Objectives  Due to the limitation of data transmission and storage cost, the sampling rates of most trajectories are low and uncertain. However, detailed urban models often require high-frequency trajectory data, for example, microscopic vehicle emission models often require trajectory data with a time interval of 1 s. Therefore, it is of great significance to reconstruct the trajectory data using the technique of interpolation.

  Methods  We propose a method to interpolate low-frequency trajectories considering the road intersections and vehicle modes. First, high-frequency trajectory data are used to train the theoretical probability model of vehicle motion modes. Second, the road intersections are used to determine the mode sequence between low-frequency trajectory points. Third, the theoretical probability model is solved by the genetic algorithm to calculate the distribution of time and distance of each mode, and then complete the high-frequency reconstruction of trajectory points.

  Results  The results suggest that the proposed method performs better than the conventional interpolation method by decreasing the root mean square error (RMSE) value with 62.9%, and better than the mode method that does not consider the road intersection by reducing the RMSE value by 12.2%.

  Conclusions  Therefore, the proposed method is valuable for the reconstruction of low-frequency vehicle trajectories.