Driver Car-Following Behavior Modeling Using Neural Network Based On Real Traffic Experimental Data

This paper presents a method used to model the driver’s behavior during car-following situations. The model is designed with a Back-Propagation (BP) neural network to reproduce the host vehicle longitudinal accelerations according to the states of Time Headway and Time-to-Collision inverse. To obtain high-quality data in real traffic for the modeling work, the authors carried out experiments with an instrumented vehicle test bed resulting in extraction of steady state car-following scenarios. The data segments were processed with Kalman filter to eliminate the measurement noise and estimate the longitudinal acceleration for the network training. The simulations with different leading vehicle speed inputs were then conducted. Results show that the neural network model is capable of simulating the driver’s car-following behavior and can adapt to normal car-following situations. These systems are designed to facilitate a comfortable distance between lead and following cars and to reduce the driver's workload in terms of manual speed and distance control actions.

Language

  • English

Media Info

  • Media Type: CD-ROM
  • Features: Figures; Photos; References;
  • Pagination: 9p
  • Monograph Title: ITS Connections: Saving Time. Saving Lives

Subject/Index Terms

Filing Info

  • Accession Number: 01140792
  • Record Type: Publication
  • Files: TRIS
  • Created Date: Sep 16 2009 4:00PM