天然災害的發生，往往伴隨著運輸路網的大規模破壞，如何迅速恢復災區路網的運輸功能，將是救援相關研究的首要課題；然而就救災實務面分析，考量路網上多種運具從事救援工作，將能改進以單一運具為基礎所分析的救援路網重建計畫，更適切地反映真實路網上的救援運具使用，亦使救援相關的路網重建與設計問題更趨近於實務上的考量，是故本研究將在多種運具的考量下，進行緊急救援路網重建與設計問題的研析。本研究屬網路設計問題，將利用雙層規劃方法的觀念建立模型，上層模型以系統最佳化為目標，探討救災路網總旅行成本最小的救援路網重建策略，下層模型考量雙限起迄對選擇與多車種救災路徑選擇行為作為限制條件，在求解方法的設計上，採用變分不等式敏感度分析、廣義反矩陣方法、對角化法及拉氏梯度投影法等技術，並舉出數值範例分析以說明本救援路網重建模型的特性與正確性，預期本研究之成果，將可對政府、國軍及相關權責單位，提供救災過程中，路網設計及重建上重要的決策參考。

Natural disasters usually result in large-scale destruction of the transportation network. How to rapidly restore the basic functions of a transportation road network is the primary topic in rescue related research. However, using multi-user classes in a network to analyze the reality of rescue will improve rescue network reconstruction plans based on a single vehicle class. It can reflect flow characteristics of the real network more properly and make rescue related network design problem considerations more practical and efficient. Therefore, this research focuses on emergency network reconstruction and design problems under a multi-user classes condition. This research addresses a network design problem which can be formulated as a bi-level programming model. The upper level model conforms to the system optimum of the minimized total travel cost on the traffic network. The lower level model conforms to O-D pair choice of doubly constrained problem and multi-user classes rescue route choices. This research adopted some solution techniques such as variational inequality sensitivity analysis, generalized inverse matrix approach, diagonalization method, and lagrangian gradient projection method to solve the problem. The results of numerical tests in this research are expected to serve as important decision-making references for network reconstruction during the rescue process for government, military and authorities.