Research On Wireless Transmission Technology And Optimization Methods Of 5g-v2x For Cooperative Autonomous Driving

The combination of Vehicle-to-Everything(V2X)and autonomous driving enables the creation of cooperative autonomous driving,which has two key cooperative features:sensing and maneuvering.Cooperative autonomous driving can provide enhanced safety and comfort,reduce energy consumption,and improve traffic efficiency.The quality of service(QoS)requirements from cooperative autonomous driving can be characterized as ultra-reliable,low latency,high traffic,and high mobility.These requirements cannot be met by the existing vehicle networking technology represented by LTE-V2X(Long Term Evolution-V2X)and DSRC(Dedicated Short Range Communication).Therefore,academia and industry began to study the evolved V2X technology based on 5G(the 5th generation mobile networks)around 2015.This thesis focuses on four problems of wireless transmission technology and its optimization methods that lie in the developing process from prototype system development to large-scale commercial deployment and to further evolution of 5G-V2X.These four problems including the wireless transmission scheme design and verification of prototype system,the power allocation scheme for coexistence scenarios of different commercial systems,the load prediction of a city-wide 5G-V2X network,and the general problem-solving tools for complex optimization models.The novelties and contributions of this thesis can be summarized as follows:(1)First of all,this thesis studies the problem of the wireless transmission scheme design and field tests of the 5G-V2X prototype system.The requirements of cooperative autonomous driving for 5G-V2X communication are transformed into the problem of joint optimization of low latency,ultra-reliable,high traffic,and high mobility.And the technical indicators to be achieved in the scheme design are clarified.On these bases,a mobility management mechanism based on hierarchical heterogeneous network architecture,a localized data transmission mechanism based on mobile edge computing,and a URLLC mechanism based on channel processing optimization and fast retransmission,are proposed.Through simulation and field tests,it is verified that the proposed scheme meets the performance requirements.This scheme supports the first cooperative autonomous driving function verification in China that all links including V2V meet the 5G-V2X requirements.(2)Secondly,this thesis studies the problem of the power allocation scheme for coexistence scenarios of different commercial systems.The practical problem of how to encourage operators to share licensed spectrum resources with non-subscribed users are considered.The power allocation in the scenario of sharing spectrum resources between Sidelink and Uplink from the perspective of economics is analyzed,and a new index of economic energy efficiency is introduced.On these bases,a new power allocation scheme based on Stackelberg Game is proposed.In this scheme,the macro station obtains the maximum utility by optimizing the interference pricing while the V2X terminal obtains the maximum utility by optimizing the economic energy efficiency.The unique equilibrium of the Stackelberg Equilibrium is proved.The optimal pricing of the macro station and the optimal transmission power of the V2X terminals are obtained.(3)Thirdly,this thesis studies the problem of the network load prediction of a citywide 5G-V2X network.The research results of deep learning based traffic flow prediction and wireless network traffic prediction are investigated.The feasibility of joint prediction of multiple network load related time-varying attributes,such as the number of users,migration trend,and network traffic,in each region of 5G-V2X network at citywide within one model is proved.The limitations of the existing CNN model caused by gird segmentation are analysed.The problem of joint prediction multiple time-varying attributes under irregular area division is defined.On these bases,a permutation based sample reconstruction method and a permutation based spatiotemporal residual network model are proposed,which extends the application scope of the existing model to the irregular segmentation scenario and supports the joint prediction of multiple time-varying attributes.It is proved by real-world data that the proposed method and model are suitable for irregular segmentation,and can improve the prediction accuracy,and also have good generalization performance.(4)Finally,this thesis studies the problem of the general problem-solving tools for complex optimization models.We focus on the Cuckoo Search(CS)algorithm,which is a widely used optimization tool for complex optimization models.The applications of the CS in the research of wireless transmission of V2X,and the main improvements and variants of CS algorithm are analyzed.On this basis,an improved CS algorithm with global convergence property is proposed,which integrates the subgroup division strategy,gray wolf optimization,and dynamic adaptive step size.Simulation results show that the proposed algorithm is superior to the existing standard CS algorithm,RC-SSCS(Cuckoo Search algorithm Based on Repeat-Cycle asymptotic Self-learning and Self-Evolving disturbance),and many other swarm intelligence algorithms in terms of convergence speed and accuracy.The results of research on power allocation,network load prediction,and optimization tools are not only applicable to 5G-V2X,but also have important theoretical significance and application value for LTE-V2X.