End-to-End Delay Analysis And Resource Management Of Vehicular Networks For Platoon-based Autonomous Driving

With the development of the Internet of Vehicles(IoV)and Artificial Intelligent(AI)technologies,designing autonomous driving applications and advanced driving assistance systems in on-road environments to enhance road safety and increase traffic efficiency has gradually become mainstream in both acadamic and industrial fields.In practical,compared to individual autonomous driving,platoon-based cooperative driving has higher stability,safety and road efficiency.Therefore,platoon-based autonomous driving has become a hot research topic and a promising technology for road-safety insurance.However,unlike traditional communication applications,platoon-based autonomous driving applications require the guarantee for lower bound of system performance,such as the end-to-end delay upper bound.Existing research work has reported that message delivery latency has a significant impact on the stability of vehicular platoons.Moreover,a timeout event of any packet can lead to huge damage to life and the economy.Consequently,when designing on-road applications,the delay upper bound of the communication and computation system must be carefully considered,and the system should reserve sufficient response time for emergencies based on the current system status.Nevertheless,the high dynamics of the on-road environment and the randomness of wireless links bring great challenges to analyzing the upper bound of the transmission delay.Existing research mostly analyses the network throughput under various protocols,but not the upper bound of the transmission latency.As for the optimization of resource allocation,Most work only considers the average network capacity optimization,but not the lower bound of the system performance.To fill the gaps mentioned above,this dissertation introduces Stochastic Network Calculus(SNC)theory to analyze the performance of IoV SNC can calculate the probability distribution of the queueing system performance lower bound.Therefore,SNC can provide theoretical models to describe the delay upper bound for various vehicular network Medium Access Control(MAC)protocols.However,there are many novel and complex operations in SNC,such as min-plus convolution and integrations.These complex calculations make the theoretical results cannot be optimized by conventional mathematical tools,such as convex optimization.Moreover,obtaining the network calculus results often needs long computation time,which makes it hard to be directly used in a real-time system.As a result,this dissertation uses Artificial Intelligent(AI)technologies,combined with SNC,to analyze and optimize the upper bound of real-time system communication latency.Furthermore,based on the system delay,this dissertation designs safety-related applications.The remainder of the dissertation is organized as follows,a)IoV transmission delay analysis based on network calculus,b)inter-vehicle distance control strategy based on delay upper bound theoretical model,c)broadcasting routing algorithm for high bandwidth and low latency applications,and d)communication and computation resource allocation for end-to-end process delay optimization.(1)This dissertation introduces network calculus theory into the vehicular network for system performance modeling and analysis,such as C-V2X Mode 4 protocol and hybrid vehicular networks.This dissertation concentrates on showing the network calculusbased analysis methods for various MAC protocols,different physical links and various applications.This dissertation innovatively analyses the transmission delay upper bound for C-V2X Mode 4 protocol with network calculus theory.To calculate the delay,the frame collision probability and send probability are first analyzed to further obtain the performance model for C-V2V.Moreover,this dissertation analyses the end-to-end delay for fixed-duration broadcasting applications in a vehicular network.The analysis of the latency for Dedicated Short Range Communication(DSRC)protocol and mm Wave network is also be presented.With the proposed novel theoretical model for parallel systems,this dissertation analyses the delay upper bound for hybrid vehicular networks.(2)This dissertation innovatively combines the results of the network calculus theory to design a safe distance control strategy,thus enabling a theoretical guarantee of traffic safety.A platoon-based inter-vehicle distance control algorithm is proposed.This dissertation proposes a reinforcement learning-based algorithm to reduce the network calculus delay upper bound.With the optimized latency,an adaptive smart inter-vehicle distance control algorithm is designed,which is consisted of a system performance prediction module,safe distance calculation module and distance mapping module.To verify the efficiency of the proposed algorithm,this dissertation implements an integrated simulation platform including AirSim and Mathematica.This platform can support autonomous driving,network calculus,artificial intelligence and vehicular communication simulations.(3)In order to solve the congestion problem of high-bandwidth broadcast applications,this dissertation combines the specific characteristics of the vehicular platoon and designs a broadcast routing strategy for high-bandwidth,low-latency applications on the routing layer.A novel routing strategy based on graph theory on the directional propagation network is proposed.This dissertation also analyses the computation complexity of the strategy and obtains the broadcasting delay upper bound of the proposed strategy with Deterministic Network Calculus(DNC).As a result,the proposed algorithm has both computation and communication theoretical delay upper bound guarantee.(4)For computation-intensive applications,this dissertation further incorporates computational latency into the optimization.In the mobile edge computing scenario,computation is a major part of the whole latency,which cannot be ignored.This dissertation studies the whole latency,which includes task offloading,computation,results returning and delivery.A novel framework for task offloading is proposed,which uses content merging technology of Information-Centric Networking(ICN)to reduce the traffic loads because platoon network usually has many same packets.This dissertation obtains the transmission latency for multi-cast applications based on the content merging technology and simplifies the results with graph theory.The proposed strategy has a different schedule for safety-related applications and non-safety-related applications.From the simulation results,the proposed strategy can improve road safety,traffic efficiency and on-road experience.