Study On Dynamic Prediction Of Road Operating Speed Based On Floating Car Data

Road operating speed prediction is the basis of traffic control and management which is one of the key technologies to realize Intelligent Transportation S ystem.The road operating speed is an important parameter of traffic state,accurate,reliable.Dynamic and real-time road speed prediction is direct impact of travel time estimation,traffic congestion and traffic state identification index estimation.The traditional road operating speed prediction is mainly based on static prediction,and the timeliness is poor;the traditional research model is mainly based on the small passenger car,ignoring the fact that the urban road is co mplicated,and the operating features of different models is different;the traditional speed algorithm including linear regression,grey prediction model and Neural Network,the majority did not consider the problem of noise in data collection of floating car which cause the prediction accuracy is not ideal,and this directly affect the traffic state recognition,road network evaluation and traffic decision-making.Therefore,the study on the road operating speed is not limited to the small passenger car itself,but should start from the whole model,study and analyze the whole vehicle model,then establish a dynamic and real-time model of road running speed for providing the basis for the construction of urban intelligent transportation system.The big data traffic of analysis and application must rely on advanced parameters collection equipment.In this paper,the data of road traffic parameters are collected by using floating car data which is characterized by wide coverage,high precision and large sample size.At the same time,analyzes the setting principle of floating car rate,acquisition time interval,in view of the fact that the original data is easy to be interfered by the environment in the actual collection process,the problems of data loss,distortion and error exist,PMM method for determining data screening methods and missing data filling.Based on the floating car data,this paper established the dynamic predictio n model of road operating speed which mainly includes: aming at the process of collecting data is easy to be disturbed by the environment,the original data are de-noised by wavelet transform,then the wavelet-ARIMA floating car speed prediction model is established,compared with the single ARIMA model without denoised,the average relative error of the proposed algorithm is decreased by 18.17%,and the average absolute error is decreased by 33.82%.The sections are divided into four types: small passenger car,taxi,bus and large passenger car,the FCM-RBF neural network model is established by using the nonlinear approximation ability of neural network.Taking the bus speed and saturation as the input parameters,the model of the neural network is established based on the output speed of the passenger car,the taxi and the large passenger car.The model car prediction average absolute error is 3.95,the average relative error is 9.21%;the average absolute error of the taxi is 4.35,the average relative error is 10.83%;the average absolute error of large passenger car is 2.65,the average relative error is 12.78%,the prediction accuracy of the model has reached the actual application requirements.