In a complex urban environment, due to the unavoidable interruption of GNSS positioning signals and the accumulation of errors during vehicle driving, the collected vehicle trajectory data is likely to be inaccurate and incomplete. A weighted trajectory reconstruction algorithm based on a bidirectional RNN deep network is proposed. GNSS/OBD trajectory acquisition equipment is used to collect vehicle trajectory information, and multi-source data fusion is used to realize bidirectional weighted trajectory reconstruction. At the same time, the neural arithmetic logic unit (NALU) is introduced into the trajectory reconstruction model to strengthen the extrapolation ability of the deep network and ensure the accuracy of trajectory prediction, which can improve the robustness of the algorithm in trajectory reconstruction when dealing with complex urban road sections. The actual urban road section was selected for testing experiments, and a comparative analysis was carried out with existing methods. Through root mean square error (RMSE, root-mean-square error) and using Google Earth to visualize the reconstructed trajectory, the experimental results demonstrate the effectiveness and reliability of the proposed algorithm.