Research Of Link Travel Time Prediction Based On Floating Car And Fixed Detector

The worsening of urban traffic makes people more and more demanding for the accuracy of traffic information.As a new traffic information collection technology in ITS system,the floating car system has become an effective complement to the traditional acquisition technology.The research of link travel time prediction based on floating car technology can provide accurate travel time prediction information for ITS system,so that it can timely dredge traffic through various countermeasures and improve operation efficiency of transportation system.This paper takes the road travel time prediction as the research goal in which the data preprocessing process of floating car,road travel time prediction based on floating car data and fixed detector and extensional study of traffic state prediction based on travel time prediction were studied.The detailed contents are as follows:Firstly,in floating car data preprocessing stage,for floating car data repair process,a data restoration method was put forward,which combined historical data and real time adjacent traffic flow data.Furthermore,in the stage of floating car data map matching,aiming at the problem of poor matching accuracy of existing map matching methods,this paper analysed the distribution characteristics of the correct matching point pairs under manual matching and the similarity of distance and average speed features were respectively fitted by exponential distribution,then a hidden Markov map matching method based on similarity of distance and average spreed was proposed.Through a series of improvements of the floating car preprocessing stage,it provides reliable data support for floating car data based travel time prediction.Secondly,a floating car data and fixed detector based travel time prediction model-MSDF was proposed.First the law of the accuracy of the estimation of the road average speed by the floating car occupancy rate was studied.According to the experimental results,the distribution law of the Logistic function was fitted.Then,Based on the distribution law,a Logistic based algorithm for calculating the average speed of road was proposed.And then,aiming at the shortcomings of prediction accuracy based on fixed detector and floating car technology alone,a prediction model-MSDF was put forward.The model used weighted average fusion method to fuse the prediction results of the two methods in order to produce higher accuracy and smaller average relative error results.Experiments showed that the proposed model effectively improves the accuracy of travel time prediction.Thridly,based on the proposed MSDF model,the extension research of link travel time prediction was studied,and the traffic state prediction method based on MSDF was put forward.Aiming at the shortage of the existing traffic status prediction method based on fuzzy C-means clustering method(FCM)because of the random selection of the initial cluster center,which leads to the loacl local optimum of the iteration,the characteristics of the initial cluster center needs to ensure that the same type of object closely and that each cluster center separately were used.And then the furthest distance strategy based initial clustering center selection algorithm was put forward to improve FCM.Finally,based on the improved FCM algorithm,the travel time predicted by the MSDF model and the traffic flow obtained by the fixed detector were taken as the input features of the FCM,and a traffic state prediction model considering the travel time was proposed.The experimental results showed that the improved FCM algorithm could achieve more accurate traffic state prediction than the existing algorithms.