Stochasticity Analysis And Modeling Of Car-following Behavior

Traffic system is a typical complex system,in which the driver is the most important decisive factor and the most uncertain factor among all factors.Driving behavior is uncertain,as drivers will make different decisions and produce different driving behaviors under different motives and external conditions.If it is simplified to a deterministic problem,there will be principled errors,and the research results will also be problematic.The stochasticity of car-following behavior is a kind of uncertainty of driving behavior.It plays an important role in the evolution of traffic flow.It not only directly affects the driver’s car-following process,but also is closely related to the emergence and development of traffic oscillation,traffic flow breakdown and other phenomena.Due to the lack of appropriate analysis indicators,there is a lack of research on the stochasticity of car-following behavior from the micro quantitative point of view,resulting in the incompleteness of the micro characteristics and mechanism of car-following stochasticity,and the lack of integrity and reliability of car-following stochasticity modeling.With the rapid development of trajectory collection technology and data analysis technology,the analysis of car-following behavior based on trajectory data and the construction of car-following model in line with the characteristics of drivers have gradually become an important way to understand car-following behavior.Therefore,based on this,the thesis carries out the stochasticity analysis and modeling of car-following behavior on the basis of trajectory data.The main research work is as follows:(1)Stochastic analysis of car-following behavior based on trajectory dataIn the framework of Newell model,combined with the experimental and empirical car-following trajectory data,the stochasticity characteristics of car-following behavior are measured and analyzed by using the wave travel time.It is found that the wave travel time is lognormal distribution;The evolution of wave travel time in a single trajectory is stochastic;It is considered that the stochasticity of car-following can be reflected by the changing rate of wave travel time,which follows the mean recovery process.The influencing factors of car-following behavior are analyzed from three aspects: the position of vehicles in the platoon,the velocity of the leading vehicle and traffic flow state.(2)Build a simulation model integrating the stochasticity characteristics of car-following behaviorBased on the trajectory data and the analysis of the shortcomings of the classical car-following model,a Newell model integrating the stochasticity analysis results of car-following is constructed,and the stochastic Newell(SNewell)model is analyzed through theoretical deduction and numerical simulation.Through theoretical deduction,it is proved that the SNewell model can reproduce the concave growth pattern of traffic oscillations;Through numerical simulation,four different scenarios are simulated,which is proved that the SNewell model has the characteristics of reproducing the empirical traffic flow.(3)Calibration and validation research on SNewell modelBased on the analysis of the shortcomings of existing model calibration and verification methods for SNewell model,an improved method is proposed.The shortcoming of objective function stability and calibration efficiency is solved by constructing the mapping between stochastic parameters and the number of runs required to obtain stable results;By optimizing the measure of performance,goodness of fit function and function form in the objective function,the shortcoming of stochasticity failure is solved.The effectiveness of the proposed calibration method and SNewell model for empirical trajectory fitting is verified by a numerical example in which Genetic Algorithm is chosen as the optimization algorithm.Based on the trajectory data,the thesis makes an in-depth study on the stochasticity of car-following behavior from the perspective of wave travel time.The research is carried out according to the main line of "trajectory analysis-simulation modeling-model verification".In theory,the thesis provides a basis for micro quantitative research on the stochasticity of car-following behavior and understanding the stochasticity of car-following behavior and its internal relationship with traffic state;In practice,the proposed car-following model and calibration method provide a powerful tool for improving the description of car-following behavior and real traffic flow state in micro traffic simulation,and provide support for the formulation of traffic management and control measures.