基于稀疏网络编码的低时延D2D无线通信研究

Low-latency device-to-device (D2D) communication is one of the most critical requirements of the next-generation wireless communication systems. Network coding, thanks to its capability of achieving the max-flow capacity of the network without the need of feedback and/or retransmissions, has a great potential in low-latency communications. However, given the fact that the data amount is small in many low-latency communication scenarios, the cost of carrying encoding coefficients in network coded packets would significantly reduce the fraction of data payload. As a result, the end-to-end latency might be negatively affected. To address the problem, this project proposes to investigate the sparse network coding design for low-latency D2D communications from three perspectives, namely the packet-level, the cross-layer design with PHY, and the design for multi-node D2D cooperative transmissions. Firstly, the project is to understand the effects of the encoding density of network coding on the end-to-end latency performance from the packet-level perspective, based on which sparse network coding schemes for reducing the coding coefficient delivery cost are to be designed. Secondly, the project is to utilize the tools from the finite-blocklength information theory to investigate the interactions between the sparse network coding and PHY, which usually uses short-packet in the low-latency communication scenarios. Lastly, the project is to design sparse network coding and the corresponding cooperative transmission schemes for the D2D cooperative transmission scenarios, where the network topologies might have circles. The goal of the project is to devise a working sparse network coding scheme to fulfill the requirements of low-latency D2D communications, and make sufficient contributions to the low-latency communication theory.

低时延D2D通信是下一代无线通信系统的关键需求之一。由于网络编码无需反馈重传且能够达到网络最大流容量，它在低时延通信中具有巨大应用潜力。但是，在数据分组较小的低时延场景下，网络编码分组中传输编码系数的开销将明显降低分组有效载荷比例，对时延性能亦有不利的一面。针对该问题，本项目拟从分组级、跨层设计、多节点协作三个方面研究基于稀疏网络编码的低时延D2D通信设计。首先，在分组级通过研究编码密度等参数对时延性能的影响机制，设计稀疏编码方法以降低系数传输开销；随后，针对低时延通信中物理层多为短包，结合有限码长域信息论研究分组级稀疏网络编码与短码信道编码的相互影响，并进行跨层联合设计；最后，针对多节点D2D协作，研究存在环路的协作网络拓扑对稀疏编码传输时延的影响规律，设计编码和传输方法。旨在通过本项目的研究，提出一套满足低时延D2D通信需求的网络编码可行方案，推动低时延D2D通信的理论发展。

本项目研究基于稀疏网络编码的低时延D2D无线通信传输设计。稀疏网络编码在编码时，充分考虑D2D通信场景的特点，使用较小的有限域或选取数据包的子集进行编码，降低编译码的复杂度。研究以“分析->设计->应用评估”为主线，从稀疏网络编码传输的性能分析方法、D2D场景中的稀疏编码设计、D2D应用场景中网络编码传输的性能评估等三个方面开展。首先，从特定D2D场景下网络编码及其稀疏形式的传输完成时间分析方法入手，围绕常见传输模型，提出构建基于吸收马尔可夫链的网络编码传输完成时间/时延分析框架。刻画了有限域大小、参与编码的分组数目、码率等编码参数对稀疏网络编码传输时间的影响，为网络编码的设计奠定了工具基础。接下来，在性能分析的基础上，面向物理层短包（short-packet）限制、传输状态反馈等D2D场景，提出了基于跨层优化的稀疏网络编码参数设计和基于强化学习的动态环境网络编码参数自适应设计。研究表明，一方面，短包场景下优化网络编码的分组大小、编码密度、有限域大小等参数以允许中间节点进行再编码并非一个明智选择；另一方面，揭示了可以基于反馈信息设计强化学习系统，在没有环境先验知识的情况下动态地实现网络编码参数调整，显示强化学习在网络编码传输设计中具有重要应用价值。最后，研究针对蜂窝网D2D内容分发系统应用，设计相应编码传输协议，并通过原型实现或网络仿真的形式，验证了网络编码在D2D内容分发场景下的性能优越性。