Research On Low Delay Congestion Control Technology Of Mobile Video Live Broadcast

At present,although the real-time video telephone service is widely used,the quality of experience(QoE)of users is still unsatisfactory.In addition,the previous research work is based on simulation experiments or small-scale real experiments to study this problem,while the research based on a large real application data is rare.T his project obtained more than 7.7tb of live video network data from this paper's partner(a major live video service provider in China).Analyzing these data,it is found that the mismatch between video coding layer and transmission layer results in the phenomenon of low QoE,such as video jam and image blur.To solve this problem,an intelligent congestion control algorithm based on reinforcement learning algorithm A3C is designed and implemented.The algorithm extracts the high-level features of video coding layer and transmission layer from the massive historical network data.Through training,the algorithm can coordinate the linkage between video coding layer and transmission layer at runtime,they are no longer independent.In addition,in order to speed up the algorithm learning process,this paper designs and implements a live video simulator,which can simulate the 24-hour live video process in 3.6 minutes.Finally,the performance of the algorithm is evaluated,and it is found that the performance of the algorithm is about 10%higher than other video transmission solutions.