Research On Prediction Of Urban Road Network Travel Time Based On Grid Area Traffic State

Travel time can directly reflect the traffic state of road sections.Accurate travel time estimation can provide the perfect routes for travelers with lower probability of traffic congestion.Current estimation of travel time mainly focus on expressways or motorways,and many travel time prediction models only provide expected travel time.Based on this,this paper studies the travel time problem based on traffic conditions.This paper divided the traffic state of the road network on the basis of the floating car data,and proposed a joint travel time probability density model based on different traffic conditions,and estimated the travel time based on examples.Firstly,this paper constructed a regional traffic state discrimination method based on macroscopic fundamental diagram after pre-processing the floating car data.The method used total travel time and total travel distance instead of weighted traffic and weighted density to construct a macroscopic fundamental diagram,and used Gaussian Mixture Model method for clustering.The results show that the traffic states in the study area could be classified into three categories: free flow,mild congestion and severe congestion,and the speed and density intervals for the corresponding traffic states were obtained by means of function fitting.Next,the frequent congestion areas in the study area were identified and the travel speed of the routes were predicted.Based on the regional road network gridded by squares,a density-based improved DBSCAN algorithm was used to identify the frequent congestion areas;different data sets based on three sets of processing were designed,based on which the travel speed of the paths were predicted using long and short time neural networks.The results show that the multiple averaging model with data smoothing gave the best prediction results.Finally,the travel times reliability of different state grids were analysed and a joint travel time probability density model based on different traffic states was developed.As the probability density functions of different distributions cannot be directly summed up,the corresponding derivation was carried out in this paper.The analysis on the travel time of grids in different states indicate that the grid travel time reveals different distribution characteristics for the three traffic states,and the best distribution functions are Gamma distribution,Weibull distribution and lognormal distribution for free flow,mild congestion and severe congestion,respectively.Taking the roads in Beijing as an example,the travel times were estimated,and the results show that the longer the path distance and the smaller the time intervals division,the higher the estimation accuracy.