实时视频多路并行传输中的码率控制与分流策略研究

With the rapid rise of industrial applications such as UAV and autopilot, the needed real-time video quality is getting higher and higher. Concurrent multipath transmission technology can effectively improve the quality of real-time video transmission in the next generation of heterogeneous wireless networks. So, it is of great significance to research concurrent multipath real-time video transmission technology. Firstly, the determination of the target video frame bits to match the concurrent multipath transmission bandwidth is the key problem of rate control in concurrent multipath real-time video transmission. To solve the problem, a concurrent multipath transmission cache theory model is built in this project. The relationship between the target video frame bits, path capacity and packet allocation scheme under the requirements of bandwidth utilization and low latency is revealed and the upper and lower bounds of target video frame bits is obtained in this project. Secondly, how to further optimize the rate control under the influence of packet allocation scheme is the second key problem to be solved in this project. To settle this problem, a joint rate control and packet allocation optimization model with the objective of video quality smoothness and low-delay is established, and a low complexity flow allocation and rate control algorithm is proposed in the project. Finally, the proposed theory and method are verified on both the simulation experiment and practical platform. This project is expected to improve the real-time video transmission quality in terms of performance such as delay and rate-distortion, and provide theoretical basis and reference for further study and applications.

随着无人机、自动驾驶等产业应用的快速兴起，实时移动视频质量要求越来越高，在下一代异构无线网络下利用多路并行传输技术可以有效提升实时视频传输质量，相关研究具有重要意义。在实时视频多路并行传输中，如何决策视频帧目标比特数匹配多路并行传输带宽是当前码率控制的关键难题。项目通过建立多路并行传输缓存区理论模型，揭示在满足带宽利用率和低时延要求下的视频帧目标比特数与链路性能、分流策略之间的关系，获得视频帧目标比特数上下界。在此基础上，如何考虑分流策略的耦合影响进一步优化码率控制是本项目解决的第二个关键问题。项目建立以视频质量平滑和低时延为目标的分流策略和码率控制的联合优化模型，并提出具有低复杂度的分流策略和码率控制算法。最后，项目通过仿真实验和实际系统平台测试相结合的方式验证所提理论和方法。本项目有望在时延、率失真等性能方面进一步提升实时移动视频的传输质量，为相关技术研究和应用提供理论依据。

本项目以异构无线网络下的实时视频传输中的码率控制与分流策略优化问题为研究对象。具体的，本项目通过研究多路并行传输中的缓存模型，揭示了在满足低时延和视频质量要求下的视频帧目标比特数与链路性能、分流策略之间的关系，获得了视频码率控制中的帧级别的目标比特数上下界。基于此，本项目在获得上下界的基础上，进一步提出了基于缓存上下溢模型的视频帧码率控制方法，实现了帧级别的码率控制。并且，本项目考虑了长期收益模型，通过分析缓冲区状态，利用长期优化理论对视频帧进行码率控制，避免视频帧码率控制因网络状态波动造成影响。在此基础上，本项目针对多径时变信道的特点，进一步提出了可变目标时延策略，通过设置动态可变目标时延优化视频码率和分流策略，保障视频传输质量。结果表明，通过设置动态可变目标时延，在实验条件下，帧丢失率可以从6.65%降低到2.06%。. 本项目进行了相关成果的落地转化，将超低延时实时视频多路并行传输技术应用于5G智能多媒体路由器中，通过多路并行传输技术实现对实时视频低时延高可靠的传输，该产品原型已应用于福建东南汽车DX9车型上实现远程驾驶。