雾计算辅助车联网中面向视频直播业务的资源分配研究  

Research on Resource Allocation for Live Streaming in Vehicular Fog Computing-Enabled IOV

作　　者：张志才 张熠宁 付芳 ZHANG Zhicai;ZHANG Yining;FU Fang(School of Physics and Electronic Engineering, Shanxi University, Taiyuan 030006, China;School of Information and Communication Engineering, Beijing University of Posts and Telecommunications, Beijing 100876, China)

机构地区：[1]山西大学物理电子工程学院,山西太原030006 [2]北京邮电大学信息与通信工程学院,北京100876

出　　处：《测试技术学报》2021年第5期450-455,共6页Journal of Test and Measurement Technology

摘　　要：提出一种车辆雾计算网络中视频直播业务的资源分配方法,通过联合优化比特率选择、用户调度和频谱资源分配,以实现在最大化视频质量的同时降低时延和视频抖动.为了降低时延和视频抖动,在效用函数的设计中将时延和比特率切换作为惩罚因子.由于网络的动态变化特性和可用的频谱资源,将上述优化问题建模为马尔可夫决策过程,采用Soft Actor-Critic深度强化学习算法获得最优资源分配策略.仿真结果证明,所提方法比现有强化学习算法具有更好的探索能力和收敛性能.This study proposes a novel resource allocation scheme for live streaming,which aims to maximize video quality while decreasing time-delays and bitrate-switches in vehicular fog computing-enabled IOV by jointly optimizing bitrate selection,vehicle scheduling,and spectrum allocation.In order to reduce delay and variance,we designed both time-delays and bitrate switches as punishment factors in utility function.Due to the dynamic characteristics of vehicular networks and available spectrum resource,we modeled the above optimizing problem as a Markov decision process.We utilized Soft Actor-Critic deep reinforcement learning algorithm to find the optimal resource allocation policy.Simulation results demonstrate the proposed scheme outperforms existing reinforcement learning algorithms in terms of exploration and convergence abilities.

关 键 词：车辆雾计算网络 资源分配 视频直播 Soft Actor-Critic算法 深度强化学习 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]