動態旅次起迄需求推估方法之研究：流量模擬法

ESTIMATION OF TIME-DEDENDENT ORIGIN-DESTINATION DEMANDS THROUGH FLOW SIMULATION APPROACH

依時性旅次起迄資料是先進交通管理系統 （ATMS） 中動態交通指派的重要資料，主要運用於路網中的交通流量預測並進行交通控制與管理，相關研究近年來皆持續發展。本研究應用流量模擬法模擬各單一起點到達多迄點流量 （One-to-All flow simulation） 進行非指派為基礎之動態旅次起迄比例值推估。模式主要特色：(1)單一起點流量模擬法提供額外路段流量資訊；(2)以迄點為基礎之周界線流量，提供僅利用起點偵測路段流量推估動態O-D比率之方法；(3)路徑旅行時間利用車輛軌跡檔，以起、迄點間最短路徑上所通過之車輛空間平均速度為計算之基準，適用於壅塞之路段以及市區號誌化路網。模式應用上，本研究建構三個路網進行模式應用與敏感度分析，包括：小路網、市區號誌化路網、市區與高速公路混合路網。結果顯示，模式於市區號誌化路網之均方根誤差 （RMSE） 值約為0.047，本研究所提出之動態旅次起迄推估模式具有合理之結果。

Time-dependent Origin-Destination (O-D) demand data is critical to thesuccess of Dynamic Traffic Assignment in Advanced Traffic ManagementSystems (ATMS), and extensive research has been conducted in the past decadeto tackle this problem. This research applies the One-to-All Flow Simulationmethod to generate the single origin to multiple destinations flows andaggregates simulated flows with detector flows to estimate possible O-D splitsbased on the non-assignment based model. The unique features of the modelinclude: (1) the One-to-All Flow Simulation, which is utilized to calculate theO-D splits supplies the additional link counts data; (2) the destination-basedCordon-line supplies a method to estimate O-D splits with sparsely-distributeddetectors. (3) the path travel time is calculated based on the space mean speedof all vehicles taking the shortest path from origins to destinations, suitable forcongested links and a signalized urban network; Numerical experiments toillustrate the proposed model are conducted in three networks: a small testnetwork, a signalized urban network and a mixed network, and severalsensitivity analyses are performed. The numerical results show that the rootmean square error (RMSE) of the signalized urban network is 0.047; theestimated O-D splits from the proposed model are reasonable and acceptable.