| Traffic congestion is one of the most difficult problems in the 21st century. Different approaches have been developed to deal with traffic congestion and manage traffic flow. In comparison with prediction based upon historical datasets only, real-time methods take vehicle operators and travelers into consideration, and develop new algorithms/models to improve accuracy for efficient traffic management. This thesis is going to first highlight current research in traffic flow prediction, and then use chaotic dynamic complexity to discuss the scale-free characteristics of traffic flow. As sharp variation points provide rich information to analyze the fluctuation and sharp variations of traffic flow, it is used in a new method developed in this thesis to guide the classification of historical datasets and to combine real-time datasets from multiple sources of traffic-relevant information. In addition, an augmented reality system is constructed to visualize traffic flow under the influence of different factors. |
