Research On Resource Allocation And Scheduling Optimization In Vehicle Platooning Networks

With the development of mobile communication technology and artificial intelligence(AI),it is difficult for the existing vehicular networks to cope with the unprecedented growth of the computation-intensive vehicular applications.Due to the large delay of long-distance communication,pure cloud computing cannot completely alleviate the pressure of vehicular networks.Edge computing closely proximity to users has been envisioned as a promising paradigm in recent years because of its flexible resource scheduling and rapid system response.Consequently,integrating edge computing into vehicular networks can push the cloud service to the network edge,promoting the ultimate goal of automatic driving.However,the phenomenal growth of the new vehicular applications significantly deteriorates the edge computing servers with limited resources.Driven by edge computing,vehicular cloud computing(VCC),which gathers the idle resources of vehicle terminals and realizes local data processing via the cooperation among vehicles,becomes a new research branch.At the same time,with the continuous improvement of the computing power of intelligent connected vehicles(ICVs)and the serious concerns on security of automatic driving,VCC has attracted great attentions to achieve the intelligence on device.However,in the scenario of VCC,the high-speed movement of vehicles leads to the short and unstable connection during of VCC link.At the same time,due to different request types and requirements,the available resources of each vehicle terminal are different and limited,which poses challenges to the reasonable and sufficient scheduling of vehicle terminal resources.To this point,this dissertation focuses on the vehicular edge computing network and carries out the research on the resource allocation optimization with vehicle scheduling based on vehicle platooning.Starting from the four aspects of network connectivity analysis based on vehicle platooning,task offloading in one platoon,edge learning in one platoon,and multi-platoon collaborative learning,this dissertation explores the impact of multi-user access on network connectivity,the competition of multi-task requests on on-board resources,the impact of terminal service capability differences on edge learning,and the impact of limited communication resources on multiplatoon distributed learning.The main work and contributions of this dissertation include the following four aspects:(1)Network connectivity analysis based on vehicle platooning.With the increasing number of ICVs,the vehicular networks are facing the problems of multi-user channel access competition and increasing packet collision probability.In view of these challenges,this network connectivity analysis based on vehicle platooning model is carried out,considering the impact caused by vehicle mobility and multi-user channel access competition.Firstly,jointly analyzing the relationship among vehicle density,vehicle speed,packet collision probability and relay probability,the network downlink performance analysis model based on relay communication is established.Defining the size of platoon,we analyze the network connectivity based on vehicle platooning to reduce the collision probability of data packets,aiming at maximizing the downlink performance.Simulation results show that the vehicle platoon model has great advantages in downlink communication by giving full play to the role of relay node,especially in the case of dense traffic flow.(2)Resource allocation and scheduling based on multi-task offloading in a platoon.Task offloading in VCC system faces the problem of low offloading efficiency caused by vehicle mobility,vehicle service capability,vehicle selfishness and multi task request.Due to the vehicle movement behavior,leading to the take-back of offloading tasks,we propose a resource allocation and scheduling scheme for task offloading and take-back to avoid the failure of offloading service due to incomplete tasks.Considering the instability of the platoon caused by vehicle movement(for instance,some vehicle leaves or joins the platoon),M/M/c queuing theory is used to model the offloading process,and the take-back is considered to avoid task processing failures.In order to encourage vehicles to contribute resources and overcome the limitation of selfishness on the utilization of vehicle resources,this dissertation proposes a task offloading scheme based on contract via defining the reward mechanism.Considering the multi-task offloading system,a multi-leader multifollower Stackelberg game(MLMF-SG)is formulated to analyse the incentives for task flows and resource allocation for platoon members.Specially,we propose an optimization scheme based on reinforcement learning(RL)to maximize the task revenues.Simulation results verify RL algorithm can avoid the situation of multi task blind bidding when the computing resources are insufficient.(3)Resource allocation and vehicle scheduling based on edge learning in a platoon.The limited communication and computing resources of vehicular networks,as well as the rising concerns about the privacy protection,bring significant challenges to the massive data training and analysis in a distributed manner.To cope with the problems above,a platoon-based distributed learning framework design for data learning is carried out,where the vacant computation resources of vehicle platooning networks are leveraged.In the proposed framework,a 2-phase Markovian stochastic process is utilized to depict the learning service heterogeneity for each participating vehicle.Meanwhile,we propose a joint scheduling and resource allocation scheme for efficiency-oriented distributed learning to maximize the learning accuracy subject to a given learning time constraint.Considering the learning convergence and remaining time,an on-demand scheduling scheme is introduced to determine the number of scheduled vehicles.Simulations results show that the proposed scheduling policy can schedule the number of participating vehicles on demand based on the trade-off between learning performance and learning latency.(4)Resource allocation and platoon scheduling based on multi-platoon distributed learning via intelligent reflecting surface(IRS).How to reduce the aggregation time of local learning models on resource-constrained edge computing systems is still a key problem in edge learning.A distributed learning system aided by IRS over vehicle platooning networks is studied.Due to the non-convexity of the phase shift optimization problem,a method based on sequential optimization algorithm(SOA)and a group-based optimization method are analyzed for single IRS aided and large-scale IRS aided communication,respectively.The relationship among model training,multi-platoon communication,and learning computing is difficult to be accurately described by mathematical formula.In order to overcome this challenge,the reliability of multi-platoon scheduling communication is described by the group-based phase-shift optimization scheme.A platoon scheduling scheme is designed based on the communication reliability and computing reliability of platoons.Simulation results show that IRS assisted communication can dramatically improve the reliability of multi-user communication network and alleviate the impact of bandwidth resource limitation.The scheduling scheme based on learning reliability balances the communication performance and computing performance of platoons,which is a reasonable user scheduling design.