Using Cellular Signaling Data to Produce Contextually Relevant High-Fidelity Demand for a Large-Scale Dynamic Traffic Assignment Model

抄録

Travel demand surges related to long-weekend holidays have clogged the entire national highway system in Taiwan, resulting in excessively prolonged travel times. As such, a large-scale simulation-based dynamic traffic assignment (DTA) was developed to evaluate various strategies and more accurately analyze their effect on system congestion. For a large-scale nationwide DTA model, obtaining demand data that is contextually relevant to long-weekend scenarios is challenging. To address this challenge, the use of cellular signaling data was explored. This paper first discusses converting the latest-generation cellular signaling data to high-fidelity trip chain data. Secondly, the process of extracting trip chain data for specific periods to develop time-dependent origin-destination matrices required for the DTA model. Model validation results indicate mean absolute percentage error (MAPE) of volume, travel time, travel speed ranging between 3.54% to 45.25%, which is deemed satisfactory for such large-scale network and data variability. A cast study on Freeway No. 5 illustrates the application of the model, and the result shows the ability to travel demand management during long-holiday.