密集异构网络中异步分布式优化关键技术研究

Future wireless networks are expected to be ultra-dense and heterogeneous. For such kind of networks, wireless designs are faced with a lot of new challenges such as large numbers of mobile terminals, complicated wireless resources and highly dynamic network topologies. Therefore, large dimensional complex network optimizations become more and more important. Based on these facts, synchronization becomes to a luxurious and ideal assumption. In this work, we will focus on asynchronous distributed optimization for ultra-dense heterogeneous networks. Based on asynchronous distributed optimization algorithms, wireless nodes can update their parameters over different intervals in a distributed and asynchronous manner to realize global network optimization even when there are synchronization errors, delays and some failed wireless links. Based on distributed signal processing theory, distributed optimization theory and factor graph theory, we aim at building a framework for asynchronous distributed optimization algorithm designs for ultra-dense heterogeneous networks to realize network resource allocation, base station selection and transceiver designs. The proposed algorithms are expected be implementable in a fully distributed manner with low complexity and overhead. Specifically, this work consists of the following research issues: 1) Asynchronous distributed network optimization algorithm design; 2）Asynchronous distributed base station selection algorithm design; 3) Asynchronous distributed linear transceiver designs for multi-cell networks; 4) Asynchronous distributed nonlinear transceiver designs for multi-cell networks. The optimization framework investigated in this work will be an important enabling technology for future ultra-dense heterogeneous networks.

密集化和异构化是当前无线网络发展的趋势，面对密集异构网络中海量设备接入、复杂无线资源管理和动态网络拓扑变化所带来的巨大挑战，大维度复杂网络优化问题的分布式求解成为了学术界的研究热点。本课题聚焦于异步分布式优化算法设计，该类型算法允许网络中不同节点可以按照不同频率更新自身局部优化问题的参数，即便是在存在同步误差、传输时延和其它无线链路不确定性的情况下，也能通过异步消息传递算法和异步参数更替算法有效地实现全网优化。本课将基于分布式信号处理理论、分布式优化理论和因子图理论，系统地建立密集异构网络中异步分布式优化的理论框架。本课题研究内容主要包括：1）异步分布式网络优化算法设计；2）异步分布式基站选择算法设计；3）异步分布式线性协作收发信机设计；4）异步分布式非线性协作收发信机设计。本课题的研究工作对于未来密集异构网络具有重要的理论意义和应用价值。

为有效应对无线网络发展的密集化与异构化趋势，本课题针对具有海量设备、复杂无线资源和复杂网络结构的密集异构无线网络，设计了高可靠的异步分布式优化算法以解决网络中的大规模协作优化问题。首先，在分布式网络优化算法设计方面，针对分布式MIMO网络中多目标优化问题，给出了基于加权切比雪夫方法的多目标优化算法；其次，在分布式基站相关算法设计方面，针对分布式异构结构下不同的信道环境与天线配置，设计了相应的导频训练以及信道估计方案，并给出了不规则三维天线阵的信道估计算法。此外还在此基础上，总结提炼出矩阵单调优化理论，有效应用于无线信道下的多种场景；最后，关于分布式线性/非线性收发信机设计方面，利用所提出的“矩阵单调性优化”和“二次型矩阵优化”两种理论工具，给出了线性/非线性混合收发信机设计的统一性最优结构。本课题对未来密集异构网络具有重要的理论意义与应用价值，仿真结果进一步有效验证了所提出算法的性能优势及理论分析的正确性。