| Summary: | 博士 === 國立成功大學 === 交通管理學系碩博士班 === 95 === This research builds a travel time forecasting model with sequential update of accident duration by fusing a variety of traffic data with Artificial Neural Networks. To consider the influence of accident to the travel time forecasting, the sequential update of accident duration forecasting model is built first. The forecasted duration can be renewed with the updated traffic data throughout the duration of the accident. The output of accident duration model, forecasted accident duration, will become one of the inputs to the travel time forecasting model. The travel time forecasting model constructs a functional relation between real-time traffic data as the input variables and real bus travel time as the output variable. Real-time traffic data are collected from the global position systems (GPS) on board of the intercity buses, vehicle detectors (VD), and accident databases. These two models will consider the time-space relationship between the traffic data and the accident to represent the traffic propagation.
To improve the model performance and save the cost in data collection, the effect of data feature reduction to the model is assessed. The methods considered for data feature reduction include the composition, cluster and selection with Genetic Algorithm. For accident occurrence and no-occurrence uses, the effect of travel time forecasting model will be assessed respectively. To reflect traveler behavior closely, partitioning the freeway into links for model development is considered a proper approach. Once the link travel time forecasting model has been built, the forecasted path travel time will be evaluated by summing the forecasted link travel time to fulfill the user’s trip characteristic.
The features of this research are considering the forecasted accident duration into travel time forecast, discussing the methods of data feature reduction and sequential update of forecasting information. This study shows very promising practical applicability of the proposed models in the Intelligent Transportation Systems (ITS) context.
|
|---|
