超密集异构网络小区间干扰消息传递抑制方法

Hyper-dense heterogeneous network is one of the key technologies to achieve the demand of high capacity, high spectral efficiency and high effectiveness in the further generation of wireless communications. However, the co/cross-tier inter-cell interference is an important problem impacting the capacity and service quality of wireless systems. Since the current inter-cell interference suppression techniques have high computation complexity, they are efficient in the static or semi-static single networks, but cannot be applied to dynamic dense networks. The recently proposed message-passing-based inter-cell interference suppression methods need a large amount of collaborative overhead between the base stations, and the methods only consider the co-tier interference. This project chooses maximizing system throughput and joint a posteriori probability density function as the cost function. After analyzing the co/cross-tier interference mechanism, we will establish the factor graph of both uplink and downlink. The combined message update rules will also be studied based on minimizing regionalized free energy with normalization and marginalization constraints, while some non-Gaussian messages, which result high computation complexity and much collaboration overhead, will be approximated by minimizing Kullback-Leibler divergence. Considering Hyper-dense heterogeneous network are dynamic, we also investigate new resources allocation strategies based on game theory for message passing scheduling to reduce coordination overhead of the messages interchange between base stations. Finally, a unified message passing framework will be proposed for transceiver to design joint co/cross-tier inter-cell interference algorithms on both uplink and downlink transmission of networks. The applicants expect that research results can provide new technologies and methods to solve the problems of co/cross-tier inter-cell interference in the hyper-dense heterogeneous networks for the next generation wireless communication system.

超密集异构网是未来实现无线通信系统高容量、高谱效和高功效的关键技术之一。该网络的小区间同层/跨层干扰问题是制约系统容量和通信质量的重要因素。传统的小区间干扰抑制技术计算复杂度过高，对静态或半静态的简单环境较为有效，但不适于动态变化的密集网络环境；现有基于消息传递的小区间干扰抑制方法协作开销过大，且仅解决了同层干扰问题。本项目以最大化系统吞吐量和联合后验概率密度函数为优化目标，分析同层/跨层干扰机理，建立上下行链路多层干扰的因子图模型；设计基于最小化区域自由能的联合消息更新规则，并针对非高斯连续型消息引起的计算复杂度过高问题，提出基于最小化KL散度的消息近似新方法；在网络动态环境下，针对基站间协作开销过大给出基于博弈论的消息调度新策略；建立收发信机统一的消息传递框架，提出系统上下行链路干扰抑制新算法。项目的研究成果为解决超密集异构网络的同层/跨层小区间干扰问题提供新的技术和方法。

超密集异构网络是下一代移动通信的关键技术之一，本项目针对超密集异构网络小区间干扰和大规模通信系统的多用户检测等问题进行较全面的系统性研究。针对超密集异构网中的大规模MIMO系统，通过分析上行链路簇稀疏-非平稳空间相关信道特征，构建了一种基于迪利克雷过程的信道概率模型，并提出了联合消息传递信道估计算法。该方法可在低导频开销条件下有效提高大规模MIMIO系统上行链路信道估计精度，进而在异构网中实现基于位置感知的导频功率分配。本项目设计了基于注水算法的上行链路导频发射功率分配算法，减轻了系统内用户间的干扰，提高了全网系统的信干噪比和系统容量。针对网络的下行链路，对MIMO-OFDM接收机进行因子图建模，通过添加辅助变量进行因子图变换，使得其能适用更合理的消息更新规则，提出了基于协作更新规则的Co-GBI接收机算法，并验证了所提协作消息更新规则的合理性。另外，针对多小区基站协作系统中用于解决超密集异构网中同层干扰问题，结合系统因子图模型统一架构设计实现了接收机协作分布式迭代干扰抑制算法；在此基础上提出了基于GAMP算法的分布式RZF预编码符号设计方法，能够以较快的收敛速度达到与RZF预编码相同的平均吞吐量，且随着系统规模的增大，所提方案能有效降低计算复杂度，从而在实际通信系统中的应用价值会更高。最后考虑到异构网可能采用毫米波通信系统，功耗问题是制约毫米波通信系统发展的关键。针对低精度量化下的毫米波通信系统，提出了基于PBiGAMP的联合接收算法，用较低的计算复杂度达到了较优的检测与估计性能，为低精度量化下迭代接收算法设计提供了参考。本项目研究工作的完成为超密集异构网干扰抑制、大规模MIMO接收机设计等问题开拓了新的思路，对超密集异构网的成熟应用和5G通信系统的商用化部署提供了理论支撑。