Research On Construction And Update Of Unstructured Road Digital Map In Enclosed Area

With the rise of the artificial intelligence technology wave,autonomous vehicles have gradually become a research hotspot in academia and industry.Due to the busy traffic on open roads and the complex environment,the first possible scenario for implementing highly automated driving is currently closed areas with relatively simple environments(such as open-pit mining areas,factories,closed parks and docks).As a high-precision,information-rich digital map for autonomous driving,high-precision digital map is considered to be one of the key technologies to realize the commercialization of L3 and above autonomous driving.The high-precision digital map is different from the traditional road-level navigation digital map.It is a lane-level digital map that contains rich road information.The accuracy of traditional road-level navigation digital maps is only meter level,and the amount of road information is not enough to support the operation of autonomous vehicles,and the usability and accuracy are low.However,high-precision lane-level digital maps based on LiDAR and Camera have some suggestions.In order to make up for the gap between traditional road-level digital maps and high-precision lane-level digital maps to some extent,this paper studies a method based on the difficulty of map,large amount of data,complex calculation,high cost of map creation,and long update cycle.High-precision GNSS-IMU(Global Navigation Satellite SystemInertial Measurement Unit)data is a closed area unstructured road digital map construction method and a map update method based on the concept of crowdsourcing.First,based on the mainstream OpenDRIVE standard format,a map format specially designed for unstructured road scenes in enclosed areas was designed.At the same time,a data table for storing map data in the corresponding database is designed to facilitate map data management.Then,the construction method of unstructured road digital map in enclosed area is studied.By collecting high-precision GNSS-IMU vehicle trajectory data as raw data,trajectory preprocessing methods,road boundary generation algorithms,boundary extension algorithms,road centerline generation algorithms,and manual labeling methods are designed to construct a unstructured road digital map in closed area.Secondly,the update method of unstructured road digital map in enclosed area is studied.Based on the idea of crowdsourcing,this paper collects the highprecision GNSS-IMU trajectory data of construction vehicles involved in the work of causing terrain changes in the enclosed area and performs trajectory preprocessing,and then designs the boundary extension,road addition and road closure methods.The map is updated.Finally,the actual vehicle verification based on a closed site.Through the map data collection vehicle and the corresponding software platform,the map construction of the unstructured road in the enclosed area is realized,and the boundary expansion,road addition and road closure map update of the constructed map are realized at the same time.