Research On Short-term Traffic Flow Prediction And Problem Of Path Selection

With the development of society and economy, the intelligent transportation system(ITS) is on flourishing. Intelligent transportation system is mainly to achieve the guidance and control of transportation in real-time, accurate and efficient. It is an acknowledged effective way to solve traffic congestion, traffic accidents and traffic pollution problems. Forecast information is the foundation and key for the city traffic control system of traffic flow, and adaptive path selection is the core technology of the city traffic guidance system.Based on the literatures and previous works, this thesis has studied the two key problems of short-time traffic flow prediction and adaptive path selection in ITS. The main contents of this thesis are as follows:(1) The background and research content of intelligent transportation system is briefly introduced, and the basic characteristics of traffic flow is analyzed, the basic parameters to describe the traffic flow characteristic are given, the significance of the short-term traffic flow prediction and path selection technology is expatiated.(2) Base on analysis of several kinds of short-term traffic flow prediction methods, the weighted combination prediction model for short-term traffic flow is proposed by combining the improved K nearest neighbor non-parameter regression with fuzzy neural network. Therefore, this model has both the strong predictive ability and the learning ability. According to the prediction error at last period time, the weights of the combination prediction model is determined, and the final prediction results are output. According to the actual traffic flow data, simulation results show that the weighted combination model can improve the prediction accuracy, can be used for real-time prediction of short-term flow.(3)To the traffic path guidance system of traffic network, considering the turning randomness, time dependence of road section and intersection, the adaptive path choice algorithm is modified. With an example, the effectiveness of this algorithm is verified.