Fast Delivery And Update Schemes Of High-Definition Map Based On Edge-Cloud Cooperation

In recent years,autonomous driving has imposed greater load demands on HighDefinition(HD)maps.Compared with digital navigation maps,HD maps are enabled with larger data volume,more complex data structure,and higher update frequency,which means that the existing cloud-based communication network architecture can barely meet the latency requirements of HD map delivery and update.In response,the thesis aims to solve the latency dilemma of HD map delivery and update on the basis of the edge-cloud-cooperation-based network architecture.Firstly,the thesis classifies HD map contents into four categories according to precision and dynamics.Secondly,the thesis analyzes the characteristics and functions of each category of the contents.Then,the thesis proposes that cooperative caching,collaborative computing,heterogeneous radio access technologies,and layered optimization are the indispensable functions of the architecture which is expected to support HD map delivery and update.Based on above analysis,the thesis designs an edge-cloud-cooperationbased network architecture to support HD maps and clarifies its structure,components,and workflows.With such architecture,the thesis optimizes the delivery and update mechanisms of HD maps,respectively.To deal with the problem of HD map delivery,the thesis proposes a static map delivery mechanism and a dynamic map delivery mechanism,respectively,considering the heterogeneity of HD maps.In the workflow of static map delivery,the thesis aims to minimize the content access latency by optimizing the cooperative caching policy and the cooperative responding policy of each edge server,considering the layered data structure and popularity of static maps.In the workflow of dynamic map delivery,the thesis aims to maximize the timely delivery rate by optimizing the caching and pushing policy of each access point in the service area of one edge server,considering the small data volume,small target area,and limited lifetime of dynamic maps.Furthermore,the thesis verifies the effectiveness and advantages of the two mechanisms by simulations.To deal with the problem of HD map update,the thesis firstly divides the workflow of the update process into two phases,namely the data collection phase and the data processing phase.Then,the thesis proposes a traffic-allocation-based fast data collection mechanism and an edge-cloud-cooperation-based fast data processing mechanism,respectively.In the data collection phase,the thesis aims to minimize the network latency of the system by optimizing the traffic allocation policy on uplinks of each vehicle so that the probable traffic congestion on saturated links can be avoided.In the data processing phase,the thesis aims to minimize the task computing latency by optimizing the task split policy and computing resource allocation policy of the edge servers and the cloud server so that the resource bottleneck of edge servers can be conquered.Furthermore,the thesis verifies the effectiveness and advantages of the two mechanisms by simulations.