Development of Freeway Travel Time Forecasting Models with Accident Characteristics Using Data Fusion Techniques

• Main Authors: Cheng-Hsien Hsu, 許正憲
• Other Authors: Chien-Hung Wei
• Format: Others
• Language: zh-TW
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/20054154419935816789

碩士 === 國立成功大學 === 交通管理學系碩博士班 === 95 === 　　With the rapid increase of motor vehicles, freeway has incline to capacity. Once the traffic congestion and accidents occur on freeway, it will cause huge time delay and social cost. When we use real time traffic information to forecast future travel time on freeway, it is the desriable information for Advanced Traveler Information System (ATIS) of the Intelligent Transportation Systems (ITS). Providing the appropriate traffic information would allow drivers to make a decision on appropriate routes and departure time to avoid congestion. This traffic information can help drivers to select the shortest travel time and elaborate the whole effects for freeway network. 　　This research fuses the real time traffic data which involve the GPS data of bus, the data of vehicle detector and the data of accident to build an accident delay forecasting model using artificial neural networks. This accident delay model is thus applied to build the travel time forecasting models of the links of the freeway. 　　After repeatedly testing and revising, the MAPE value of accident delay forecasting model constructed in the research is 10.70%, belonging to “good forecasting” level. The MAPE values of travel time forecasting models with accidents are 5.79% to 28.16%, belonging to “reasonable forecasting” level. The MAPE values of ordinary travel time forecasting models are 8.27% to 14.72%, belonging to “good forecasting” level. Based on the resulting performance, this research can be applied to real-time freeway travel time estimation.