1月18日下午在F706举办了实验室本学期的最后一次Workshop，共有6位同学在会上进行了报告。

报告人：陈彬强
题目：D2D通信中的缓存策略优化：基于学习用户喜好
摘要：基于缓存的终端直传（D2D）可以提升网络吞吐量和用户体验。通过将文件提前缓存在用户本地，当某个用户发起请求时，附近的其他用户可以将缓存在本地的文件通过D2D链路传输给此用户。现有的文献基于文件流行度对用户端的缓存策略进行优化。文件流行度定义为文件库中的每个文件被所有用户请求的概率分布，并不能反映单个用户的兴趣，因此现有存储策略不能充分挖掘缓存所带来的性能增益。本报告将介绍通过学习得到用户喜好，进行缓存策略优化。用户喜好是指文件库的所有文件被给定用户请求的概率分布。本报告首先介绍给定用户喜好时以最大化分流比例所建立的优化问题。此问题属于NP-hard的范畴，采用贪婪算法进行求解。然后，介绍为了预测用户喜好，通过概率潜在语义分析（pLSA）对用户请求行为进行建模，通过期望最大化（EM）算法对模型参数进行求解。仿真表明，pLSA可以快速学习得到用户喜好，以及利用学习得到的用户喜好进行缓存优化带来的性能增益。  

报告人：佘昌洋
题目：超可靠地延时通信的跨层设计 (Cross-layer Design for Ultra-reliable and Low-latency Communications)
摘要：实现超可靠低延时通信是第五代移动通信的主要任务之一。超可靠低延时通信不仅仅要保证极短的端到端延时还要做到极低的丢包率。其中，端到端延时包括在无线接入网（Radio Access Network）中的上下行传输延时和排队延时，以及骨干网和核心网中的路由选择与传播延时。我们的设计主要是面向无线接入网。在无线接入网中，造成丢包的原因有以下三个方面。由于传输时间很短，信道编码的长度是有限的，因而一部分数据包不能正确解码。由于信源生成数据包是随机的，一部分数据包会因为排队延时过长而无法按时传完。考虑到信道是随机的，且发射功率是有限的，一部分数据包会在信道较差时被主动舍弃。在保障以上服务质量需求的前提下，我们设计了排队策略和传输模式。在不同排队策略和传输模式下，通过分别优化上下行传输时间和排队时间，带宽和功率分配，以及天线配置，从而最小化了超可靠低时延通信所需的系统资源或者最大化了系统可用范围。  

报告人：孙乘坚
题目：Energy Efficient Resource Allocation for Uplink and Downlink in Ultra-reliable and Low-latency Communications
摘要：Ultra-reliable and low-latency communications (URLLC) is expected to be supported without compromising the resource usage efficiency. The quality of service (QoS) requirement imposed by the end-to-end delay and the packet loss probability is so stringent that, none of the queueing delay, backhual delay, uplink (UL) and downlink (DL) transmission delay, and UL and DL packet loss probability can be ignored, which are affected by both UL and DL resources. To achieve such challenging requirement, we study how to maximize energy efficiency (EE) for URLLC under the QoS constraint by jointly optimizing the UL and DL resource allocation. Since the URLLC can be supported by low data rate which only depend on the packet arrival process, EE can be maximized by minimizing the power consumption. Power and bandwidth allocation and the number of active antennas are optimized to minimize the weighted sum of the power consumed in UL and DL transmission. Since the achievable rate with finite blocklength channel codes is not convex in transmit power and bandwidth, it is challenging to optimize the resource allocation. To circumvent such difficulty, approximation and upper bounds are introduced to the QoS constraints, with which an optimization problem is formulated, though still non-convex but with desirable property. The optimal resource allocation is then obtained by exploiting the relation among the optimization variables.  

报告人：薛一凡
题目：MIMO信道中基于多维空间嵌套的预编码设计
摘要：This talk introduces our recent work about the nested alignment method for precoding design. Our work mainly includes two parts, both of which are based on the geometric explanation of covariance matrix. In our presentation, we will first introduce the basic mathematical tools. Then the distributed precoding method will be introduced considering a specific situation. Finally, we will talk about the NOMA (non-orthogonal multiple access) technique in MIMO-BC (multiple-input multiple-output broadcast channel), where a sufficient condition is provided to achieve the capacity by nested method, and a mixed transmission scheme combining nested and orthogonal method is further introduced.  

报告人：刘娟
Title: On the Gain of Full-duplex Network Over Half-duplex Network
Abstract: Recent work has experimentally demonstrated that full-duplex (FD) can achieve nearly doubled spectral efficiency of half-duplex (HD) systems for point to point communications. This presentation rethinks the benefits of FD over HD from the viewpoint of network performance under practical bidirectional traffic demands. To this end, a flexible HD-FD hybrid transmission mode, namely XD mode, is introduced and optimized. The results show that directly replacing current HD BSs with FD BSs only provides marginal gain in practical systems, which sheds light on the ways of applying FD technology in cellular communication systems.  

报告人：黄磊
题目：基于全双工通信的异构网串行跨层干扰消除技术
摘要：在异构网络中，宏基站对微用户的跨层干扰是限制系统性能的瓶颈问题。报告将首先介绍一种基于微基站全双工通信的跨层干扰抑制技术（full-duplex based ICIC, fICIC），它可以由每个微基站分布式实施，不需要宏基站的参与，因此可以在不损失宏用户性能的前提下有效提高微用户的性能。在此基础上，我们提出了一种基于全双工的串行跨层干扰消除技术（full-duplex based successive ICI cancellation fSICIC），并进一步提出了将fSICIC技术和fICIC技术相结合的混合干扰消除方法。