Reliable And Efficient Wireless Communication For Vehicular Network

With rapid increase in the number of vehicles,vehicular networks,as typical application scenario of the fifth generation(5G)wireless network,become increasingly important for improving driving safety and traffic efficiency.Consequently,recent research priority of automotive industry is focused on the development of intelligent connected vehicles,which rely on autonomous driving capabilities,varieties of vehicle sensor and wireless communication devices.By leveraging Vehicle-to-Everything(V2X)communications,intelligent connected vehicles can collect surrounding information and then have an in-depth understanding of their surrounding environment to make suitable control decisions.However,due to the high complexity and dynamic network topology of vehicular networks,current vehicular communication technologies cannot effectively support diverse quality of service(QoS)requirements of V2X services.Thus,we believe it is essential to investigate reliable and efficient wireless communication mechanisms for vehicular networks.Firstly,we design an intelligent network slicing architecture for V2X services,through combing both network softwarization and network intelligentization technologies.Then,we focus on dealing with related challenges rasied in the realization of the proposed architecture.In the micro aspects,to promote the driving safety,we optimize raido resource allocation of V2V communication to guarantee the latency and reliability requirements of safety related services,and investigate the optimization of vehicle platooning control.On the other hand,in the macro aspect,we propose a deep reinforcement learning based network slicing deployment scheme according to the diverse requirements of V2X services.The major contributions of this dissertation are given as follows:1.Intelligent:Network Slicing Architecture Desgin for V2X ServicesBenefiting from the widely deployed LTE infrastructures,the 5G wireless networks has been considered as a promising candidate to support emerging V2X services.However,existing LTE networks cannot efficiently support the stringent but diverse and varying requirements of V2X services.One effective solution to this challenge is network slicing,where different services could be supported by logically separated networks.To mitigate the increasing complexity of network slicing,we propose to leverage the artificial intelligence(AI)technologies for automated network operation.Specifically,we present the concept of intelligent network slicing architecture for V2X services,where the multi-dimension network resources are virtualized and divided for different network slices.Furthermore,to perform the deployment and management of network slices fast and efficiently,we propose a hierarchical control layer,which can be further divided into two sub-layers:slicing deployment controller(SDCon)and slicing management controller(SMCon).From a two-timescale perspective,SDCon,an AI agent performs global control of network slicing,is responsible for deployment of network slices on a long-term timescale,while SMCon,a real-time regional control entity,directly manages multi-dimension resources in each network slice on a short-term timescale.In achieving optimized slicing intelligently,several critical techniques,including ensuring requirements of safety related services and design of AI algorithm,are discussed to tackle the related challenges.2.Latency and Reliability Guaranteed Resource Allocation Scheme for Safety-related V2V ServiceIn this section,we propose jointly optimizing the radio resource,power allocation,and modulation/coding schemes of the V2V communications,in order to guarantee the latency and reliability requirements of safety related service while maximizing the information rate of cellular user equipment(CUE).To ensure the solvability of this optimization problem,the packet latency constraint is first transformed into a data rate constraint based on random network analysis by adopting the Poisson distribution model for the packet arrival process of each VUE.Then,utilizing the Lagrange dual decomposition and binary search,a resource management algorithm is proposed to find the optimal solution of joint optimization problem with reasonable complexity.Simulation results show that the proposed radio resource management scheme can reduce the interference from V2V communication to CUEs and ensure the latency and reliability requirements of V2V communication.3.Joint Radio Resource Allocation and Control for Vehicle PlatooningTo improve safety and efficiency of transportation system,an effective approach is vehicle platooning,in which a group of vehicles maintains a pre-defined moving pattern by minimizing tracking error of each vehicle.As a networked control process,platooning control relies on the cooperative awareness messages(CAM)periodically exchanged in the vehicle platoon.In meeting communication and control expectations,we propose to jointly optimize the radio resource allocation in LTE V2V network for the CAM transmission and control parameters of vehicle,in order tominimize the tracking error while guaranteeing the reliability requirements of V2V communication and string stability of platoon.To solve the formulated problem,we develop a sub-optimal solution.Firstly,we present a tracking error based scheduling criterion and then obtain a radio resource allocation scheme by bipartite graph matching.Then,control parameters of vehicles are adapted to minimize the tracking error by heuristic gradient descent based method.Simulation results show that the proposed scheme can reduce the tracking error as well as ensure the platoon stability compared to the existing schemes.4.Deep reinforcement Learning;for Wireless Resource Management in Network SlicingBased on the proposed intelligent network slicing architecture,we investigate network slicing deployment policy for wireless radio resource assignment and parameter setup of network function,with the purpose of ensuring QoS requirements of V2V services while maximizing the long-term revenue of network operator.The network deployment policy optimal problem is formulated as an infinite-horizon discount-reward partially observed Markov decision process(POMDP).Considering the unknown and complex environments dynamics of vehicle networks,we propose an actor-critic reinforcement learning approach to learn the policy by interacting with vehicular networks.Since vehicular network status is very complex and partially observable,the dynamic of observation state is non-Markovian.Therefore,we incorporate Long Short-Term Memory(LSTM)network into the deep neural network to capture temporal dependencies between current observation state and former states.Finally,we develop a simulation platform to illustrate the effectiveness of our proposed deep rein forcement learning based intelligent network slicing.