Design of Mobility-Aware Map Partition and Distribution System for Smooth Automated Driving

High-Definition (HD) maps play an essential role in automated driving systems. It contains 3D point clouds that help automated vehicles achieve centimeter-level localization precision using LiDAR scan matching. Currently, it is unrealistic for vehicles to store an entire end-to-end map due to its tremendous data size. This implicates the need for these maps to be partitioned to reduce their size. Existing methods simply evenly cut map tiles and distribute them via legacy WLAN. However, map partitioning and distribution need a more intricate design that takes into account the vehicle's speed and the network's performance. In this paper, we state the necessary conditions for map partition and propose a strategy to guarantee an overlap between partitions and optimize the overlap's length to avoid localization failure. To complete the service chain, we further design a vehicle context-based millimeter-wave (mmWave) map distribution system under edge/cloud orchestration. Our proposed system is demonstrated on roads with advanced Intelligent Transportation System (ITS) equipment. The results show that our overlapping strategy gets more effective as the vehicle velocity increases, and mmWave can distribute a map of 267 MB within 2 seconds at common vehicle velocities in the urban area.