Urban Road Traffic Flow State Identification And Decision-making Method Research

As urban traffic problem increasingly prominent, the research of Intelligent Transportation Systems has been paid more and more attention. Especially, the research and development of advanced transportation management system and traveler information system are the basis to solve the traffic problem. The accurate method of traffic prediction and state recognition is the key to realize real-time induction, which is also an important content of dynamic management.The identification method of urban traffic flow is researched from traffic flow predictive and real-time traffic conditions, which is proceed with the analysis of characteristics of short-term traffic flow, combining with the basic demand of dynamic management and traffic guidance.Firstly, wavelet-based stochastic traffic flow combination prediction methods is proposed, for existing methods could not meet the prediction accuracy and the need of time scale of prediction by analyzing existing prediction methods of traffic flow on the basis of the randomness and the volatility of short-term traffic flow. The method decomposes short-term traffic flow into random and stability by using wavelet analysis. According to the characteristics, respective time series are predicted by using Markov chains and RBF neural network. The predicted results are combined for the final predictive value. And the effectiveness of the method is checked in the short-term traffic flow prediction by using experimental data.Secondly, considering more extensive traffic flow characteristics, single forecasting method may not meet the needs of the prediction accuracy. Therefore, weight combination forecasting model is applied to traffic flow forecasting. After considering the prediction accuracy and computational complexity, the exponential smoothing, adaptive gray prediction method and short-term traffic flow prediction methods based on wavelet analysis are combined using the least squares method, and also using experimental data to validate combination forecasting methods. Compared to short-term forecasting method which is based on wavelet analysis, in the forecast accuracy, the improvement of combined forecasting method which also increase the amount of computation is not very ideal; At the same time, it verifies the widely adaptability of the forecasting method which is based on wavelet analysis.Finally, several ways of the traffic state division were analyzed and the traffic state decision model based on the prediction was put forward. The model calculates the difference of traffic flow parameters of the predicted value between two adjacent sites during several moments later in a row. And which is used to determine whether there has been traffic congestion and distinguish the type of traffic congestion using the size of difference; The traffic state decision-making model is validated by using of experimental data, and the results show that the model has good performance.