Urban Expressway Traffic State Recognition And Forecasting Based On Taxi GPS Data

Traffic congestion is not only a problem for traffic managers,but also a hot topic for discussion.Faced with the increasing number of vehicles and the basic formation of the existing road network,it is a "fantasy" to alleviate traffic congestion by increasing the capacity of roads.c and forecasting of the traffic state fast and accurately has become an important means to solve traffic congestion.As the "backbone" part of urban road network,urban expressway bears a high proportion of urban traffic volume.If the problem of urban expressway congestion can not be solved,it is likely to evolve into the traffic congestion of the whole road network.Starting from solving the actual traffic congestion problem and combining GPS data with traffic flow theory analysis,this paper establishes a model of recognizing and forecasting the traffic state,which includes the following main contents:1.Acquisition of valid expressway GPS data.Firstly,the characteristics of GPS data are analyzed.Secondly,a GPS data repairing method is proposed for redundant data,shifting data,missing data.Then the spatial optimization of GPS data is carried out with the help of map matching algorithm based on the ArcGIS platform.And then,the RBF-SVM classification algorithm is used to effectively separate the expressway data from the road surface data,so as to extract the effective GPS data on the expressway,which provides the data basis for the analysis of the expressway traffic state.2.Recognition of the expressway traffic state.Firstly,the travel speed and velocity fluctuation are selected as the characteristic parameters based on the analysis of the characteristics of GPS.According to the theory of mathematical statistics,the minimum sample size required for actual traffic state division is determined.After that,the methods of the road section length selection and the determination of the time interval are put forward through case study.Then the relationship between taxi sample travel speed and real road travel speed is determined by regression analysis.Ultimately,a model of recognizing traffic state is established by using the improved fuzzy C-means clustering algorithm based on genetic algorithm,and the model is verified by experiments.3.Establishment of the expressway traffic state forecasting model.Based on the analysis of traditional neural network,LSTM neural network is selected as the method of forecasting the traffic state.Then the spatial-temporal correlation of GPS data is analyzed respectively.Considering the impact of ramp,a travel speed forecasting model based on LSTM neural network is established.The traffic state of the next time interval is forecasted by the forecasted travel speed and the speed variance of the current time interval.Finally,this paper designs the experimental scheme of traffic state forecasting,and makes a comparative study of the experimental results.Based on the taxi GPS data of Changchun City,this paper analyzed the urban expressway's traffic state recognition and forecasting model respectively.The traffic state was divided into four states: severe congestion,congestion,slow-moving and unblocked.The traffic state of different locations in working days was forecasted.The test data consist of 3840 sets of data,and the accuracy rate is 98.67%,which confirms the validity of the model.And the results of this study have certain theoretical and practical application value for the effective traffic management of urban expressways.