旅行時間預測之研究：模擬指派模式之應用

Development of Simulation-assignment Based Travel Time Prediction Models

旅行時間預測為先進交通管理系統（Advanced Traffic Management Systems,ATMS）與先進旅行者資訊系統（Advanced Traveler Information Systems,ATIS）應用中重要的課題，其結果可運用於市區路網及交通走廊中避免擁擠、事故，並增進整體路網之效率，相關研究亦日趨重要。本研究發展模擬指派為基礎之旅行時間預測模式，以DynaTAIWAN為基礎提出兩種旅行時間預測方法，分別為車流為基礎（flow-based）以及車輛為基礎（vehicle-based）之模式，主要針對路口號誌之市區交通走廊路網，並可適用於壅塞及事故路段。本研究利用動態旅次起迄推估模式建立O-D資料，以DynaTAIWAN模式模擬車流之移動，並使用實際電子收費（Electronic TollCollection,ETC）資料加以驗證模式之精確性，經由平均絕對誤差（Mean Absolute Percentage Errors,MAPE）以及均方根百分誤差（Root Mean Square Percentage Errors,RMSPE）顯示，車輛為基礎模式之RMSPE小於26％，MAPE小於20％；車流為基礎模式之RMSPE小於12％，MAPE小於10％，本研究所發展之模式具有合理之預測結果。

Travel time prediction on traffic corridors and urban arterials is important as it can allow motorists to avoid congestion and incidents as well as improve network efficiency. It is also a basic output component in application of Advanced Traveler Information Systems (ATIMS) and Advanced Traffic Management Systems (ATMS). This research aims at constructing a simulation-based travel time prediction model for traffic corridors. The simulation-assignment model, DynaTAIWAN is utilized to predict the travel time based on two approaches, a flow-based model and a vehicle-based model. Dynamic origin-destination (O-D)estimation and prediction procedure is developed to prepare O-D demand data,and the estimated O-D flows are used within DynaTAIWAN to simulate vehicle movements. The developed framework is illustrated for a traffic corridor and validated through empirical data. Empirical data for signalized urban network and the travel time from electronic toll stations are used to validate the model.Mean absolute percentage errors (MAPE) and root mean square percentage errors (RMSPE) are less than 20% and 26% for the vehicle-based model, and less than 10% and 12% for the flow-based model, respectively. The results show the proposed models can produce accurate predictions with minimum mean absolute percentage errors and root mean square percentage errors.