Comparing the calibration methods for intelligent driver model using Beijing data

Abstract

Safe and comfortable journeys in the traffic systems greatly depend on the drivers' behaviour. Car following mode (CFM) describes the drivers' behaviour by following the paths of preceding drivers in a traffic flow. Intelligent driver model (IDM) is the most popular CFM for safe and comfortable journeys. This research compares the calibrated methods with each other for Beijing data by using genetic-algorithm (GA), sequential-quadratic-programming (SQP) and simultaneous-perturbation-stochastic-approximation (SPSA). Findings reveal that the IDM simulation parameters, such as maximum-acceleration, maximum-deceleration, desired-speed, minimum-headway and minimum-jam-distance, differ from the IDM calibration parameters by -66.21%, -36.57%, -44.98%, -98.77%, -9.76% respectively for SPSA. Findings also show that the negative percentage values represent the decrease from the IDM simulation parameters, and the positive percentage values represent the increase from the same parameters. The comparison results show that the IDM calibration parameters are more precise with SPSA than GA and SQP for safe and comfortable journeys.