Travel Time Reliability Of Urban Road Based On Statistical Distributions

With the rapid development of global economy and the increasing needs for the timely service and demand,while people are traveling,they not only consider the average travel time,but also pay more attention to smooth,safe and on time to arrive destination.Travel time reliability is considered to be an important factor when people make choices on travel routes,methods and departure times.Moreover it is an major indicator for evaluating the performance of road network,thus it provides related information about the travel time reliability for travelers.It helps to decrease unplanned travels,avoid the traffic jam,choose the suitable travel times and routes and make the traffic situation more stable and faster.Meanwhile,it is beneficial for managers to implement new transport policy to insure smooth traffic and cut the environmental pollution.At present,among most of the researches,time data were dealt with by the methods of normal distribution and log-normal distribution.The travel time reliability was judged by the nature of these distributions,such as the variance,standard deviation,variation coefficients.However,as many scholars were working on the deeper analysis regarding travel time data,they realized that the travel time distributions are asymmetrical and often show a large positive bias characteristics of the long tail,some of which show obvious bimodality.Therefore,travel time data can not be well fitted by the normal distribution and the lognormal distribution.In this paper,the performance of urban road networks is assessed on the basis of the real travel time data,the knowledge of probability and statistics reliablity theories and the theory of system reliability is used to discuss the travel time reliability of urban road network.It also sets up travel time reliability models of urban road network and the problems of travel time reliability of urban road are analysed quantitatively and qualitatively.The paper provides the theoretical basis for the management operation and service of urban road transport network.On the basis of the travel time reliability theory,the existing index system and evaluation methods are analyzed.And because of the commonly used distributions can not fully describe the actual travel time data characteristics,we try to look for some distributions which fit the urban travel time at best and builds some models to evaluate the travel time relibility of the urban road.A power log normal distribution is provided to deal with the skewed and heavy tailed travel time data.With the better properties of this distribution,it not only can capture the characteristics of travel time data,but also has a clear definition of of the travel time reliability indexes.So it provides a useful model for the evaluation of urban road traffic performance.In addition,aiming at the bimodality which was showed in the real travel time data.The comparison and analysis of the commonly used statistical methods of bimodal normal distribution and the new proposed alpha skew normal slash distribution for fitting results of actual travel time data was made.Through experimental verification,the alpha skew normal slash distribution can represent the characteristics of the travel time data.It also found out the factors which will affect the travel time distribution for the data of given road links,such as the ones with traffic signals,the lengths and the delay on the crossing,which will make the bimodality among the travel time distribution.Based on the properties of the power log normal distribution and the alpha skew normal slash distribution.The travel time reliability is expressed by the buffer time index,and the relationship between the traffic congestion index,the lengths of the link and the buffer time index is considered.And the empirical equations which are related to the travel time reliability,the lengths of link and the congestion index are built by means of multivariate regression analysis,and the judgment of its effects has been analyzed as well.The travel time and travel time reliability are measured in the form of currency,and the generalized travel cost is defined as the weighted sum of the value of travel time,the value of travel time reliability and monetary cost is developed.Then,a user equilibrium model which was based on the generalized travel cost is established and a solution algorithm is designed,which provide theoretical basis for traffic flow prediction and traffic demand management.This paper analyzes the characteristics of urban road travel time,according to the long tail and bimodal characteristics of travel time data,the fitting test was carried out using different statistical distributions.The results show that power lognormal distribution and alpha skew normal slash distribution are the best evaluation model.And the travel time reliability is calculated in detail.Based on the properties of the best fitting distribution,the travel time reliability and the regression equation of the link length and the congestion index are established with the aid of R and Matlab statistical analysis software.The results indicate that the congestion index has a positive effect on reliability,while the length of the road has negative correlation with it.The travel time reliability data of the commuter and the monetary value of travel time reliability are obtained by using the questionnaire survey method,and the generalized travel cost is determined.The research of this paper provides a new way for the evaluation of travel time reliability,and the evaluation model of travel time reliability is further refined.It provides effective reference for people's travel decision and traffic demand and management.