基于广义因子图的非正交多址接入技术研究

In this project, the generalized factor graph is introduced as an analytical tool to construct the optimized transceiver of the non-orthogonal multiple access and improve its performance. First, this projects plans to propose a novel superimposed modulation scheme at the transmitter, which is based on the optimization of the modulation base vector with the help of the weighted edges of generalized factor graph, a novel successive iterative multiuser detection technique, which exploits the difference of the reliability of multiple users, a novel detection performance evaluation technique, which is based on the SINR calculation after detection, to construct the analytical framework of the non-orthogonal multiple access. Then, to improve the multi-antenna transmission performance of non-orthogonal multiple access, this project maps the multiple-antenna link into the mulriple edge in generalized factor graph and plans to propose a joint optimization technique of the multiple antenna precoding matrix and the modulation base vector, a hybrid message passing multiuser detection algorithm with low complexity, and a user-data-aided channel estimation technique. Finally, this project plans to propose the channel adaptive transmission strategy in non-orthogonal multiple access, which includes the adaptive coded modulation scheme, the user selection and resource scheduling method and the low-cost feedback codebook design. The research output of this project tends to provide theoretical support and valuable references for the evaluation of the non-orthogonal multiple access in 5G system.

本项目引入广义因子图作为分析工具，构建非正交接入条件下传输链路优化方法，提高非正交接入性能。首先，针对非正交接入，采用叠加调制，构造广义因子图中的权重连接边，探索基于调制基向量优化的发送信号设计方法，采用串行检测方法，优化迭代多用户检测性能，提出基于检测后信干扰比估计的多用户性能评估方法，构建非正交接入收发设计理论；其次，将多天线非正交接入映射为广义因子图上的多重连接，提出多天线预编码和调制基向量联合优化的发送信号设计方法，采用混合消息传递，设计低复杂度多用户检测算法，并利用数据译码结果辅助信道估计，实现非正交多天线传输链路优化；最后，提出自适应非正交调制编码方法，构建非正交链路用户选择和资源调度策略，设计低冗余反馈码本，实现非正交多址的自适应传输优化。通过本项目的研究，为5G非正交多址技术的发展提供重要的理论支持和参考。

本项目主要研究基于广义因子图的非正交多址接入技术。非正交技术是实现大规模接入的物理层关键技术，本项目重点对非正交接入中的码本设计、多用户检测算法、多用户功率分配和信道反馈技术进行了研究。主要成果包括以下三个方面：1、基于广义因子图，给出了基于叠加调制的非正交接入信号模型，提出了成对差错概率最小化和离散输入连续输出互信息容量最大化两种码本设计准则，设计了新的码本映射方法，仿真结果表明，本项目设计的码本性能优于多种现有码本；2、设计了具有低复杂度的多用户检测算法，对于稀疏码分多址接入的场景，采用多重判决方法减小了误差传播，提高了检测性能；对于交织器多址接入的场景，采用软符号串行干扰抵消的方法降低了用户间干扰，并能够采用低复杂度频域实现的方法，改善了误比特率性能；对于多用户检测与译码器的联合迭代接收机，提出了对检测结果的缩放因子进行优化的方法，加快了迭代的收敛速度；3、以深度学习神经网络作为工具，设计了免迭代的多用户功率分配注水方法，提高了算法的并行度；将注意力机制引入到多层神经网络中，用于信道反馈压缩，降低了信道反馈冗余。在本项目的研究中，研究内容始终追踪学术界与工业界的前沿，本项目的码本设计成果已经写成了提案，提交到了6G标准化工作组中，具有持续发展的潜力。