A data-driven traffic modeling for analyzing the impacts of a freight departure time shift policy

Abstract

• We assess the impact of container transport demand on traffic using a data-driven traffic prediction model based on a graph neural network. • The model is implemented for the Port of Rotterdam using its container trip database combined with loop detector data for traffic flows. • The application of the model informs departure time shift policy, including the optimal shift volumes and patterns. • An optimized peak avoidance scheme can lead to significant congestion reduction while allowing to compensate the affected freight carriers. This paper proposes a data-driven transport modeling framework to assess the impact of freight departure time shift policies. We develop and apply the framework around the case of the port of Rotterdam. Container transport demand data and traffic data from the surrounding network are used as inputs. The model is based on a graph convolutional deep neural network that predicts traffic volume, speed, and vehicle loss hours in the system with high accuracy. The model allows us to quantify the benefits of different degrees of adjustment of truck departure times towards the off-peak hours. In our case, travel time reductions over the network are possible up to 10%. Freight demand management can build on the model to design departure time advisory schemes or incentive schemes for peak avoidance by freight traffic. These measures may improve the reliability of road freight operations as well as overall traffic conditions on the network.