運輸需求預測是運輸規劃的核心模組，也是智慧型運輸系統中先進運輸資訊子系統的重要依據。伴隨著科技的進步與演算法效率的提升，運輸需求預測也逐漸從長期性的規劃性質朝向短期性的操作性質發展。為了因應此發展趨勢，運輸需求預測必須克服過去存在的三大課題，即：循序性運輸需求預測程序中界面的不一致性、演算法效率不符合即時性要求、未提供合理的路徑導引資訊。由於運輸需求整合模型可克服介面不一致性、導入時間維度的依時性的模型可呈現精細的運輸需求預測結果、而成對替選區段交通量指派（traffic assignment by paired alternative segments, TAPAS）演算法又具有快速精確的演算效率以及提供合理的唯一路徑解資訊，因此，本論文嘗試將TAPAS演算法應用於依時性單限旅次分佈交通量指派整合模型，依序提出依時性時空路段成本函數、整合模型數學架構、超級路網概念，發展延伸性的交通量指派演算法，然後以雙三角形小路網驗證其正確性。並且通過參數驗證，找出適用於內湖路網的依時性成本函數參數，並提出可能的通用參數a值及β值較大時，較符合路側VD結果。文末並提出研究結論與未來改善建議。;Transportation demand forecasting is a core module in transportation planning and an important basis for advanced transportation information subsystems in intelligent transportation systems. With the advancement of technology and improvement in algorithm efficiency, transportation demand forecasting has gradually evolved from long-term planning to short-term operational applications. To address this development trend, transportation demand forecasting must overcome three major challenges that existed in the past: inconsistency in the interfaces of sequential transportation demand forecasting procedures, algorithm efficiency not meeting real-time requirements, and lack of provision of reasonable path guidance information.
Since the integrated transportation demand model can overcome interface inconsistency, time-dependent models incorporating the time dimension can present detailed transportation demand forecasting results, and the Traffic Assignment by Paired Alternative Segments (TAPAS) algorithm has fast and accurate computational efficiency while providing reasonable unique path solution information, this thesis attempts to apply the TAPAS algorithm to a time-dependent integrated single-constrained trip distribution and traffic assignment model.
The study sequentially proposes a time-dependent spatiotemporal link cost function, the mathematical framework of the integrated model, and the super network concept. It develops an extensible traffic assignment algorithm and then verifies its correctness using a small double-triangle network. Through parameter validation, it identifies suitable time-dependent cost function parameters for the Neihu road network and suggests that when the possible universal parameter a value and β value are larger, the results are more consistent with roadside vehicle detector (VD) data.
The paper concludes with research findings and suggestions for future improvements.
