Evaluating Parametric Car-Following Models in Naturalistic Congestion: Insights in Driver Behavior and Model Limitations

TL;DR

This study evaluates five parametric car-following models (Gipps, IDM, ACC, OVM, FVDM) on naturalistic congestion data using RMSNE, revealing similar predictive performance after calibration. It uncovers systematic model-reality gaps linked to driver behaviors such as coasting and idle creeping, alongside distraction and momentum effects that perturb model predictions. Time-series clustering with Dynamic Time Warping identifies five distinct driver-maneuver clusters corresponding to disagreement periods, offering insight into temporal transitions during congestion. The results suggest incorporating coasting and idle-creep dynamics into parametric models to improve low-speed congested traffic modeling and provide reproducible code via GitHub.

Abstract

Car-Following is a broadly studied state of driving, and many modeling approaches through various heuristics and engineering methods have been proposed. Congestion is a common traffic phenomenon also widely investigated, both from macroscopic and microscopic perspectives. Yet, current literature lack a unified evaluation of Car-Following models with naturalistic congestion data. This paper compares the performance of five parametric Car-Following models: IDM, ACC, Gipps, OVM, and FVDM, using a rich naturalistic congestion dataset. The five models in question is found to perform similarly when optimized over the same RMSNE metric. Sub-sequences of Car-Following where models noticeably disagree with driver behavior is noticed and separately investigated. A review of corresponding front-facing and cabin video data reveals distraction and driving with momentum as potential reasons for model-reality difference. We further show that drivers often employ coasting and idle creep under Car-Following in different speed ranges, which existing parametric models fail to capture. Finally, time-series clustering is performed and analysis of result clusters align with empirical findings. Our findings highlight the necessity to consider vehicle dynamical properties including coasting and idle creep abilities, which drivers take extensive use of under low speed congestions. Future research could integrate such parameters with traditional parametric models to improve congestion modeling performance. We also suggest future research into investigating temporal correlations between clustered blocks to reveal behavioral transition patterns exhibited by drivers in congestions. Source code for this study can be found on Github.

Figures (10)

• Figure 1: Distributions of Per-Sequence RMSNE with high-speed and congestion optimized parameters.
• Figure 2: Example of Space Gap w.r.t Time before and after calibration on congestion dataset.
• Figure 3: Example Space Gap w.r.t Time for model-reality difference cases.
• Figure 4: Example Space Gap/Velocity w.r.t Time for model-reality difference cases. The first subfigure shows the space gap case where driver drives with momentum, and the second subfigure illustrates the acceleration pattern adopted by the driver and models.
• Figure 5: Example Space Gap/Velocity w.r.t Time for model-reality difference cases. Driver in this sequence elects to lift and coast as visible in subfigure 2.