Research On Video Caching And Transmission Strategy For Edge-cloud Collaboration In Internet Of Vehicles

With the development of 5G communication,intelligent transportation systems(ITS)and mobile edge computing(MEC),video services play an important role in Internet of Vehicles,commercial entertainment and intelligent transportation applications.However,the fast mobility of vehicles,the dynamic characteristics of wireless networks,and the limited spectrum resources and storage resources make the quality of experience(Qo E)of users not guaranteed when watching videos.Based on scalable video coding(SVC)technology,this thesis studies the video caching and transmission strategy in the Internet of Vehicles.The specific work is as follows:(1)An unequal error protection algorithm of expanding window fountain code with feedback is proposed to solve the problem of packet loss during SVC video transmission in the Internet of Vehicles scenario.According to the popularity of the video,the video files are cached in cooperation with the cloud and edge servers,and the video is forwarded from the cloud to the roadside unit(RSU)for encoding or directly encoded by the RSU according to the vehicle request.The source data packets of different layers of the SVC video requested by the vehicle are subjected to unequal error protection.When the vehicle receives encoded packets,the base layer data packets are preferentially decoded.After the base layer data is successfully decoded,single-bit feedback information is introduced,and the RSU only encodes and sends the enhancement layer data after receiving the feedback information.The simulation results show that,compared with the expanding window fountain code without feedback,the proposed scheme can improve the probability of successful decoding of enhancement layer data,and reduce the amount of RSU data packets sent and the transmission time when decoding is successful.(2)A resource optimization scheme is proposed for the video service in the vehicle-to-infrastructure(V2I)communication system in the Internet of Vehicles.All videos are stored in the cloud according to the popularity of the video content,and some videos are dynamically stored in the edge server in the RSU,and sent collaboratively according to the requests of different users.Considering factors such as limited communication resources and storage resources,video popularity,and picture quality,a utility function to measure the user’s Qo E is designed,and it is used as an optimization target to achieve reasonable resource allocation,dynamic adjustment of cached video content in the MEC server and optimized selection of video layers for each user.Resource optimization based on the Deep Asynchronous Advantage Actor-Critic algorithm to maximize the utility function.The simulation results show that compared with other algorithms,the algorithm has better performance in improving video quality,reducing latency,and the reward value is about 0.025 and 0.1325 higher than the PG algorithm and the no-learning scheme,respectively.(3)A communication resource optimization scheme is proposed for the intelligent transportation video service in the vehicle-to-vehicle(V2V)communication system in the Internet of Vehicles.The V2 I link sends the video of road conditions to the RSU,which can cache the video on the edge server or forward it to the cloud.Vehicles on the road utilize the V2 V link for road condition video transmission.This thesis defines the joint optimization problem of resource allocation and video layer selection under the premise of V2 V sharing V2 I communication resources,and proposes a V2 V communication resource optimization mechanism based on the Distributed Proximal Policy Optimization algorithm.The model proposed in this thesis can use this algorithm to output continuous and multi-dimensional actions,improve video reception quality,reduce transmission delay and interference to vehicle-infrastructure communications.The simulation results show that the proposed method is about 0.2 higher in reward value than other algorithms.