Study On Reliability Of Path Travel Time Base On Copula Model Of Variable Correlation Structure

Path travel time is very important for real-time and reliable route guidance,and can provide theoretical basis and technical support for complex traffic management and people’s travel.Based on the data of floating vehicle and considering the time and space dependence of route travel time,this article proposes a method of route travel time prediction.In order to evaluate the reliability of route travel time,the copula route travel time reliability model based on variable correlation structure is builded.The mixed copula function is used to replace the single copula function to describe the relationship between variables with sophisticatedand variable correlation patterns.The defect that the single copula function can only represent the relationship between variables with fixed dependence structure is solved.Firstly,the floating car data preprocessing.Understand the detailed information of floating car data and analyze the feasibility of floating car data for travel time in terms of time and space;based on the characteristics of floating car data in time,space and speed,analyze the r easons for data exceptions and design abnormal data processing methods;Due to the floating car positioning error,the data is corrected based on the map matching algorithm;in order to obtain the historical travel time sample data,the length of the vehicle’s coincident trajectory coverage is considered,and a travel time parameterless estimation method is proposed.Secondly,considering the time-space dependence of path travel time,a path travel time prediction algorithm is proposed.The algorithm takes the travel time data modeled by the path structure as input,uses the convolutional neural network to capture the spatial dependence of the path,and uses the long and short-term memory network to capture the time dependence of the travel time.At the same time,due to the time drift of data,attention mechanism is introduced into the long-term memory network to correct the effect of time drift.The final series fusion model predicts the path travel time of the next slotThirdly,considering the correlation between the travel time of road segments,the reliability model of path travel time is established.Based on the rank correlation coefficient,the model analyzes the non-linear dependence between the travel time of continuous links,select the mixed copula function which can describe the uncorrelated mode to aggregate the travel time distribution of each road segment to estimate the travel time distribution of the route,and then establishes the path travel time reliability model.Finally,an example analysis of the reliability model of path travel time.In the path travel time prediction algorithm,the true travel time of 500 routes in Harbin is verified,and the average accuracy is more than 90%,which is 18.6% and 22.46% higher than the machine learning algorithm in average absolute error and decision coefficient,respectively.In the example verification of the path travel time reliability model,the path travel time distribution based on the M-Copula model is closer to the empirical path travel time distribution than the convolution model and the single Copula model,and can accurately describe the change relationship of the path travel time reliability.