摘要
针对实际网络中网络吞吐率的变化有很大程度的随机性,引入了分析模型预测控制(model predictive control,MPC)方法。通过对历史数据规律进行归纳总结并且将历史数据估计方法作为深度预测模块的补充,提出了一种多层感知的深度预测模块。相对于同期最佳模型,所提模型能够提高11%的预测准确度。将所提模型在真实网络中进行实验验证,结果表明,所提供的方法能够有效提升视频质量并降低重缓冲概率,从而提升用户体验。
As to the significant randomness of countermeasure changes above network throughput in actual networks,the analytical model predictive control(MPC)method was introduced.By summarizing the regularity of historical data and supplementing the deep prediction module with historical data estimation methods,a multi-layer perception deep prediction module was proposed.Compared with state of the art(SOTA)models,the model reported herein can improve the prediction accuracy by 11%.The pro-posed model was experimentally verified in a real network.The results show that the proposed method can effectively improve the video quality and reduce the probability of rebuffering,thereby improving user experience.
作者
陈铤沛
张书豪
袁一平
向文馗
杨力军
唐东明
CHEN Tingpei;ZHANG Shuhao;YUAN Yiping;XIANG Wenkui;YANG Lijun;TANG Dongming(School of Computer Science and Engineering,Southwest Minzu University,Chengdu 610041,China)
出处
《中国科技论文》
CAS
2024年第2期193-199,共7页
China Sciencepaper
基金
中央高校基本科研业务费专项资金资助项目(校2021118)
四川省科技计划项目(2023YFSY0049)。