上海交通大学学报

上海交通大学学报（自然版）

• 自动化技术、计算机技术 • 上一篇    下一篇

基于POMDP的可伸缩视频流优化决策调度

范凤军1,邹君妮1,汪敏1,熊红凯2   

  1. （1.上海大学 通信与信息工程学院， 上海 200072； 2.上海交通大学 图像通信与信息处理研究所， 上海 200240）
  • 收稿日期:2009-05-31 修回日期:1900-01-01 出版日期:2010-03-30 发布日期:2010-03-30

Scheduling Optimization for Scalable Video Streaming Based on Partially Observable Markov Decision Processes

FAN Fengjun1,ZOU Junni1，WANG Min1，XIONG Hongkai2   

  1. (1.School of Communication and Information Engineering, Shanghai University, Shanghai 200072， China; 2.Institute of Image Communication and Information Processing, Shanghai Jiaotong University, Shanghai 200240， China)
  • Received:2009-05-31 Revised:1900-01-01 Online:2010-03-30 Published:2010-03-30

摘要： 针对可伸缩视频流在无线通信中的数据包调度问题，提出了一种基于部分可观测马尔可夫决策过程（POMDP）的决策调度策略，在用户状态不确定或部分可观测条件下，采用POMDP建立了完整的最优化调度模型，以实现视频接收质量的最佳化.仿真实验结果表明，与传统的调度算法相比，该方法有效提高了视频流的平均峰值信噪比(PSNR).随着调度规模的扩大，其性能能够逐渐逼近用户状态完全确定的理想马尔可夫决策过程（MDP）模型.

关键词: 可伸缩视频编码, 部分可观测马尔可夫决策过程, 优化调度, 无线广播

Abstract: Packet scheduling remains a critical issue for the transmission of scalable video streaming over wireless broadcasting channel. To achieve the optimal receiving video quality, a packet scheduling algorithm based on the optimization strategy of partially observable Markov decision processes (POMDP) was proposed. It adopts POMDP to establish a complete optimization scheduling mathematical model on the condition that the state of the user is uncertain or partially observable. Compared with traditional packet scheduling algorithms, the experimental results show that the proposed scheme can efficiently improve the average PSNR. With the increase of the number of users, the performance of the proposed POMDP model asymptotically approaches to the ideal MDP model in which the state information is completely observable.

中图分类号: