基于度量优先树搜索的高性能Gbps软输出MIMO检测器研究

In next-generation wireless communication systems, gigabit data rate transmission can be achieved by increasing modulation order and employing multiple antennas at both the transmitter and the receiver. Soft-output multiple-input multiple-output (MIMO) detector is of great importance for the deployment of commercial MIMO systems. However, soft-output MIMO detection algorithm has posed the bottleneck for algorithm research and integrated circuits design of future wireless communication systems, because the huge complexity brings the difficulty of hardware implementation. In this project, we focus on the modified metric-first detection algorithm which has the best efficiency among the tree search algorithms for MIMO systems, and plan to propose a low complexity and high performance parallel algorithm and architecture for metric-first tree search detector. Firstly, according to the search ability both at horizontal and vertical orientation, we investigate the parallel sub-space detection in breath orientation, and the parallel level detection in depth orientation. Secondly, in order to decrease the number of visiting nodes during the tree search, we design the complex node enumeration approach and channel-adaptive early termination strategy for the detection algorithm. Then we will propose the efficient binary heap based storage mechanism and storage architecture to solve the bottleneck for the VLSI architecture design. After the cross-layer optimization between algorithm and VLSI architecture, we will develop Gbps soft-output MIMO metric-first tree search detector with near optimum performance and low hardware consumption instead of using multiple detectors. The expected research achievements can provide the theoretical basis for the development of next-generation wireless communication system algorithms and chip design.

下一代无线通信系统在发送和接收端都采用多根天线和高阶调制技术来支持超过1Gbps的传输速率，其走向实用化的关键是软输出MIMO检测器。然而软输出MIMO检测算法复杂度高，硬件实现难度大，成为无线通信与微电子领域迫切需要研究的难点。本项目以树搜索算法中具有最高效率的度量优先检测为研究对象，研究高性能、Gbps吞吐率的度量优先软输出检测器。根据度量优先检测同时在横向和纵向进行搜索的特点，探讨子空间并行和层间并行算法；然后通过低复杂度复数域节点枚举方法和信道自适应提前终止策略的研究，降低访问节点的个数；并针对度量优先检测VLSI设计的瓶颈，重点研究基于二叉堆的高效节点存储机制与存储结构。通过算法和VLSI结构的协同优化，在不依靠多个检测器叠加的方式下，设计近似最优检测性能、低面积消耗的Gbps软输出MIMO检测器。研究成果可望为下一代无线通信系统算法研究和芯片设计提供新思路并奠定理论基础。

在自然科学基金青年项目支持下，本课题组分步骤、分层次进行了软输出度量优先检测的低复杂度并行算法与VLSI架构的研究。提出了一种度量优先树搜索检测的并行化策略，基于此方法，设计了一种并行软输出堆栈算法用于4x4天线、64-QAM调制方式的MIMO系统信号检测，并完成了VLSI硬件实现。在算法层面，课题组提出了基于多个局部堆栈的并行树搜索策略，针对该并行树搜索策略分别使用了无效节点删除，复数域节点列举和叶子节点并行展开等方法，并修正了软输出的近似公式，提高了搜索效率和检测性能。在VLSI架构层面，采用多级流水线架构，基于查找表和并行计算结构进行设计，提高了检测器数据吞吐率。通过对整个检测器进行并行流水线的VLSI架构设计，并采用65nm的CMOS工艺库进行后端实现，实现结果表明提出的检测器在检测吞吐率和硬件效率上优于其他文献中的度量优先检测器。. 进一步，课题组将检测算法与信道解码级联为迭代接收机来提升接收机性能，基于开环和闭环模式的MIMO信号检测器，提出了多模检测器架构的概念；基于同频干扰存在下的迭代MIMO接收机，提出了利用信号检测器和信道译码器之间的迭代交换软信息来逼近最优的性能；基于低复杂度的LDPC译码算法和高性能的译码器，提出了一种计算最小值和近似的次小值的算法。这些研究成果可望下一代无线通信系统算法研究和芯片设计提供新思路并奠定理论基础。. 自获自然科学基金青年科学基金项目支持以来，申请人以通信作者发表SCI收录期刊论文7篇，其中IEEE期刊论文3篇；发表国际会议论文2篇；申请发明专利4项。通过课题的实施，课题组共培养博士研究生1名，硕士研究生4名。