Calibrating the Local and Platoon Dynamics of Car-following Models on the Reconstructed NGSIM Data

TL;DR

This study calibrates car-following models using reconstructed NGSIM data, focusing on leader-follower dynamics and platoon behavior, employing multiple calibration methods to improve model accuracy.

Contribution

It introduces a comprehensive calibration approach for IDM and FVDM models using reconstructed NGSIM data, including platoon-level calibration methods.

Findings

Platoon calibration improves model fit to real data.

Reconstructed data reduces outliers and inconsistencies.

Different calibration methods yield varying model accuracies.

Abstract

The NGSIM trajectory data are used to calibrate two car-following models - the IDM and the FVDM. We used the I80 dataset which has already been reconstructed to eliminate outliers, unphysical data, and internal and platoon inconsistencies contained in the original data.We extract from the data leader-follower pairs and platoons of up to five consecutive vehicles thereby eliminating all trajectories that are too short or contain lane changes. Four error measures based on speed and gap deviations are considered. Furthermore, we apply three calibration methods: local or direct calibration, global calibration, and platoon calibration. The last approach means that a platoon of several vehicles following a data-driven leader is simulated and compared to the observed dynamics.