Research On Urban Road Travel Time Prediction Model Based On OBD Data Map Matching

As one of the effective ways to alleviate urban traffic pressure,Intelligent Traffic System(ITS)has become the direction of urban traffic development.Travel time prediction is one of the core research contents of ITS,which plays an important role in traffic guidance system and signal coordination control system.At present,most of the research on travel time prediction at home and abroad is based on floating car and fixed detector data,rarely using the On Board Diagnostic(OBD)data,while OBD can obtain the traffic data under the actual operation status of vehicles,with large amount and wide coverage,the OBD data is very suitable for traffic prediction research.The main research contents of this paper include:(1)The total amount and the distribution of OBD data are analyzed,in order to improve the quality of OBD data,using threshold method and statistical results to identify and process the outliers and missing values.(2)Based on vehicle positioning error,the electronic map is divided into grids.By setting the index rules of positioning points with roads,calculating and determining the deviation distance and road driving direction,the OBD data are matched to the electronic road network.Based on the endpoint time interpolation estimation method,different calculation methods of vehicle passing through endpoints are set up according to the different locations of vehicle locating points near the endpoints.Taking Beijing Fuchengmen Avenue as the experimental road,the travel time data of 20 working days has been calculated.The reliability of the travel time data is verified by correlation analysis.(3)In the study of urban road travel time prediction,a new Innovation Kalman filter travel time prediction model is proposed.By observing the change of innovation,the noise covariance is adjusted in real time to improve the stability of Kalman filter.At the same time,a travel time prediction model based on cuckoo algorithm to optimize Elman neural network is established,The optimal nest location is introduced to reduce the strong randomness of cuckoo algorithm.The improved cuckoo algorithm is used to optimize the initial value of Elman neural network parameters.(4)The road travel time data obtained from OBD data are used to test the established model.The results show that the mean absolute percentage error(MAPE)of the two models is less than 4%,and the MAPE of Kalman filter adjusted by innovation noise covariance is 3%lower than that before adjustment.The MAPE of Elman neural network prediction model optimized by cuckoo algorithm is 3.3%lower than that before optimization.Compared with genetic algorithm,cuckoo algorithm is more effective in optimizing the initial parameters of Elman neural network.