A Study of Dynamic O-D Estimation and Prediction for General Networks

• Main Authors: Chi-yu Chang, 張琪玉
• Other Authors: Ta-yin Hu
• Format: Others
• Language: zh-TW
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/86517911689480136827

碩士 === 國立成功大學 === 交通管理學系碩博士班 === 95 === Intelligent Transportation Systems (ITS) aim to utilize the transportation system efficiently by strengthening the connection between traffic control measures and available information, such as real-time information and historical flow information. Traffic flow distributions are detected by surveillance systems, and the information is transmitted to the traffic management center. However, applications based on static OD flows do not capture the dynamics of build up and dissipation of congestion, time-dependent OD demands are extremely important in Dynamic Traffic Assignment, a core model in ITS to analyze dynamic flow distributions. To generate O-D demand data through field surveys is a time consuming process. Dynamic O-D estimation can save human resources and reduce expense. Most of dynamic OD estimation methods are constructed based on traffic flow counts; however, the interrelationships between link traffic counts and OD are not clear. In this research, a time-dependent O-D estimation algorithm, based on Ashok’s algorithm, is constructed under mixed traffic flow conditions. The framework is a Kalman Filter based approach, and the deviations of OD flows from historical estimates are used in an autoregressive process. The algorithm is a assignment-based model, in which Dynamic Traffic Assignment Models need to be incorporated within the algorithm. DynaTAIWAN, a simulation-assignment model, is used to generate time-dependent assignment results for dynamic OD estimation. Numerical experiments and sensitivity analysis are conducted to illustrate the algorithm in a wide variety of scenarios.