Deriving features of traffic flow around an intersection from trajectories of vehicles

Abstract

Features of traffic flow, such as travel time, traffic volume, road choice, etc., are always important information for traffic planning and management. Conventional methods of collecting traffic flow features include loop detectors, closed-circuit televisions, field investigation, etc. They record vehicles' mobility in an indirect and infrastructure-based manner, and have difficulties in producing road choice data of vehicles or cover the whole road network. On the other hand, the advances in positioning and communication technologies make it possible to collect trajectories of moving vehicles. The collected trajectories can record the above information in a ‘natural’, direct, and vehicle-based manner. So far, various applications of location-based services have been implemented and meanwhile, a large number of vehicle trajectories have been collected. Therefore, it is possible to bring trajectories of vehicles into traffic study as an important supplement to conventional traffic data. To date, several fundamental works have been conducted to handle vehicle trajectories, such as map-matching algorithms, trajectory data representation, and trajectory data indexing and query language, etc. Based on them, the paper aims at illustrating a practical application of trajectories in analyzing traffic flow around a traffic intersection. A typical intersection in Shanghai with heavy traffic is selected as an example. More than 10GB trajectory data collected with GPS receivers of about 2000 taxis in Shanghai for 6 days is obtained. All trips around the intersection are extracted first. Then, a map-matching algorithm is employed to represent each trip as a journey record, i.e. a network route consisting of road segments and the intersection. Third, all journey records are classified into several groups according to their turning directions around the intersection, and then, specific analyses can be made. Since temporal information is inherent in each trip, all these analyses are time-dependant, such as to summarize the number of vehicles turning left or the average time for vehicles going straight across the intersection from a certain road segment during rush hours. In our case study, coupling the analyzing results with the surrounding characteristics of the intersection, we have tried to figure out and explain various patterns behind traffic flow. This example illustrates the potential of applying vehicle trajectories to traffic study.