Experimental Analysis of Car-Following Dynamics and Traffic Stability

Abstract

The study of car-following dynamics is useful for capacity analysis, safety research, and traffic simulation. There is also growing interest in its applications in intelligent transportation systems, such as advanced vehicle control and safety systems and autonomous cruise control systems. A large number of car-following models have been developed in the past five decades. Some of them were investigated and validated against experimental data; nevertheless, the results were not that consistent for some models, e.g., those for the General Motors (GM) model. As a part of the problem, the data acquisition and calibration techniques were not advanced then. The past few decades have seen remarkable advancements in these techniques, e.g., the use of the differential Global Positioning System (GPS) for position measurement, the use of Doppler's principle for speed measurements, and the use of genetic algorithms for optimization. It might be useful to reassess some outstanding issues in car-following dynamics in light of the latest technological advancements. This paper attempts to investigate car-following dynamics on the basis of the real-time kinematic GPS data collected from test track experiments. The GM model was evaluated along with some well-known simulation models, including the Gipps model and the Leutzbach and Wiedemann model. A genetic algorithm-based optimization technique was adapted for calibration. The sensitivities of drivers to their speeds and spacings from the vehicle ahead were found to vary among drivers. The interpersonal variations in model performance were significant. The GM model parameters were identified with improved reliability. The stability of traffic flow was analyzed experimentally.