基于图协同过滤模型的D2D协作缓存策略

针对设备到设备（D2D）缓存中基站信号覆盖范围有限导致的难以获得足够数据来预测用户偏好的问题,提出了一种基于图协同过滤模型的D2D协作缓存策略。首先,构建图协同过滤模型,通过多层图卷积神经网络捕捉用户-内容交互图中的高阶连通信息,并利用多层感知机学习用户和内容之间的非线性关系来预测用户偏好。其次,为了最小化平均访问时延,综合考虑用户偏好和缓存时延收益,将缓存内容放置问题建模为马尔可夫决策过程模型,设计基于深度强化学习的协作缓存算法进行求解。仿真实验表明,与现有的缓存策略相比,所提缓存策略在不同的内容种类、用户密度和D2D通信距离参数下均取得了最优的性能效果。     

移动边缘计算中基于联邦学习的视频请求预测和协作缓存策略

随着互联网社交平台的崛起和移动智能终端设备的普及,自媒体短视频、直播等视频业务蓬勃发展,人们对高质量视频服务的需求也急剧上升。与此同时,连接到核心网络的大量智能设备增加了回程链路的负载,传统的云计算难以满足用户对视频服务的低延迟要求。移动边缘计算(MEC)通过在网络边缘部署具有计算和存储能力的边缘节点,通过在更靠近用户的边缘侧提高计算和存储服务,降低了数据传输时延进而缓解了网络阻塞。因此,基于MEC架构,该文充分利用网络边缘资源,提出了基于联邦学习的视频请求预测和视频协作缓存策略。通过利用多个边缘节点对提出的深度请求预测模型(DRPN)视频请求预测模型进行联邦训练,预测视频未来的请求情况,然后量化缓存内容所带来的时延收益并协作地以最大化该时延收益为目的进行缓存决策。该文分析了真实数据集MovieLens,模拟了视频请求缓存场景并进行实验。仿真结果表明,相比于其他策略,所提策略不仅能有效降低用户等待时延,在有限的缓存空间中提高内容多样性,从而提高缓存命中率,降低缓存成本,还能降低整个系统的通信成本。     

基于时延优化的蜂窝D2D通信联合用户关联及内容部署算法

针对蜂窝网络传输性能及基站(BS)缓存能力受限,多用户内容请求难以满足用户服务质量(QoS)需求等问题,该文提出一种蜂窝终端直通(D2D)通信联合用户关联及内容部署算法。考虑到位于特定区域的多用户可能对于相同内容存在内容请求,该文引入成簇思想,提出一种成簇及内容部署机制,通过为各簇头推送热点内容,而簇成员基于D2D通信模式关联簇头获取所需内容,可实现高效内容获取。综合考虑成簇数量、用户关联簇头、簇头缓存容量及传输速率等限制条件,建立基于用户总业务时延最小化的联合成簇及内容部署优化模型。该优化问题是一个非凸的混合整数优化问题,该文运用拉格朗日部分松弛法,将原优化问题等价转换为3个凸优化的子问题,并基于迭代算法及Kuhn-Munkres算法联合求解各子问题,从而得到联合成簇及内容部署优化策略。最后通过MATLAB仿真验证所提算法的有效性。     

面向蜂窝网络的D2D多播通信的分簇和中继选择方法

传统蜂窝网络中,信道衰减的随机性和不确定性导致小区边缘用户的接收性能很差,尤其是面向视频传输等速率要求较高时其弊端更加凸显。D2D通信因其配置灵活性可作为传统蜂窝网络架构的有利补充,能有效改善边缘用户的性能。该文针对D2D通信的多播传输,分析了系统最小时延成本下的中继数量和分簇算法,提出一种基于分簇和中继选择的低时延D2D多播方案。该方案可以自适应选择多播重传中的中继的数量和中继节点到基站的距离,同时给出最优的带宽资源分配机制。仿真结果表明,与其他方案相比,所提方法能有效减少系统时延,提高边缘用户体验和系统性能。     

基于用户中断概率的D2D通信干扰协调与资源分配

在LTE系统中引入设备直传(D2D)通信技术,会因为D2D用户复用蜂窝用户资源进行通信而产生同频干扰.在现有的干扰协调与资源分配研究中,都需要基站获取各个通信链路的信道状态信息(CSI),但这样无疑会增加基站的信令负担.为减小干扰与基站的信令负担,提出了一种基于用户中断概率的干扰协调与资源分配算法,首先在保证蜂窝用户正常通信的情况下,通过限制D2D用户到基站间的距离来降低干扰;其次通过遍历所有蜂窝用户的频谱资源,选择能使D2D用户的总中断概率最低的频谱资源进行复用.仿真结果表明,所提算法能够在保证蜂窝用户正常通信的情况下,明显降低D2D用户的平均中断概率,同时还能够降低基站信令负担.     

D2D协作边缘缓存系统中基于传输时延的缓存策略

为了满足5G系统低时延高可靠的需求,针对单缓存终端直传(D2D)协作边缘缓存系统,提出了一种基于传输时延的缓存策略。运用随机几何理论,将请求用户和空闲用户的动态分布建模为相互独立的齐次泊松点过程,综合考虑内容流行度、用户位置信息、设备传输功率以及干扰,推导出用户的平均传输时延与缓存概率分布的关系式。以平均传输时延为目标函数建立优化问题,提出了一个低复杂度的迭代算法,得到平均传输时延次优的缓存策略。仿真结果表明,该缓存策略在传输时延方面优于常见的几种缓存策略。     

过时CSI下全双工中继辅助D2D网络的物理层安全

中继辅助终端直通（devicetodevice,D2D）网络通过与蜂窝网络共享频谱提高D2D用户的频谱效率和蜂窝用户（cellular user,CU）的物理层安全性。为进一步改善其性能,可以在基站和D2D链路的中继节点采用天线选择以及在中继节点采用全双工技术。然而,由于存在反馈时延和移动性,用于蜂窝链路和所有D2D链路天线选择的信道状态信息（channelstateinformation,CSI）均可能是过时的,针对该场景下的物理层安全性和可靠性问题,提出一种主动窃听和过时CSI场景下基站和中继节点均采用发射天线选择的全双工中继辅助D2D网络安全模型,推导CU的中断概率、遍历容量、非零安全容量概率、安全中断概率、渐近安全中断概率的解析表达式。数值计算与仿真结果均表明,基站发射天线数、中继干扰天线数越多,CU的安全性能越好;过时的CSI会降低CU的中断性能和安全性能。     

D2D中基于社交关系的按需用户发现策略

无线多媒体业务的爆炸式增长使得D2D(Device-to-Device)通信技术成为下一代移动通信系统的研究热点。用户发现策略,作为D2D关键技术之一,旨在有效发现潜在的D2D邻居用户。基于现有用户发现策略建立D2D链路,正面临着有效性和安全性的巨大挑战。本文考虑到每个用户不可避免地处于移动社交网络中,提出了基于社交关系的按需用户发现策略。主要根据用户请求的文件、具有的社交关系以及基站反馈和更新系统用户的状态信息表,来发现潜在的D2D用户对并建立D2D链路。理论分析及仿真结果表明,该策略可以更有效地发现邻居用户,并提高了D2D链路建立的安全性。      

雾无线接入网中基于内容流行度和信息新鲜度的缓存更新策略

将边缘缓存技术引入雾无线接入网,可以有效减少内容传输的冗余。然而,现有缓存策略很少考虑已缓存内容的动态特性。该文提出一种基于内容流行度和信息新鲜度的缓存更新算法,该算法充分考虑用户的移动性以及内容流行度的时空动态性,并引入信息年龄(AoI)实现内容的动态更新。首先,所提出算法根据用户的历史位置信息,使用双向长短期记忆网络(Bi-LSTM)预测下一时间段用户位置;其次,根据预测得到的用户位置,结合用户的偏好模型得到各位置区的内容流行度,进而在雾接入点进行内容缓存。然后,针对已缓存内容的信息年龄要求,结合内容流行度分布,通过动态设置缓存更新窗口以实现高时效、低时延的内容缓存。仿真结果表明,所提算法可以有效地提高内容缓存命中率,在保障信息的时效性的同时最大限度地减小缓存内容的平均服务时延。     

蜂窝网络中D2D通信的分析与建模

设备与设备间（D2D）通信因其具有诸多优点如有效提高频谱效率、减少通信时延、降低设备能耗等而作为下一代蜂窝网络中的关键技术。本文首先分析了蜂窝网络中D2D通信的分类。其次,运用随机几何数学工具,对蜂窝网络中的蜂窝用户和D2D用户进行数学建模并使用基于距离的模式选择策略分析蜂窝网络中D2D通信的最优平均信息速率及最优的蜂窝用户与D2D用户配比。最后通过MATLAB仿真平台验证结论的正确性及可行性。     

雾无线接入网中基于Q-learning的协作式内容缓存策略

为了提高雾无线接入网（Fog-Radio Access Networks,F-RAN）的边缘缓存效率,提出一种基于用户偏好预测和内容流行度预测的协作式内容缓存策略。首先,利用主题模型中隐含狄利克雷分布（Latent Dirichlet Allocation,LDA）模型动态的预测用户偏好;其次,利用网络中不同设备之间的拓扑关系和已预测的用户偏好以在线的方式预测内容流行度的变化,然后再结合基站之间的相关度,以减少缓存内容文件的重复率;最后,以最大化缓存命中率为目标,利用强化学习中的Q-learning算法获得了最优的内容缓存策略。仿真结果表明,与其他内容缓存策略相比,该内容缓存策略能有效的提高缓存命中率。      

Cooperative caching algorithm of UAV and user in UAV-assisted cellular network

Aiming at the problem of mass data content transmission and limited wireless backhaul resource of UAV in UAV-assisted cellular network,a cooperative caching algorithm for cache-enabled UAV and user was proposed.By deploying caches on UAV and user device,the popular content requested by user was cached and delivered,which alleviated the backhaul resource and energy consumption of UAV,reduced the traffic load and user delay.A joint optimization problem of UAV and user caching was established with the goal of minimizing user content acquisition delay,and decomposed into UAV caching sub-problem and user caching sub-problem,which were solved based on alternating direction method of multiplier and global greedy algorithm respectively.The iterative way was used to obtain convergent optimization result,and the cooperative caching of UAV and user was realized.Simulation results show that the proposed algorithm can effectively reduce user content acquisition delay and improve system performance.     

Proactive content caching by exploiting transfer learning for mobile edge computing

To address the vast multimedia traffic volume and requirements of user quality of experience in the next‐generation mobile communication system (5G), it is imperative to develop efficient content caching strategy at mobile network edges, which is deemed as a key technique for 5G. Recent advances in edge/cloud computing and machine learning facilitate efficient content caching for 5G, where mobile edge computing can be exploited to reduce service latency by equipping computation and storage capacity at the edge network. In this paper, we propose a proactive caching mechanism named learning‐based cooperative caching (LECC) strategy based on mobile edge computing architecture to reduce transmission cost while improving user quality of experience for future mobile networks. In LECC, we exploit a transfer learning‐based approach for estimating content popularity and then formulate the proactive caching optimization model. As the optimization problem is NP‐hard, we resort to a greedy algorithm for solving the cache content placement problem. Performance evaluation reveals that LECC can apparently improve content cache hit rate and decrease content delivery latency and transmission cost in comparison with known existing caching strategies.     

Community oriented in‐network caching and edge caching for over‐the‐top services in adaptive network conditions to improve performance

Over‐the‐top (OTT) services such as Netflix, Amazon Prime, and YouTube generate the most dominant form of traffic on the Internet today. There is increasingly high demand for resource intensive 3D contents, interactive media, 360 media, and user‐generated contents. As the amount of contents keep increasing in multiple folds, it is important to cache contents intelligently. Caching algorithm needs to exploit in‐network caching, community‐based pre‐caching, and a combined approach. Hence, we survey CDN‐based edge caching infrastructures including OpenConnect (Netflix) and Google Edge, followed by CCN based in‐network caching. We implement and compare four different approaches for caching contents including (1) in‐network caching, (2) edge caching, (3) community‐based in‐network caching, and (4) community‐based edge caching. We run our algorithms on adaptive network conditions with different topologies, cache size, content popularity, and request arrivals in and compared the delay for all these four approaches. We verify our model by calculating important performance parameters including hop count, redundancy, and hop count variances. Hopcount is an important performance parameter as it influences the processing, queuing, and transmission delays. We focus on determining if an in‐network caching approach is any better than edge caching. We reach several conclusions. First, in most of the scenarios, community‐based in‐network caching performs the best. Second, if the cache size is lesser than 30% of the total content size then community‐based edge caching is better for less popular contents. Finally, our statistical analysis also reveals that a community‐based edge caching mechanism is least affected by varying cache sizes and dynamic user behavior, which makes it a better choice for providing Service Level Agreement.     

基于实测数据集预测用户请求行为对主动边缘缓存的影响

由于无线边缘节点的缓存空间很小，在流行度已知时主动缓存策略的性能远优于被动缓存。最近，业界开始研究在文件流行度等用户请求行为未知、需要进行预测时的主动边缘缓存，发现主动缓存依然优于被动缓存。然而，大多数工作基于合成的数据集或者在推荐系统等领域采集的开源数据集，难以反映无线用户的请求行为。本文采用一个在局部区域每秒记录用户请求视频次数的实测数据集、利用神经网络预测用户在未来短期内的个体和群体行为，基于预测的用户行为信息在宏基站或微基站进行主动缓存。研究结果表明，当采用实测数据集时，由于用户请求行为具有很强的时间局部性、甚至是猝发性，所造成的虚警、漏警和加性误差使被动缓存优于主动缓存、且在宏基站缓存时增益更大；一旦采用合成的静态数据集，主动缓存明显优于被动缓存。这意味着不能仅用加性误差刻画预测流行度的不确定性，要实现主动边缘缓存的性能增益，更重要的是降低虚警和漏警。      

Cooperative caching game based on social trust for D2D communication networks

Content sharing via device‐to‐device (D2D) communications has become a promising method to increase system throughput and reduce traffic load. Due to the characteristic of spectrum sharing in D2D network, confidentiality is becoming a key issue in content transmission. Secure communication in D2D networks is generally guaranteed by a physical‐layer security mechanism. However, this method sacrifices the system transmission rate while ensuring security. Since mobile devices are carried by humans, we can leverage their trust relations to enhance the security of communications. As much, considering the psychology structure and social attributes of mobile users, we build a multidimensional trust evaluation mechanism to evaluate the trust relationship between users, and we pick out the trusted users based on the decision‐theoretic rough sets. By sharing content only between trust users, we can enhance the security of content transmissions without relying on physical‐layer security measures. Meanwhile, content caching is now widely used to improve accessing efficiency and reduce traffic load on cellular networks. However, caching content for other users incurs additional cost, which results in selfish and noncooperative behavior in users. Considering such selfishness, we introduce a cooperative caching game based on multidimensional trust relations to motivate users to cache contents for other devices. In this game, the trust relations and physical distance between two users are considered to formulate the cost function. Furthermore, we introduce an incentive caching algorithm based on social trust to minimize the total cost in the D2D network.     

命名数据网络中基于局部请求相似性的协作缓存路由机制

该文针对命名数据网络(Named Data Networking, NDN)应答内容的高效缓存和利用问题,依据内容请求分布的局域相似特征,提出一种协作缓存路由机制。缓存决策时,将垂直请求路径上的冗余消除和水平局域范围内的内容放置进行有效结合。垂直方向上,提出基于最大内容活跃因子的路径缓存策略,确定沿途转发对应的最大热点请求区域;水平方向上,采用一致性Hash协同缓存思想,实现应答内容的局域定向存储。路由查找时,将局域节点缓存引入到路由转发决策中,依据内容活跃等级动态执行局域缓存查找,增大内容请求就近响应概率。该机制减小了内容请求时延和缓存冗余,提高了缓存命中率,以少量额外的代价换取了内容请求开销的大幅下降,仿真结果验证了其有效性。     

基于随机几何理论的流行度匹配边缘缓存策略

异构网络边缘缓存机制是解决传统回程传输链路负载过大的可靠技术之一,但已有的缓存策略往往不能与被请求数据的流行度相匹配。为了解决这一问题,该文提出一种流行度匹配边缘缓存策略(PMCP),该策略能够根据流行度参数匹配对应的文件缓存概率以最大限度提升通信可靠性并降低回程带宽压力。基站的平面位置通过随机几何建模,文件的被请求概率则通过齐夫分布建模。蒙特卡罗仿真结果表明缓存机制能够有效降低回程带宽压力,且所提出缓存策略的可靠性优于对比策略。