Investigation of longitudinal dynamics of vehicles in disordered traffic

Abstract

Vehicles in developing countries have wide variations in their static and dynamic characteristics, and drivers tend to not follow lane discipline. Models for driving behavior under such disordered traffic conditions need to include the vehicle dynamics and their interactions with the surrounding environment. Calibration of those models is necessary to evaluate their predictive power and suitability for analyzing traffic flow under disordered traffic. The present study aims to calibrate a longitudinal dynamics model, the High-Speed Social-Force Model (HSFM) using a vehicle trajectory dataset collected from Chennai city. The HSFM was calibrated by minimizing the deviations between the simulated and observed longitudinal coordinates of vehicles using a genetic algorithm. The observed and simulated vehicle trajectories were compared using a goodness of fit function of the positions. The convergence of the objective function has been illustrated with the help of fitness landscapes. The calibration errors were found to be within the acceptable range and the optimal parameter values were found to be consistent. The outcomes of the study indicate that the model can capture the influence of non-overlapping leaders under disordered traffic conditions.