Research On Road Vehicle Queue Estimation

Recent years, as the development of economy and urbanization, the congestion problem in china become more and more prominent. There are many factors that cause traffic jam. In its essence the problem is caused by unbalance between traffic demand and traffic resource, but from the form of jam, the cause is various, include traffic plan (the bad-using of land), traffic design (irrational set of traffic signal and traffic state, traffic facilities and so on), traffic management (the disorder running of traffic flow). The process is usually from a road bottleneck. A queue is formed at the bottleneck and trace to upstream, and choke the intersection once queue back reaches intersection upstream. If it happens, traffic congestion inevitably diffuse around road network, so called"domino"come into being. From the analysis above we conclude that if we can detect real-time queue back and take some measurement before queue back reach intersection upstream, to some extent the congestion can be avoided, this is what the base point of this research.1st chapter analyze the demand of queue estimation. There is an important characteristic in China's road structure: short average road length and strong relationship of roads. This nature easily causes the problem of queue trace-up. As to the nature, the research contents are designed: queue estimation research and road queue relationship research.Through literature reading, 2nd chapter summarize three theories the queue research based on: micro, macro and integration of the above two. Based on these theories, several methods were formulated: block theory; accumulated curve method; shock wave method and probability method. At last, the merits and demerits were analyzed. The work of this chapter was basis of following research.Because the detectors are still used comprehensively, 3rd chapter research queue back estimation method. In order to obtain real-time queue back position from point traffic information, the traffic state when queue is formed is analyzed first, through the analysis the mapping relationship is defined and hence the estimation method of queue back when arrival is fixed is obtained. But in the real world arrival is not always the same, so we establish estimation method at variable arrival. The result show that in some detect interval (10s) the estimation method work well.4th chapter research the relationship between urban roads. It proceeds from analysis of adjacent road directions. After the analysis two viewpoints is formulated on the relationship between upstream road and downstream road: one is that inputs and outputs; the other is that cause and results. As the traffic is influenced by many factors such as traffic control, traffic management and so on, in order to simplify the problem, we establish two models based on above viewpoints: one is neural network road relation model and the other is that bayesian network road relationship model. PCA, GMM and probability method are used to realize these models. The validation result shows that because of the randomness of road network, bayesian network model based on probability theory can grasp traffic characteristic better.5th chapter summarize the work done in the thesis and show the on-going work. Research achievements include mainly queue back estimation method and road relationship models. These methods can be used into ITS to enhance the reliability of information collection system and decrease system cost.