Research On Traffic State Forecasting Towards Urban Freeway

With the rapid growth of the socio-economic, urban traffic congestion has become acute. Construing the traffic state reasonablely is an important basis for collaboration of Traffic Control and Route Guidance System. By predicting the perspective of traffic state and releasing it as guidance content, will support traffic management and department to make a correct decision and help the traveler to reduce their travel time and improve their travel quality, thereby reducing road congestion.Urban freeway as the main part of the road network, is taking on the majority of urban transportation, thus traffic state of the freeway responses to a certain extent of the traffic status of the whole network. In this paper, based on the data collected by RTMS, the thesis studies the method to forecast traffic state towards freeway with LOS, the main research work are as follows:First, the state of traffic flow data collection and pretreatment have been studied, and then under statisticsing and analyzing the data of traffic flow in different cycles, different dates, different time period ,the thesis concludes time-varying law of the traffic state, which provides a theoretical support and data infrastructure for sub-mode forecasting.Second, after classified the traffic state of Beijing freeway, the thesis puts forward sub-mode ARIMA forecasting model, which forecasts the traffic flow in order to predict the traffic state indirectly corresponding LOS evaluation methods. The forecasting model not only achieves a good experimental result in predicting traffic state, but also in traffic flow. The model suits the multi-object forecasting.At the end, the paper proposes a new forecasting method based on the maximum entropy model with temporal and spatial characteristics. In order to suit time-varying traffic state, the paper also researches the sub-mode maximum entropy forecasting model considering the introduction of Macro-LOS and spatial connectivity matrix, which not only achieves the higher accuracy reliability in short-term prediction, but also gets the better adaptability in the long-term traffic state forecasting across the experiment upon a section of Beijing freeway. But the model only forecasts the traffic state, thus it mainly suits the single target.