Research On Optimization Technology Of VR Video Transmission Based On SDN

Virtual reality(VR)video has attracted more and more attention with its immersive experience,and its requirements for high bandwidth and low latency have brought great challenges to network transmission.In order to meet the Quality of Experience(QoE)requirements of VR video users under limited resources,real-time and effective network resource management strategy is required.The next generation mobile networks 5G can offer the network and the computation resources according to service requirements,which will be the communication technology for the VR industry.In 5G architecture,the introduction of Software Defined Networking(SDN)extracts the control function of network device and centralizes them into controller.The controller has the ability of global network resource management.It can actively allocate resources for VR video to optimize transmission performance.In order to meet the quality of service requirements of VR video services,this paper uses the centralized management ability of SDN and the requirements of VR video transmission on network conditions to reasonably allocate network resources.VR video service identification is the premise of optimized network resource allocation.In this paper,we use the flow statistical features combined with machine learning to identify VR video services.Through theoretical analysis and experimental analysis,we analyze the statistical features of VR video and Non-VR video,and then propose a set of flow statistical features to distinguish them.After that,the flow counter in OpenFlow switch is used to periodically read the basic flow statistics information.We calculate the flow statistics features according to sampling information.After human labeling,four machine learning algorithms are used to train and test.The experimental results show that our proposed flow statistical features can achieve an accuracy of 99%for VR video service identification.After identifying VR video service,we design an appropriate VR video routing scheme according to the requirements of VR video transmission on network conditions.The routing scheme is used to guide the packet forwarding of VR video,so as to allocate network resources reasonably and effectively.In this paper,a QoE-driven Fine-grained routing(QFR)scheme based on SDN has been proposed.It can improve the performance of VR video transmission without modifying the Dynamic Adaptive Streaming over HTTP(DASH)client.The core of QFR is the route calculation algorithm and the route allocation strategy.The route calculation algorithm is a two-stage adaptive routing algorithm.In the first stage,by means of an improved Dijkstra algorithm,the algorithm calculates k paths with the shortest delay.In the second stage,the k paths with the shortest delay are ranked according to the predicted QoE of each path.In addition,tile-based VR video provides a prerequisite for fine-grained routing scheduling.Through differentiated routing of Field of View(FoV)video streaming and Non-FoV video streaming,we develope a fine-grained route allocation strategy.The route allocation strategy determines how to allocate the sorted k paths with the shortest delay according to the residual bandwidth.Comparative evaluation of QFR is conducted to verify its preponderance over several existing routing schemes,in terms of download bitrate and QoE of VR video.