Research On Edge Collaborative Computing System For Connected And Autonomous Vehicles

As the product of autonomous driving technology combined with V2 X technology,connected and autonomous vehicles(CAVs)can achieve safe,efficient,comfortable and energy-saving driving.With the rapid growth of the CAV technology,more and more thirdparty applications are being developed and deployed on CAVs.They not only improve the user experience,but also provide more helpful services.However,the data processing mode based on cloud computing brings a lot of network delay,affecting the driving safety.Therefore,the current vehicles tend to use the data processing mode based on edge computing.But,with the increasing computation-intensive third-party applications,the burden of vehicle computing units(VCU)increases sharply,leading to the increase of applications' response latency,which seriously affects the driving safety and user experience.Inspired by the idea of edge computing,this thesis proposes two vehicle edge collaborative computing systems,which enhance the onboard VCU by utilizing different types of edge nodes near vehicles.The work and contributions of this thesis are as follows:1)Proposing an edge collaborative computing system inside the vehicle,MobileEdge.Considering the increasing computing capabilities of mobile devices such as smartphones,the MobileEdge system enables on-board VCU to integrate idle computing resources on mobile devices within the vehicle.It supports two types of computing,general-purpose computing and AI computing.And it designs a lightweight computing offloading method for mobile devices that consists of code migration and data offloading.In addition,considering the mobility of mobile devices and the variability of available resources,,the MobileEdge system also provides dynamic management of mobile devices,real-time monitoring of the status of device resources,and the development interface of customized task scheduling strategies,to support optimal task scheduling.2)Proposing an edge collaborative computing system outside the vehicle,XEdge.Because the performance of the MobileEdge system is limited by the number of mobile devices inside the vehicle,more edge computing resources need to be mined to supplement it.The XEdge system enables on-board VCU to collaborate with edge servers outside the vehicle.It also supports AI computing and general-purpose computing,and uses container technology to improve the isolation of tasks performed on edge servers and the system scalability.In addition,XEdge also designs an application partition function,which divides the application into two parts,the vehicle side and the server side,so as to take full advantage of the computing advantages of on-board VCU and the edge server.In this thesis,the system prototypes are implemented for MobileEdge and XEdge respectively.The experimental results show that the MobileEdge system can effectively reduce the computing burden of on-board VCU and the response time delay of CAV applications by offloading computation onto mobile devices;the XEdge system can partition the application,and effectively reduce the application response latency and improve the driving safety and user experience to some extent.