Traffic Information Collection Based On Information Fusion Research

Intelligent transportation system (ITS) is an information-based system, where each sub-system and its function attach most importance to the application of traffic information. As a result, the validity of ITS depends highly on the quality of traffic information. To collect real-time, all-around, and accurate traffic information is the key of making urban transportation intellectualized and the important premise of implementing ITS successfully. This dissertation makes researches on the information fusion of urban traffic information. Several related research works, such as FCD (Floating Car Data) algorithm and traffic information prediction etc., have been studied in this dissertation as well.The main works in the thesis are introduced as follows:1. To study the basic theory of information fusion technology. From a point of view, the hierarchy, the functional model, the structural model and the mathematical model of information fusion are discussed, respectively,2. To study the traffic information collection method based on FCD.(1) A vehicular speed estimation algorithm based on FCD is proposed.(2) An arc-based Dijkstra route searching algorithm is proposed. In the algorithm, a prior knowledge and the restricted traffic information of the road network are included in the structure with arc relation, and the searching efficiency and accuracy is improved. This algorithm can be used both in the vehicular speed estimation and in route planning.(3) The vehicular speed prediction algorithm is applied to traffic information collection in urban area on a large scale.3. To study the traffic information fusion based on D-S evidence theory and Kalman filter.(1) A GPS/DRS/MM integrated navigation system is introduced in order to overcome the "navigation blind region" problem and supply high precision to vehicle navigation in urban area.(2) A traffic information fusion method by using evidence theory and least squared support vector machine (LS-SVM) is proposed. By integrating the traffic volume information induced from loop detectors and the traffic speed information collected from floating cars data system, more complete and accurate traffic speed information is obtained. Firstly, the relationship curve between traffic speed and traffic volume is obtained by LS-SVM regression, and the critical speed of this curve is induced. According to the above critical speed and the association matrix which converted the loop detector's volume into traffic speed, we define the basic probabilistic functions of the two sources. Then the D-S theory is used to integrate these two sources of traffic information and new traffic speed is obtained.4. To study the traffic information prediction based on an RBF neural networks.(1) A genetic gradient based learning algorithm for RBF neural networks is proposed. This algorithm makes full used of the global optimizing capability of GA and the local searching capability of gradient descent algorithm.(2) An RBF neural network predicting model is proposed, and we applied it to traffic speed information prediction.