Analysis And Modeling Of Stochastic Characteristics Of The Car-following Behavior On Urban Expressways Based On Trajectory Data

With the rapid growth of urban residents' travel demand,the urban traffic congestion has become one of key problems restricting the urban development.The traffic congestion not only increases the time cost of travelers,but also worsens the urban air pollution and seriously affects the sustainable urban development.As a representative city in China,the road traffic in Beijing has typical mixed traffic flow characteristics.Mixed motor-and non-motor-vehicles as well as numerous Chinese style traffic violations seriously affect the road traffic safety and the motor-vehicle passing efficiency.The traffic jam caused by traffic chaos has become an important factor aggravating the traffic congestion.However,due to limitations by problems such as the data collection and the difference of the traffic environment,existing driving behavior models calibrated with local model parameters in applications cannot fully capture the characteristics of site-specific driving behavior and the traffic state evolution characteristics in China.Particularly,these models lack in-depth representation and understanding of the characteristics of the traffic instability and the impact mechanism of stochastic characteristics of driving behaviors.The analysis of the site-specific micro-driving behavior characteristics and their impact in Beijing can help understand the intrinsic relationship between the traffic congestion and the driving behavior at the micro level.In addition,it can provide a strong support to improving the ability of the traffic simulation to depict the dynamic traffic congestion characteristics and develop effective congestion mitigation strategies.The objective of the research in this dissertation is to establish a quantitative relationship between the stochastic characteristics of the micro-driving behavior and the evolution patterns of the macro traffic state by analyzing the stochastic characteristics of the site-specific car-following behavior and modeling the driving behavior in the context of mixed traffic flow.The research is intended to be used for evaluating the impact ofstochastic characteristics of the micro-driving behavior on the stability of the traffic state,so as to provide a support to the study of urban road traffic simulations and the congestion mitigations.The research contains the following main components:(1)The vehicle trajectory data were extracted and reconstructed.First,the video data of motor vehicles on the basic segment of the expressway in Beijing were collected,and the vehicle trajectories were extracted.Then,an approach to the cleaning and reconstruction of the vehicle trajectory data considering the restrictions of the driving behavior characteristics,resulting in the trajectory data set of vehicle driving behaviors for Beijing.(2)The stochastic and asymmetry characteristics of the motor vehicle's car-following behavior were analyzed quantitatively.First,the probability distribution functions for the following speed,spacing headway,and time headway were established respectively based on the speed intervals.Then,the Dynamic Time Warping Algorithm was used to extract the perception-response time and the disturbance propagation speed of the car-following behavior.Thus,the asymmetry characteristics of the driving behaviors under acceleration and deceleration states were verified comparatively.Further,as a discussion case,the difference of the driving behavior characteristics between Beijing and L.A.and its impact on the traffic state were analyzed through a combined use of NGSIM data.(3)Two stochastic car-following models were improved respectively for different application scenarios.First,the maximum speed constraint condition based on the speed-spacing headway relationship was proposed in light of the defect of the speed constraint in the Gipps model.Then,the safety factor is defined based on the steady state condition to describe the estimated difference of the expected decelerations between the leading vehicle and the following vehicle in the Gipps model.Furthermore,an improved Gipps car-following model was proposed to describe the stochastic characteristics of drivers by means of the probability distributions of the safety factor and the jamming headway.The improved model can capture the external heterogeneity and the internal randomness of drivers.Subsequently,a stochastic Newell car-following model considering the driving behavior stochastic characteristics was established based on the probability distribution of the disturbance propagation speed,which improved the ability of the Newell model to describe the stochastic characteristics of the traffic disturbance and time headway.(4)The quantitative relationship between the micro driving behavior and the macro traffic state was derived analytically.First,the quantitative relationships between the traffic disturbance and the traffic wave,as well as the traffic wave and the expressway capacity,were established respectively based on the density-flow relationship.It provided that the longer the average perception-response time of drivers is,the lower the capacity is.Further,the dynamic characteristics of the traffic breakdown at the bottleneck and the congestion dissipation rate were analyzed using the RTMS data.The Weibull probability distribution was used to fit the probability distribution functions of the pre-breakdown flow rate,the minimum flow rate and the recovery flow rate respectively.It was found that the average recovery flow rate decreased by 5.4% compared with the average pre-breakdown flow rate.(5)The probability model of the traffic breakdown that captured the stochastic characteristics of the micro driving behavior was established.First,based on the improved Newell model,the probability model of the traffic breakdown was established and the parameter sensitivity analysis was conducted.Then,the impact of the stochastic driving behavior on the congestion duration was studied through a case of the simulation.The simulation results showed that the more stochastic of the driving behavior would cause not only the higher probability of traffic breakdown but also the higher probability of a longer congestion time.Hence,the simulation that neglects the asymmetry of the driving behavior would result in an underestimation of the probability of the traffic breakdown.From the perspective of the congestion caused by the chaos,this dissertation analyzed the stochastic characteristics of the driving behavior quantitatively.It provided the stochastic car-following models for different application scenarios.Further,it established the probability model of the traffic breakdown.The combined effort of these fundamental research work revealed the internal relationship between the micro driving behavior and the macro traffic state,thus providing a new support of approach to the studies on the urban traffic congestion and and the congestion mitigation.