Prediction And Evaluation Method Of Simulation-based Traffic Running

In recent years, many cities are building the intelligent traffic system to meet the gradually serious traffic congestion. The reasonable prediction of the traffic flow and traffic state can provide powerful technical support for the formulation of management scheme of the intelligent transportation system. At present, the real-time forecast of traffic flow has been developed rapidly, all kinds of the prediction method in the prediction has different advantages on accuracy or speed. However, for the same traffic flow may correspond to different traffic state, it is necessary to predict the traffic state.Traffic simulation has become a powerful tool in the traffic field and its effect become more and more obviously. Although the traffic simulation function is powerful, it cannot be used to predict directly, however, once the traffic flow or other data is predicted through the other way and become an input value of traffic simulation, the simulation software can simulate the corresponding period traffic state. Traffic simulation provides a new idea for traffic condition prediction.Fist, Based on consulting a large number of domestic and foreign literature, and summarizes the method to forecast the traffic state, then determine to use an indirect prediction method:By using the secondary development of TransModeler, the short-term forecasting module is implanted into TransModeler. Then the simulation software is run with the forecast value to forecast traffic state.Secondly, analyzing the related knowledge of short-term traffic flow prediction, firstly of all, summarized some kinds of traffic flow prediction model which is widely used and comparative its advantages and disadvantages, then selected kalman filtering algorithm as the prediction model. Considered that kalman filtering algorithm requires the data is continually, but in fact some times the traffic flow data will be leakage or deviated, the historical data are used to smooth and estimate the original value to ensure that the kalman filtering algorithm can recursive continuously with leakage data. Then the next time period data are predicted with the first two period data.Thirdly, based on the analysis of the traffic simulation software of each level, the advantage of TransModeler to predict the traffic state is expounded. Parameters of the simulation software are corrected before running the simulation software; it can make the simulation more realistic. Combined with the characteristics of the output data of simulation software, set up an evaluation system with simulate evaluation index to evaluate the intersection, road, road network state respectively.Finally, select the Xuefu Road and one two one street district of Kunming city as an example to verify the method, using C++programming to accomplish the data extraction and evaluation indexes analysis, it makes work efficiency. Compared with the measured data and prediction data, the error is less then20%. Though the examples, it shows that input the prediction flow to simulation software to forecast the traffic state in next period is feasible.