti-interference ability and better convergence. Then, for the problem of massive D2D users simultaneously handover on cell edge with high-speed will lead to congestion, the interference coordination technology in the D2D communications is deeply studied. We propose a speed-aware joint handover approach that leverages D2D multicast and D2D clusters. The proposed solution can effectively alleviate the pressure of broadcast signalling storm, and improve the handover QoS in terms of interruption probability and handover latency. In addition, we attempt to address the challenge of a large D2D UE population in mode selection. We employ a gametheoretic approach and formulate an evolutionary game, based on which a distributed device-controlled algorithm is designed for D2D mode selection. Using three mode for uses to selecte which are cellular mode, direct reuse mode and relay mode. The simulation results show that the proposed scheme can achieve the highest utilities compared to the max-SINR scheme, the distance based scheme and the random scheme. At the last, for the problem of different layer interference management of complex network structure, the research on interference avoidance technology under ultra-dense networks is carried out. On the one hand, we construct a system model under ultra-dense networking. Then according to the complex resource allocation problem in the networking, a resource allocation algorithm based on conflict factor graph is proposed. The resource conflict graph is constructed by using resource conflicts, reduced the complexity by using the factor graph, and solved the multi-objective problem of resource allocation and power allocation. The simulation results show that the proposed algorithm not only improves the edge throughput, but also improves the average throughput of the whole network, and improves the spectrum efficiency while reduce the interference between users. On the other hand, an reinforcement learning power allocation algorithm based on graph signal processing is proposed. The graph signal processing tool is used to analyze the network interference, and the interference analysis result of the whole network is obtained as the state to optimal reinforcement learning power allocation. The simulation results show that the proposed algorithm improves the throughput of edge users while ensuring the peak rate, thus improving the throughput of the whole network. Key Words: Interference management; Interference cancellation; Interference coordination