Abstract
Vehicle trajectory data contain extensive spatiotemporal information and are of great significance for analyzing the operational patterns of urban traffic networks, optimizing traffic signal control and achieving refined traffic management. However, due to the low penetration rate of probe vehicles and the limited coverage of fixed sensors, it remains challenging to obtain comprehensive full-sample vehicle trajectory data. To address this issue, this paper proposes a multi-source data fusion-based vehicle trajectory reconstruction method, which comprises vehicle trajectory state estimation and a self-optimization algorithm. First, the trajectory states of undetected vehicles are categorized into four types based on the trajectory states of adjacent probe vehicles. Four corresponding trajectory estimation models are then established using an extended Intelligent Driver Model to reconstruct the initial trajectories of undetected vehicles. Second, a particle filter-based trajectory self-optimization algorithm is proposed, integrating upstream and downstream fixed sensor data to iteratively correct and optimize the initial trajectories by minimizing errors, thereby enhancing trajectory accuracy and smoothness. Case studies demonstrate that the proposed method achieves outstanding performance under low PV penetration rates and in complex traffic environments. Compared to baseline methods, MAE, MAPE, and RMSE are reduced by 14.31%, 22.87%, and 13.36%, respectively. Furthermore, the reconstruction errors of the proposed method gradually decrease as traffic density and PV penetration rates increase. Notably, PV penetration has a more significant impact on model accuracy. These findings confirm the robustness and effectiveness of the proposed method in complex traffic scenarios, providing critical technical support for refined urban traffic management and optimized decision-making.