Fully Sampled Trajectory Reconstruction for Manned Container Trucks in Mixed Traffic Flow at Ports

In the mixed traffic environment of a port’s container road network, sensor malfunctions and unstable transmission signals can cause manned container trucks (MCTs) to suffer from much more severe GPS data loss than unmanned container trucks (UCTs). Producing fully sampled reconstructions of the trajectories of MCTs is still a critical challenge in advancing data-driven intelligent transportation systems for digital ports. Considering the specific traffic conditions of a given port, this paper introduces an innovative approach for the reconstruction of MCTs’ trajectory. Firstly, an improved generalized adaptive smoothing method (GASM) model is employed to construct a velocity field of the port road network. By adjusting kernel function parameters and constructing virtual free-flow trajectories, the issues of low data density and poor data quality are resolved. Secondly, candidate trajectory lines for each trajectory point are designed by integrating macroscopic velocity field conditions with microscopic vehicle operating behaviors. Finally, trajectory offset weights are assigned based on the position of each trajectory point, including the lane-changing position if a lane change occurs, to fuse the candidate lines and estimate the vehicle’s probable trajectory. The lane-changing position is determined as the point exhibiting the maximum velocity difference between two lanes within the lane-changing spatiotemporal zone. The proposed trajectory reconstruction method was validated using real data. It leads to a substantial reduction in key error metrics, including mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE), demonstrating a remarkable enhancement in the accuracy of trajectory reconstruction. Moreover, it can effectively deduce optimal vehicle lane-change positions based on data analysis.