Identification of Common Paths for Rideshare Routing by Mining Vehicle Trajectories
Trajectory data generated from various location sensors are increasingly available nowadays, providing comprehensive information on detailed itineraries that can be used in traffic and mobility analyses. This paper presents a study on the identification of underlying traffic stream clusters by exploring vehicle trajectory data. It aims to break down the spatial distribution of mobility flows and recognize the common paths according to which tentative rideshare routes could be set up. A step-by-step methodological framework, which includes map matching, measuring path similarity at the geometry level, trajectory clustering according to overlapping segments, locating trunk routes on the network, and classifying new trajectories on the identified routes, is proposed. An experiment has been conducted using trajectories collected in the real world. The results showed that different traffic streams with regard to direction and travel distance can be identified clearly. Overall, this study offers a comprehensive means of leveraging trajectory data to analyze itinerary patterns and make comparisons between trajectories. Compared with previous efforts, the proposed method is effective in distinguishing between traffic streams even with regard to bidirectional flows along the same route. In addition, it has the capacity to identify the common routes of underlying traffic streams instead of simply picking out single hot links or origin–destination pairs. This study is an example of data-driven research that uses digital trajectories to detect common path issues and offers detailed guidance on its application in this respect.