| Popis: |
The estimation of freight trip distribution is an important issue in freight transport studies, and collecting data on the spatiotemporal aggregated number of freight vehicles without trip information is relatively easy and inexpensive. In this paper, we propose a new model for freight distribution estimation, referred to as “TripCFDM.” Unlike the traditional collective flow diffusion model (CFDM), which solely relies on aggregate data, TripCFDM relies on aggregate data along with a small amount of trip data. CFDM estimates the probability of latent zonal freight trips based on aggregate data corresponding to the number of vehicles in each zone and time step. The use of the aggregate and small amounts of trip data ensures a realistic data acquisition environment. Our model is applied to freight trips in the Tokyo Metropolitan Area using the National Survey of Roads and Streets Traffic Conditions (Road Traffic Census). The results reveal that TripCFDM yields a significant improvement in the estimated probability compared with CFDM, particularly for intrazonal freight trips. However, the estimation results of interzonal trips obtained by CFDM and TripCFDM tend to be similar, suggesting that CFDM alone may be adequate to estimate the probability of interzonal trips with a sufficient degree of accuracy; nevertheless, TripCFDM is a better method for freight trip distribution estimation, including for intrazonal trips. The results of CFDM are not significantly affected by the time steps in the input data, whereas TripCFDM can yield higher accuracies with longer time steps. Regarding the number of zones and zone size, sufficient zone aggregation may be effective in improving the estimation accuracy because a detailed zonal division leads to a small probability of origin–destination trips with extensive computations. Based on this modified CFDM approach, origin and destination matrices for freight transport can be generated without large-scale surveys. |
