Vehicular mobility scenarios are utilized to study vehicular networks and transportation systems. However, the generation of vehicular simulation scenarios at scale poses several research challenges. Large-scale vehicular datasets (in geographic coverage and time span) are not easily or publicly available, which hinders the generation of data-driven scenarios. In this paper, we introduce a systematic method, called En Route, to generate vehicular mobility scenarios from traffic datasets such as one derived from thousands of available traffic webcams covering major cities around the world. Our framework includes datadriven components for estimation of traffic density, flow, road occupancy, as well as origin-destination (O/D) matrix estimation, trip generation, and route/navigation calculations. By applying the framework, we explore the city of London using the dataset of —100 traffic cameras throughout the city. We utilize available taxicab trips and traffic measurement datasets as guidelines for reasonable estimation of flow values, trip generation and O/D matrix. Our initial study shows reproducible step-by-step procedures, detailing parameter choices and settings, and measuring effects of changing those settings on the scenario outcomes. The results show a clear relation between flow and occupancy, and that trip duration follows a Lognormal distribution. Also, using traffic-aware routing (vs. shortest path) results in less congestion and more completed trips for a given simulation time. Queue distributions are obtained showing that over 90% of the intersection queues are 15 meters long (have average of 4 cars), and 90% of the roads carry less than 20 vehicles/km with average speed of —22.3 mph. Future studies shall provide multi-city simulations with further analysis.
En route: Towards vehicular mobility scenario generation at scale
2017-05-01
1291385 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
VEHICULAR TRAVEL ROUTE DISPLAY METHOD AND VEHICULAR TRAVEL ROUTE DISPLAY DEVICE
Europäisches Patentamt | 2020
|