傳統之運輸規劃模型具有程序運作特質，稱之為四步驟模式（FSM），而此一傳統的運輸規劃模式也常受到評論與挑戰，其主要來自兩方面：其一為程序式操作之本質所導致之前後程序中數值(例如起迄間之旅行時間)兩者前後並不一致之矛盾現象。其二為傳統模式並未能討論動態之課題。在動態交通量指派之理論與應用日漸成熟後受到強烈之挑戰。因此，傳統四步驟之處理仍以長期觀測穩定狀態反應交通績效之表現，僅可大致顯示空間分佈之狀態，卻無法顯示依時間變化之反應。 本研究以傳統程序性總體程序運輸需求模式並結合動態交通量指派，透過回饋處理並探究其能否改善一致性之問題。測試兩種類型路網，其一為虛擬路網，目的為先行測試整體程序運作流程，針對單一及多路徑路網探討，以了解不同路網之差別；其二為近年可取得實際規劃資料之桃園實際路網，測試各種回饋方式，比較其結果分析。本研究所欲探討課題分為下列兩點： 1.探討回饋變數：處理交通量指派後所產生的績效變數與指標形成回饋變數。 2.探討回饋方式：將依時性的動態交通量指派回饋至運輸規劃程序中，藉此探討如何將回饋變數回饋到運輸規劃程序。 本研究經由文獻回顧歸納，挑選較佳之兩種回饋方法，分別為連續平均法(Method of Successive Averages)以及固定權重(Constant Weights)法，其中CW方法權重另設0.6及0.3。以旅次數及旅行時間RSE和RMSE數值來觀察整個程序並比較分析其結果。在虛擬以及實際路網數值分析之結果綜合如下： 1.觀察到計算過程中，各遞迴所推估的變動皆有收斂的趨勢，而MSA收斂降幅較緩，CW(0.6)及CW(0.3)收斂型態較類似，呈現持續下降趨勢。並以CW(0.3)方法之收斂幅度以及最終收斂位置皆較低，為較佳之回饋方法。 2.與Boyce(2007)文獻之靜態實際路網處理之最終結果相同，亦是CW(0.3)方法較佳，顯示在動態之情況結果亦同。 本研究與過去研究之差別為考量動態之情況，在動態交通量指派處理依時性、依路徑資料之結果。顯示在DTA輸出回饋結果，回饋程序確實可改善不一致性，且亦可保持旅次分佈之狀態。 Since its introduction in late fifties, traditional sequential urban planning model has been implemented in a four steps manner as called Four-Step Model, or FSM in short. Although it has been recognized and widely accepted among transportation engineers and planners, an inherent drawback is its inconsistency of travel time (impedance) utilized in its sequence, notably between the trip distribution and traffic assignment step. This particular issue has been addressed in academic field but less known in practice. Several approaches have been proposed to tackle this problem of which the introduction of “feedback” loop seems particular appealing due to its simplicity and performance. In addition, recent development of dynamic traffic assignment has demonstrate its power to describe more real-world like traffic performance in road network to reveal the traffic flow pattern in both time and space dimensions. The later development complicates the inconsistency issue by adding time-varying and route-dependent travel time between any fixed given O-D pairs. In this study, various feedback approaches implemented in static planning model were examined and compared to be implemented under dynamic network context. A simulation-based dynamic traffic assignment model DynaTAIWAN was utilized to replace the usual traffic assignment procedure in FSM. Two particular problems will be addressed: (1) the representative travel times (indices) between given O-D pairs and (2) the feedback processes. The adequateness of these feedbacks introduced for the modified FSM was analyzed by two cases, one as a simple virtual network for primary diagnosis and the other a sized-down network form a selected city network (Tao-Yuan) for real-world operation exercise. After applying the continuous average method (Method of Successive Averages, MSA) as well as fixed weight (Constant Weights, CW) method with two alternative weight ratios (of 0.6 and 0.3 respectively), it has been shown that CW (0.3) outperformed the other methods in terms of lowest convergence level and fewer number of iterations to reach convergence. The results were generally in accordance with the previous related studies with static network cases, in particular the noted study by Boyce (2007).
