Car-following models like the Intelligent Driver Model (IDM) describe the longitudinal behavior of an ego-car that reacts to a leading vehicle. They have been applied on a broad range of mobility-related tasks like the analysis of traffic phenomena, microscopic traffic simulations or single vehicle behavior prediction. Although car-following models can be formulated with explicit delays, the IDM is generally used in form of a time-continuous differential equation without delays. One reason is that in early IDM parameter calibration work it was found that the effect of the explicit delay is negligible. In this paper, we reopen the question of the importance of delays for accurate trajectory modeling, based on an analysis with real trajectories. We show that indeed both trajectory matching and prediction errors are strongly reduced when considering delays, enabling and improving their usage for safety-relevant scenarios like required for behavior prediction in intelligent Adaptive Driving Assistant Systems (ADAS). In addition, from optimizing the match of the model with real trajectories we gain a delay distribution that resembles the distribution of human driver reaction times.
Robustifying Car-Following Models with Explicit Reaction Times
2022-10-08
1533563 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Robustifying the deployment of tinyML models for autonomous mini-vehicles
BASE | 2021
|Robustifying the deployment of tinyml models for autonomous mini-vehicles
BASE | 2021
|Optimal state-rate feedback explicit model-following
AIAA | 1984
|