The Study of Traffic Incident Duration Predicting Models in Freeway

• Main Author: 謝乾駿
• Other Authors: 陳昭華
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/38150758058522141604

碩士 === 中華大學 === 科技管理研究所 === 91 === Freeway traffic congestion has been a critical problem for both commuters and traffic system managers in Taiwan. Such a problem is even more serious when there involves an incident. Therefore how to take proper traffic management strategies to relieve the traffic impact at the early stage is a vital issue for traffic managers. As the completion of national freeway system in the near future, freeway traffic network will be larger and more complicated than before. The need for an efficient incident duration prediction model to aid in traffic incident management system to take proper and timely actions after the occurrence of an incident will then also become a critical issue. Hence, how to propose an efficient and cost-effective incident duration prediction model to advance the effectiveness of traffic incident management system will be a vital need. The impact of traffic incident can be composed of a variety of factors and performed in various patterns according to different situations. However, the duration of the impact is dependent upon what the timely actions being taken into response the incident. In general, the incident duration is a process that composes of four periods of time, which are incident detection, verification, response and removal. As the four stages could involve many different factors, it is clear that the effect of each factor to the duration would be interactive and vague to be identified. Hence, after the review of the related literature Radial Basis Function Network (RBFN) and Principal Component Analysis (PCA) that can take multivariate variables and vague interaction relation in parallel is chosen to model the traffic incident duration. Prior to the model construction, this research will begin from the investigation and identification of the critical factors that affect the duration of traffic incident. At the stage of model construction, an RBFN model with proper structure and adaptive nature will be proposed to fulfill the real time prediction requirement. And the RBFN model will be trained and verified by using historical real data organized by this research from related traffic institutes or simulation data generated by using traffic simulation model such as CORSIM in case the real world data are unobtainable. Finally, the findings and recommendations for future research will be proposed.