A Comparative Study of Non-Linear Car Following Models in Real-Driving Scenarios

Car following models (CFMs) are the most prominent microscopic traffic flow models that capture the follower behaviour through detailed representation of leader–follower interactions. Models vary in their interaction logic, but it is generally assumed that all established models can produce realistic vehicle responses under real-world driving conditions. In this study, the efficacy of three well-established CFMs—nonlinear Newell model, the Optimal Velocity Model (OVM), and the intelligent driver model is evaluated in real driving conditions represented by Worldwide harmonized light vehicle testing cycles (WLTC). The choice of leader vehicle profile such as WLTC, captures speed variations corresponding to driving conditions such as rural, urban and highway. The model responses to WLTC were investigated for extreme behaviour analysis, characterized by high acceleration or jerk values. Model robustness is compared using nominal range sensitivity analysis and the response surface method, yielding insights into reducing model complexity during calibration exercises. The results reveal OVM to be the least robust model of the three. The findings highlight unphysical and unrealistic model outputs, offering critical insights to inform model selection and guide improvements for more accurate and reliable microscopic traffic simulations.