面向可伸缩视频编码传输的DDPG无人机服务增强机制  

UAV service enhancement mechanism deep deterministic policy gradient for scalable video coding transmission

作　　者：闫俊杰 朱亚新 冯艳茹 刘汉永 邓钧忆 王欢 YAN Junjie;ZHU Yaxin;FENG Yanru;LIU Hanyong;DENG Junyi;WANG Huan(School of Electronic Engineering,Guangxi University of Science and Technology,Liuzhou 545006,China;Shanxi College of Applied Science and Technology,Taiyuan 030062,China;Dazu Power Supply Company of State Grid Chongqing Electric Power Company,Chongqing 402360,China)

机构地区：[1]广西科技大学电子工程学院,广西柳州545006 [2]山西应用科技学院,山西太原030062 [3]国家电网重庆市电力公司大足供电分公司,重庆402360

出　　处：《电信科学》2023年第8期69-81,共13页Telecommunications Science

基　　金：国家自然科学基金资助项目(No.62106053);广西自然科学基金资助项目(No.2022GXNSFBA035642);广西科技计划项目(No.2021AC19167);广西中青年教师基础能力提升项目(No.2022KY0325,No.2021KY0364)。

摘　　要：无人机(UAV)因其机载存储、通信和计算能力成为未来空天地一体化网络的重要组成部分。然而,现有无人机辅助的边缘网络研究总是更多地从网络的视角出发,缺少从用户视角出发带来的变革。为此,从用户角度出发,提出一种面向可伸缩视频传输的DDPG无人机服务增强机制。首先,结合无人机提出一种基于可伸缩视频编码(SVC)的弹性视频传输方法,提高用户的差异化体验。其次,以最大化用户接收增强层视频的数量为目标,提出一种基于深度确定性策略梯度(DDPG)算法的无人机路径规划设计,保证UAV对热点区域增强覆盖的有效性。仿真结果表明,所提出的算法与深度Q网络(DQN)算法、最短路径(SP)算法相比,在不同的用户分布下得到的增强层数量平均分别可以提高47.9%、76.4%,该研究达到了热点区域覆盖增强和用户差异化体验的效果。Unmanned aerial vehicles(UAV)are an important component of future air-space-ground integrated net-work due to their onboard storage,communication,and computing capabilities.However,existing research on UAV-assisted edge networks often focuses more on the network perspective and lacks consideration of user perspectives and their requirements.Therefore,a UAV service enhancement mechanism based on deep deterministic policy gradient(DDPG)for scalable video coding(SVC)transmission was proposed from the user’s perspective.Firstly,an elastic video transmission method based on SVC was proposed in conjunction with UAV to improve differentiated user experiences.Secondly,a UAV trajectory planning design based on the DDPG algorithm was proposed to max-imize the number of users receiving enhanced layer videos and ensure effective coverage enhancement by UAV in hotspot areas.Simulation results show that compared with both the deep Q network(DQN)algorithm and shortest path(SP)algorithm under different user distributions,the proposed algorithm can increase the average number of en-hanced layers received by 47.9%and 76.4%,respectively.This study successfully achieves improved coverage in hotspot areas while also providing differentiated user experiences through its proposed methods.

关 键 词：深度确定性策略梯度 无人机 可伸缩视频编码 路径规划 

分 类 号：TN929.5[电子电信—通信与信息系统]