Calibrating Car-Following Models by Using Trajectory Data

Kesting, Arne; Treiber, Martin · 2008 · OpenAlex-citations

DOI: 10.3141/2088-16

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the calibration of microscopic car-following models using high-resolution trajectory data, a shift from traditional macroscopic calibration methods. The authors aim to evaluate the reliability and robustness of two specific models: the Intelligent Driver Model (IDM) and the Velocity Difference Model (VDIFF). The motivation stems from the increasing availability of microscopic traffic data and the need to establish rigorous criteria for benchmarking these models, particularly regarding parameter stability and the role of driver reaction time. The methodology utilizes three publicly available trajectory datasets recorded in Stuttgart, Germany, by a vehicle equipped with a radar sensor. These datasets capture relative speed and distance to leading vehicles during peak-hour city traffic, including complex scenarios with acceleration, deceleration, and standstills. The authors calibrate the IDM and VDIFF by minimizing the deviation between observed driving dynamics and simulated trajectories using a genetic algorithm for nonlinear optimization. To assess robustness, they employ three distinct objective functions: relative error, absolute error, and a mixed error measure, focusing exclusively on vehicle gaps rather than speeds. Additionally, the study investigates the impact of an explicit reaction time delay and validates the models by applying parameter sets calibrated on one trajectory to the others. The results indicate that calibration errors range between 11% and 29%, consistent with previous studies. A key finding is the significant difference in model robustness: the calibrated parameters of the VDIFF vary strongly depending on the optimization criterion used, whereas the IDM parameters remain stable across different error measures. Furthermore, introducing an explicit reaction time parameter yielded negligible improvements in fit, suggesting that drivers compensate for human reaction delays through anticipation. Validation tests revealed that "intra-driver variability" accounts for a substantial portion of calibration errors, as parameters optimized for one driver’s trajectory did not necessarily fit another’s well. The IDM generally demonstrated superior performance and stability compared to the VDIFF, particularly in Dataset 3. The significance of this work lies in its contribution to the benchmarking of microscopic traffic models. The authors argue that model quality should be judged not only by fit errors but also by the consistency and robustness of calibrated parameters across different optimization criteria. The finding that reaction time is negligible in model fitting challenges conventional assumptions, implying that anticipation mechanisms are critical in car-following behavior. These insights provide a framework for evaluating the reliability of traffic simulation models and highlight the importance of considering individual driver variability in calibration processes.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• following distance

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: behavioral performance data
• Methodological Resource: validation psychometrics
• Theoretical Contribution: computational model