Travel Time Analysis And Prediction Of Urban Traffic Based On Statistic Model

Travel time prediction is an important content of ITS, and it is the basic of traffic control system and traffic flow guidance system. The road travel time can response the current traffic condition directly, so accurate and real-time travel time prediction can not only induce the travelers choose reasonable travel routes, reduce traffic congestion, improve the efficiency of travel, but also helps to judgment the abnormal traffic status and traffic events. This paper studys the travel time prediction of intersection road under signal control, builds the travel time estimation and prediction model, and get some good results in final.This paper maily includes three parts:first of all, it develops a travel time estimation approach based on vehicle license plate auto recognition system, this method extracts triple information of travel time(Capture Time; Camera Number; Plate), extracts valid data under the limit of space time relationship, and matchs vehicle license plate by NoSQL database technology, this approch has greatly increased the travel time prediction efficiency.The second, this article analysis the traffic state and vehicle trajectories of intersection road under signal control according to traffic flow theory, builds a model for the probability distribution of travel times between any two locations on an aterial link, then estimates the model parameters using the actual travel time data, obtain the probability distribution trend of travel time at last, this information is significant for the travel time reliability analysis, at the same time, we can get the average travel time.At last, this paper proposes a method to predict travel time based on particle filtering. In the algorithm, each particle corresponds to the experienced travel time at one moment on the history day, we calculate and update particle weights by analysing the correlation between traffic flow and travel time, comparing the real time traffic flow with the history flow data, thus get the travel time predictive value. This approach does not need to build state-transition model between prediction travel time and history travel time, just uses the change trend of history data to implement state transition. In the end, we use the travel time data and traffic flow data in different traffic conditions to verify the model, the results shows a good prediction effect.