Uncertainty Analysis Of Dynamic OD Matrix Estimaiton Based On RFID Data

With the development of informationization construction, advanced data acquisition devices and data processing methods make us get richer traffic information. Wherein, OD (Origin-Destination) flows are the main basis data for modern traffic planning and management. It has been a research hotspot for scholars that how to obtain, optimize and apply OD flows. Aiming at this problem, in the theory of dynamic estimators of OD flows, this thesis focuses on dynamic optimization and forecasting method of OD flows using traffic counts with the unban network vehicle dynamic OD demand flows and traffic assignment matrices computed by radio frequency identification devices (RFID) technology and uncertainty analysis of the forecasting dynamic OD flows. Research contents and results are as follows:In terms of literature, the articles mainly from the following three categories are reviewed:traditional acquisition method not based on traffic counts, OD estimation method using traffic counts and OD matrix estimation method based on data acquisition facilities. A detailed analysis for current situation and characteristics of above researches is given. Considering the necessity for uncertainty analysis in the decision-making process, research status for uncertainty analysis of OD estimation method using traffic counts is introduced. The review points out the current research in this area are still very inadequate.In terms of dynamic estimators of OD flows based on RFID technology, according to the definition defined in Cascetta's paper which was published in 1993, the thesis firstly proposes a computing method of dynamic OD flows and dynamic traffic assignment matrices which are the basic data of dynamic estimators of OD flows. With the deep analysis of obtained data, it is found that the obtained data has a same trend with the real one but the amount of obtained data is insufficient. The obtained dynamic traffic assignments show little change in same time interval of different days. According to the theory of dynamic estimators of OD flows using traffic counts, the thesis proposes the framework of dynamic estimators of OD flows using traffic counts based on RFID technology. Based on the theory of off-line dynamic estimators of OD flows, taking the characteristics of computing data, thesis proposes a GLS optimization model and uses a genetic algorithm to solve it. From the analysis results, it is shown that the proposed model and algorithm can maintain the trend of the raw data and make up the lack of amount of data. Based on the theory of real-time dynamic estimators of OD flows, the state space model and one-step Kalman Filter algorithm are given. Taking the optimized OD flows as the priori information, forecasting process is done. According to the analysis of the forecasting results, it is shown that the given model and algorithm have a good effect.In terms of uncertainty analysis of dynamic OD flows estimation using traffic flows, the thesis proposed a quantitative approach for uncertainty analysis of dynamic OD flows forecasting results. Based on the theory of time series analysis, the approach acquires the 95% confidence interval of forecasting traffic counts and the confidence is evaluated. Putting the upper and lower bounds of obtained 95% confidence interval into dynamic OD flows estimation process, the interval estimators of dynamic OD flows are obtained. Through the reliability analysis of the forecasting interval, it is showed that the optimized results are almost entirely within the interval. The forecasting interval has a good reliability.