| 摘要: | 本研究以Chen and Hsueh(1996)發展之動態用路人均衡模型為基礎，分別構建動態多重用路人均衡模型與路段等候長度動態用路人均衡模型。動態多重用路人均衡模型主要探討當用路人為多種類(multiclass)多準則(multicriteria)之模型，以各用路人主觀決定旅行時間與旅行成本之權重，組成該用路人之旅行負效用，在該基礎下進行動態用路人均衡模型求解，以數值範例進行測試與分析。 就時間向度而言，路段流入率、流出率與車輛數是同一實體三種表象之轉換，過去研究已探討路段流入率與路段流出率容量之模型，因此本研究嘗試將容量限制於路段車輛數上，構建路段等候長度動態用路人均衡模型，利用擴張拉式對偶法及懲罰法兩種演算法求解，以測試範例驗證模型與演算法之正確性，分別比較兩種演算法之績效，及三種限制下構建之模型的結果與績效差異。 本研究參照Chen et al.(2002)提出之信賴度分析程序，結合蒙地卡羅模擬、動態用路人均衡模型、敏感度分析、不確定分析等以獲得各項信賴度資訊，探討路網容量信賴度。分析架構主要以蒙地卡羅模擬產生路段容量變數，獲得路網退化資料後，進行動態用路人均衡模型求解，並利用卓訓榮(1991)提出之廣義反矩陣法(Generalized Inverse Approach)求算敏感度分析資訊，其後結合各項信賴度分析方法，進一步獲得路網各項信賴度分析資訊進行評估。 Dynamic user equilibrium problem explored travelers’ choice behavior involving temporal dimension consideration, which has been tackled for years. Contrast to link flow as the decision variable in static traffic assignment, three variables are identified in the dynamic user equilibrium problem, i.e., link inflow, number of vehicles and exit flow. However, by concept of flow propagation, these three flow variables can be transformed each other. For example, the number of vehicles and exit flows can be deemed as subsequent states of link inflows. With only one decision variable adopted, the structure of dynamic user equilibrium model can be significantly simplified. Based on the results that have been developed so far, this study attempts to strengthen the applicability of the dynamic user equilibrium model by taking three more issues into study. The first issue is to relax the assumption of single class of travelers, thus yielding a multiclass or multicriteria dynamic user equilibrium model. The second issue is to retrospect the link capacity requirement from the practical point of view. In the past, either link inflow or exit flow was taken for link capacity constraint in the dynamic user equilibrium problem. Here, we take the number of vehicle (or equivalently link queue length) as the link capacity constraint instead, and solve it accordingly by the penalty and augmented Lagrangian methods. The results show that under the assumption that time dependent network constructed by the actual link travel times, the outputs obtained from the link capacity constraints formed by link inflow, exit flow or the number of vehicles are basically comparable. The last issue is to deal with link capacity reliability analysis. The framework is mainly borrowed from Chen et al. (2002), which involves sensitivity analysis embedding a generalized inverse approach, uncertainly/risk analysis within a Monte Carlo simulation. This systematic procedure is general in that the reliability on connectivity and travel times is constructed as special cases. The findings associated with the three issues are summarized and few remarks are given in the end to conclude the research. |
