Calibration of the IDM car-following model using trajectory data

Luís Vasconcelos, Francisco Martins, Luísa Cruz-Lopes

Last modified: 2024-05-06

Abstract


Car-following models describe the longitudinal movement of vehicles and are a major component of microscopic simulation packages. As car-following models seek to replicate the behaviour of individual drivers, their mathematical formulation usually includes a large set of adjustable parameters. The calibration of the model is essential to achieve accurate results, but as it may be a complex and expensive task, users often rely on default values or on simple techniques that offer poor transferability.
In this paper we describe a calibration technique for the Intelligent Driver Model (IDM) that explicitly accounts for the physical meaning of each parameter. Trajectory data was collected for a sample of Portuguese drivers using an instrumented vehicle and covers the most relevant cases, such as unconstrained acceleration and deceleration manoeuvres and car following in steady-state conditions. A two-step calibration technique was followed: first, subsets of parameters with clear physical meanings were manually adjusted to replicate the velocity profiles of simple driving patterns; second, the results were used to define the bounds of values within an automatic calibration procedure for normal driving conditions.
First results show that the calibration procedure allows to accurately replicate the real trajectories. There is still the concern with the transferability of results and further work is required to understand how to reach the best compromise between the model’s descriptive and predictive capacities.

Keywords


Car-following; Intelligent Driver Model; Calibration; Genetic Algorithm

Full Text: PDF