A Car-Following Model Considering Preceding Vehicle’s Lane-Changing Process

Dihua Sun,Min Zhao,Xuan-Jin Wang,Shi-Hui Wang

doi:10.1109/access.2019.2924659

Abstract

Car following and lane changing are two common driving behaviors in the traffic flow. Preceding vehicle's lane changing is affected by the surroundings and will have a greater influence on the followers' driving decision. The existing car-following theory does not fully take it into consideration that the followers' driving behavior may change during a lane-changing process. In order to reflect the driving decision in a complex traffic flow more precisely, the influence on the following vehicle during the preceding vehicle's lane-changing process is studied. First, the different types of stimulus during the preceding vehicle's lane-merging (LM) process and the space gain effect produced by the preceding vehicle's lane-passing (LP) behavior are analyzed. Then, the LM-FVDM and LP-FVDM are proposed based on the classical car-following model-FVD model. Finally, the linear stability theory, numerical simulation, and NGSIM data sets are used to analyze and validate the performance of the LM-FVDM and LP-FVDM. The numerical simulation results show that the model can reasonably reflect the driving decision of the following vehicle in various scenarios, and verification based on NGSIM shows that the $R$ -squared of vehicles' speed and distance is significantly better than the FVD model, which can more effectively reflect the speed adjustment process of the following vehicle during the preceding vehicle's lane-changing process in the real traffic flow.

Highlights

The car following model is the basis of microscopic traffic flow theory, which is mainly used to describe the following vehicle’s driving response caused by the change of preceding vehicle’s driving state in a single lane, where the overtaking behavior is limited
According to the applicability of car-following behavior, car-following models can be divided into single-lane carfollowing models (SLM) and multi-lane car-following models (MLM)
MODEL VALIDITY VERIFICATION BASED ON NGSIM Using GA genetic algorithm to correct the basic FVDM model parameters, the membership function [30] to identify drivers’ competitive coefficient, the NGSIM data set as the input of LM-FVDM and LP-FVDM, and the longitudinal speed of the following vehicle is output

Summary

INTRODUCTION

The car following model is the basis of microscopic traffic flow theory, which is mainly used to describe the following vehicle’s driving response caused by the change of preceding vehicle’s driving state in a single lane, where the overtaking behavior is limited. Cnl keeps the car-following state with the preceding vehicle C(n+1)l before C(n+1)l reached the target lane, and the driving behavior of Cnl is determined by the distance xnl,(n+1)l , the speed difference vnl,(n+1)l , and lateral distance Wn,n+1. B. MODEL PERFORMANCE UNDER LM-2 STIMULUS Adjust the initial space xnl,(n+1)l+1 (t0) > 15m and xnl,(n+1)l+1 (t0) < 46.25m, other conditions are the same as an experiment (1), in which the initial spacing and the competitive coefficient of the following vehicle are set as shown in TABLE 1.The reaction of the following vehicle under the control of different parameters of FVDM and LM-FVDM. The reaction degree to the preceding vehicle’s lane-changing process is different: some drivers are more conservative to ensure the safety, while others are more competitive to strive for driving space ahead when the LP vehicle is changing lane. The sensitivity coefficients in our model can reflect the different competitive levels of different drivers in the actual traffic flow, and at the model ρ can make a correct response to the lateral separation Wn,n+1 of the preceding vehicle

MODEL VALIDITY VERIFICATION BASED ON NGSIM

EFFECT OF LP-FVDM

CONCLUSION