運輸需求預測整合模型統合架構之探討

A Combined Model for the Four-step Travel Demand Forecasting Procedure

基於用路人行為之均衡原則，本研究構建包括旅次發生、旅次分配、運具選擇與交通量指派在內之四階段整合模型，以克服傳統旅運預測之四階段界面不一致性的問題。整合模型經由改良後之超級路網的詮釋，可視為一個延伸性交通量指派問題。本研究以Frank-Wolfe演算法求解小規模路網，分別針對線性與非線性路正是成本函數加以驗證，並歸納出需求函數之係數的限制，文末亦比較整合模型與反饋式循序性核型之優劣點，並做出結論與建議。

Based on Wardrop's first principle, this research formulates a combined model covering the four steps, i.e., trip generation, trip distribution, modal split, traffic assignment, embedded in the traditional travel demand forecasting procedure. Using a modified supernetwork representation, the proposed combined model can be naturally regarded as an extension of the fixed demand traffic assignment model. The proposed combined model and the associated Frank-Wolfe algorithm are demonstrated with a hypothesized small network characterized by both linear and nonlinear link cost functions. The experiment also indicates that a certain restriction must be imposed on the parameters of origin-destination travel demand functions to ensure the convergence of applicable solution algorithms. A comparison of our proposed combined model and the sequential travel demand forecasting with feedback was also made. A few remarks are given in the end.