DynamicMap 2.0: A Traffic Data Management Platform Leveraging Clouds, Edges and Embedded Systems

摘要

Various embedded sensors such as LiDAR, cameras and GPS are mounted on vehicles today. And, vehicles with communicationequipment can exchange sensor data between not only surrounding vehicles but also road-side-units. To achieve traffic safety andcomfort driving, utilization of these traffic data are important for autonomous driving and advanced driving assistance systems.Our research group has been developing a city-scale traffic data management platform named DynamicMap 2.0. This paperproposes a software architecture of our platform. Our platform uniformly treats road maps, static information, dynamic informationand predicted information. Road maps consist of three-level granularities: links, lanes and physical shapes. Especially, adata representation of lane-level road maps is newly proposed. Static information includes fixed properties about objects on themap. Dynamic information includes streaming data from sensors. And, predicted information is computed by prediction algorithmsand simulations. Our platform provides query language for searching these traffic data. A developer of traffic applicationswrites a query (continuous query or one-shot query) which specifies integration and filtering of data. As a use case of ourplatform, we developed a prototype of a cooperative merging assistance system at intersections, and verify that our platformoperates correctly.