Research on car-following model based on molecular dynamics

Abstract

The car-following model has always been a research hot spot in the field of traffic flow theory. Modeling the car-following behavior can quantify the longitudinal interaction between cars, thereby understanding the characteristics of traffic flow, and revealing the inherent mechanisms of traffic congestion and other traffic phenomena. In fact, there is an asymmetry problem in the driver’s acceleration and deceleration operation. The existing car-following model ignores the difference between the acceleration and deceleration of cars. To solve this problem, the cars driving on the road are compared to molecules with interactions. Based on the molecular interaction potential function and the wall potential function, we construct a molecular car-following model. We use NGSIM data set to calibrate the parameters of the model through the genetic algorithm. Finally, we analyze the evolution rule of the disturbance in the traffic flow in different states with the help of the time-space diagram, and compare the molecular model and the classical optimal velocity model. The results show that the molecular car-following model can better describe the car-following behavior from the micro level.