Research On Mechanisms For QoE-driven VR Video Transmission Over Wireless Networks

Virtual reality(VR)is a three-dimensional artificial environment generated by a computer.It represents a brand-new way of interaction,and is commonly regarded as one of driver applications for 5G services.The most popular application of VR at present utilizes 360-degree video to create an immersive artificial world in head mounted display(HMD).However,due to integrated characteristics of high resolution,high frame rate,low latency and multiple degree of freedom et al,VR video cannot be directly supported by current wireless transmission technology;Even in 5G network optimized for single scenario(e.g.,enhanced mobile broadband,e MBB),Qo E will not be acceptable without tuning of transmission.Thus in foreseeable future,VR video has to be transmitted through wireline or displayed locally,which critically restricts large-scale applications of VR technology,especially in complicated environments like intelligent manufacturing,and calls for effective approaches to VR video transmission over wireless networks.Researches on VR video transmission over mobile wireless network have following insufficiencies:(1)Little attention was paid to uplink transmission,although user generated uploading and sharing is more and more popular;(2)For wireless RA and source coding bitrate assignment,different view regions contribute differently to user's Qo E but were not differentially treated;(3)Current buffering strategies are not smart enough for VR video downlink adaptive streaming;(4)Novel structure of 5G network was not fully leveraged in VR transmission designing.To address aforementioned issues and for the sake of Qo E optimization,in this thesis,VR video transmission over mobile wireless network is in depth studied regarding source coding,wireless RA,buffer management,cooperative multicast and unicast.The main contributions are outlined as follows:(1)For the first time,the mechanism for VR video uplink transmission over LTE is investigated,targeting at the increasing popularity of user generating VR video for sharing.Based on the traditional Qo E model and the feature that viewers have various quality expectation on video tiles,a quality of content(Qo C)model is defined,and a novel scheme for VR video delivery over cellular network based on Qo C is also proposed.Moreover,the RA and tile-based source coding bitrate assignment is formulated as a joint optimization problem.Furthermore,two solutions based on greedy and approximate convex approaches are presented.Simulation results show that the proposed scheme is feasible and effective,and can achieve higher utility under limited bandwidth.(2)Furthermore,mechanism for VR video uplink transmission in 5G heterogeneous cloud radio access network(H-CRAN)is researched,combining small cell feature of H-CRAN and UE mobility.First,a content-sensing based joint source coding and resource allocation scheme is presented,in which the importance of different VR video regions(i.e.Tiles)is sensed according to the saliency detection.This scheme takes into account g-NB group RA,remote radio head(RHH)/g-NB association,sub-channel assignment,power allocation,and tile encoding rate assignment,and is formulated in an optimization problem;then a three stage heuristic algorithm is proposed to solve the problem.Simulation results show that the proposed scheme ensures the total utility of VR video uploading,and the multi-stage heuristic algorithm is an effective solution with low complexity and faster convergence.(3)To improve intelligence of buffering strategy,VR video downlink adaptive transmission mechanism based on viewport prediction and hierarchical buffer is researched.First,viewport prediction based on historical trajectories and similarity between cross user motion behaviors is explored.Based on hierarchical buffer strategy,an adaptive algorithm for UE by considering bandwidth,buffer status and predicted viewport is proposed,and hierarchical buffer adaptive scheduling problem is formulated into an optimization problem,then is solved separately according to different current buffer status.Simulation results show that the proposed approach can effectively improve the viewport prediction accuracy;and proposed adaptive algorithm based on hierarchical buffer and viewport prediction can maintain stable Qo E in fluctuating bandwidth condition.(4)Combining the heterogeneous feature of H-CRAN and dual connectivity supported for UEs,for the first time,a cooperative multicast and unicast with viewport prediction(CMU-VP)mechanism for VR video streaming is proposed: video content with the highest bitrate is proactively cached at mobile edge computing(MEC)server,then is pushed on corresponding g-NB or RHHs after real-time transcoding;a basic version of video is transmitted to all users through the g-NB in a multicast session,and tiles of enhanced-version in predicted viewport are transmitted to each viewer in a unicast session through its stationed RHH.The scheduling problem of the proposed mechanism is formulated as an optimization problem,and two near-optimal solutions are presented.Simulation results show that the CMU-VP ensures better Qo E under constrained bandwidth compared with pure multicast or unicast mechanisms,and the proposed near-optimal solutions can efficiently solve the problem with low complexity and comparable performance.In summary,VR video uplink and downlink transmission in LTE and 5G networks are systematically formulated,solved and verified by simulations in this thesis.In particular,optimization for VR video uplink transmission is thoroughly studied;A multicast cooperative transmission mechanism combined with MEC is proposed.This thesis' s work on exploring mechanisms of VR transmission over wireless network devotes valuable contributions to push VR technology into practice.