This dissertation describes a new approach of the online traffic flow modelling based on the hydrodynamic traffic flow model and an online process to adapt the flow-density relation dynamically. The new modelling approach was tested based on the real traffic situations in various homogeneous motorway sections and a motorway section with ramps and gave encouraging simulation results. Traffic flow models have been developed employing various off-line modelling approaches by numerous researchers and provided the basis for understanding traffic flow. Nowadays, traffic flow models are needed for practical traffic control systems as well as for the investigation of characteristics of traffic flow. Sound traffic flow models are a decisive component for the performances of traffic control systems. Traffic flow models developed until now can not be satisfactorily applied to practical traffic control systems because of the transferability and post-prediction problems caused by the calibration of parameters. The parameters of traffic flow models have dominant effects on the simulation results of the models and should be calibrated depending on traffic data sets. These properties of traffic flow models are crucial for the online application of the models. In this dissertation a new modelling approach was developed which can alleviate the calibration problems of the current traffic flow models. It was shown by the model evaluation based on real traffic data that the new model works using various traffic data sets with no additional calibration necessary and gives satisfying accuracy for online traffic control systems. This work is composed of two parts: first the analysis of traffic flow characteristics and second the development of a new online traffic flow model applying these characteristics. For homogeneous motorway sections traffic flow is classified into six different traffic states with different characteristics. Delimitation criteria were developed to separate these states. The hysteresis phenomena were analysed during the transitions between these traffic states. The traffic states and the transitions are represented on a states diagram with the flow axis and the density axis. For motorway sections with ramps the complicated traffic flow is simplified and classified into three traffic states depending on the propagation of congestion. The traffic states are represented on a phase diagram with the upstream demand axis and the interaction strength axis which was defined in this research. The states diagram and the phase diagram provide a basis for the development of the dynamic flow-density relation. The first-order hydrodynamic traffic flow model was programmed according to the cell-transmission scheme extended by the modification of flow dependent sending/receiving functions, the classification of cells and the determination strategy for the flow-density relation in the cells. The unreasonable results of macroscopic traffic flow models, which may occur in the first and last cells in certain conditions are alleviated by applying buffer cells between the traffic data and the model. The sending/receiving functions of the cells are determined dynamically based on the classification of the traffic states by employing fuzzy logic and the shock wave theory. The model is extended to describe also the propagation of congestion in the motorway sections with ramps by considering the capacity reduction caused by the interaction between the traffic flows of the main-stream and the ramps. This research represents the potential of the macroscopic traffic flow models for the application to online traffic control systems by applying the dynamic flow-density relation. The new modelling approach alleviates a critical problem, i.e. the parameter calibration problem, of existing traffic flow models.
«
Login
BibTeX