Can we trust the speed–spacing relationship estimated by car-following model from non-stationary trajectory data?

This paper studied the reliability of estimating individual speed-spacing relationship by calibrating car-following models with non-stationary data. Empirical trajectories with long durations are used to extract the near stationary data and the dynamic trajectory data. Three car-following models, the Optimal Velocity Model (OVM), the Full Velocity Difference Model, and the Intelligent Driver Model (IDM), are applied for the model calibration. A root mean square error-based indicator is introduced to measure the performance of the model estimation for the stationary speed-spacing relationship. It is found that both the OVM and the IDM perform well in estimating the individual speed-spacing relationship. The IDM has the vantage in the estimation under the situation far away from the stationary traffic state. The results of linear regression indicate that the Stationary-Data-Coverage and Multiple-Dynamic-Type are beneficial to the reliability of the estimation for the individual speed-spacing relationship.