抖动
计算机科学
溪流
实时计算
调度(生产过程)
数据流挖掘
计算机网络
电信
数学优化
机器学习
数学
作者
Zhuang Tian,Wujun Yang,Juan Guo,Yuanzheng Cheng
标识
DOI:10.1145/3573942.3573992
摘要
For the problem of using time-aware shaping combined with credit shaping-based hybrid traffic scheduling to achieve efficient transmission of hybrid streams, an improvement of CBS shaper based on deep reinforcement learning is proposed. By setting appropriate reward values, this mechanism greatly satisfies the different requirements of time delay for various types of audio and video streams and greatly reduces the delay and loss rate of BE streams without affecting the Time Triggered Flow (TT). The experimental results show that the algorithm improves the delay of AVB streams under the worst conditions. At the same time, the minimum delay and jitter are greatly improved compared with the classical model.
科研通智能强力驱动
Strongly Powered by AbleSci AI