Research On Stochastic Characteristic Of Traffic Flow And Travel Time Reliability In Highway

In recent years, the increasing amount of travel demand in highway system constantly exceed the designing capacity of highways, which result in deteriorating condition of traffic flow, the increment of travel delay and the decreasing social and environmental satisfaction, especially during festivals or holidays(such as the Spring Festival, the National Day). For travelers, the phenomenon of congestion brings the increase of travel time, seriously disrupts the schedule arrangement and adds travel cost. Therefore, aiming at exploring uncertain of travel time, this dissertation studies the principle of stochastic travel time in highways from the perspective of macroscopic traffic and microscopic driver, then reveals the dynamic of spatial traffic evolution and develops the evaluation model of travel time reliability. The research results provide not only comprehensive information of travel time for traveler, but also the theoretical support for identification and prediction on travel time reliability in highways.Firstly, the internal law of stochastic fluctuation is studied from the facets of macroscopic traffic and microscopic driver. Form driver’s behavior, the concept of time headway in the pattern of car-following is proposed, then this article analyzes the shape of density function of headway. Form macroscopic traffic, the traffic condition is divided into two states that include free flow and congestion state, in which,accordingly, the spatial and temporal evolution models are built based on dynamic headway under different states:(1) In the state of free flow, stochastic process is introduced in the equation of expected velocity to describe drivers’ autonomous control under undisturbed environment;(2) In the state of congestion, the DDH(Driving by Dynamic Headway) model representing compact car-following pattern is developed. The DDH model, based on the rule of collision avoid, operates in three iterative steps: headway update, velocity update, position update. These steps entirely describe stochastic fluctuation of reaction and dynamic motion between leading vehicle and following vehicle. And this dissertation utilizes the MATLAB software to reproduce vehicles’ trajectory in specific environment of segment in highway. The simulation result accords with basic laws and empirical phenomenon in real traffic flow.Then, travel time data is extracted from microscopic trajectories, which is used to fit the distribution function of travel time by EM algorithm. Moreover, travel time reliability is evaluated under the conditions of different travel flow respectively. The results provide theoretical support for evaluating travel time reliability.Further, combined probabilistic characteristic of travel time with traffic flow propagation law of network, the Tree-Augmented Naive Bayesian Network(TAN) is developed to analysis the route travel time reliability. In the aspect of network structure model, based on first-order homogeneous Markov Chain properties, the probabilistic association model is built to represent the relationships among interconnected links. In the aspect of network parameter reasoning, the equations of probabilistic association are deduced by Bayes theorem. Then, to evaluate the route travel time reliability, this dissertation attains posterior distribution of the link or route travel time by priori probability distribution of link travel time and probabilistic association equations.This dissertation utilizes analog simulation method according to dynamic headway and association reasoning of links to establish the corresponding traffic flow models and the method of calculating travel time reliability, which can provide the reference value for transportation systems analysis and travel information service system.