貨櫃運輸是海運業中成長最快的部門，然而由於近年來歐美地區與亞洲間貿易不平衡的 情況加劇，造成貨櫃航商需額外支出巨額成本以搬運空櫃至缺櫃港口，加上燃油價格飛 漲以及環保意識抬頭，如何有效率的管理空櫃調度問題已成為貨櫃航商的重要課題。遠 洋貨櫃航商之調度部門通常分為掌管區域間貨櫃調度的總管制中心與配合總管制中心 並掌管區域內調度的區域管制中心兩級。如此階層式依序決策的架構與雙層規劃的特性 相符，因此本計畫擬以雙層規劃來建構空櫃調度之決策模型。模型考量之因素包括多期 間、未來需求預測、空櫃安全存量、多種貨櫃規格、租還櫃成本與限制，模型之目的則 在求解各港口之空櫃調度與租櫃數量及其期間。為了反映不同決策階層間之互動性，本 計畫擬採用模糊決策方法以求解本問題；此一解法允許決策者參考其他層級之決策並藉 以修改本身決策以達成整體之最佳化。本計畫亦將採用傳統之KKT 最佳化條件轉換法 來求解本問題，以作為比較的基礎。顧慮到問題求解效率，本計畫亦將擬定結合啟發式 解法與問題分解技巧的演算法作為備案；該演算法將低階決策問題分解為眾多子問題， 並擬利用其多元商品網路流量問題的結構，應用專屬之演算法求解。本計畫擬針對個案 公司蒐集實務營運資料以評估模型之可行性。 Container shipping is the fastest growing sector in liner shipping. Owing to the trade imbalance between Europe/America and Asia, container carriers have experienced a great additional expense for repositioning empty containers from ports with surplus empty containers to ports with a shortage. Fuel price inflation and environmental concern also raise the importance of the empty container management issue. Container carriers generally divide the control of container movement to a global controller and many area controllers, where the global controller is in charge of the repositioning of empty containers between areas, while each area controller allocates the empty containers between ports within its area. Such a sequential decision manner is coincided with the bi-level programming approach, and thus this project plans to apply bi-level programming to modeling the container repositioning problem. The factors considered in the model include multi-periods, demand forecasting, empty container safety stock, multi-type containers, and leasing and returning costs of empty containers. The objective of the model is to find solutions of quantity and time of container allocation at each port. To facilitate the interaction between the upper-level and the lower-level decision-makers, this project intends to use the fuzzy decision approach to solve the problem. The fuzzy decision concept enables a decision-maker to adjust his decision based on the decision from other decision-makers at other levels, and to achieve an overall satisfactory solution. This project will also adopt the traditionally used KKT conditions transformation method to solve the problem. The solution from this method will serve as a comparison basis to the fuzzy approach. Concerning the efficiency of the above solution procedures, an alternative procedure which combines a heuristic method and a problem decomposition technique will also be investigated. This alternative solution procedure will decompose the lower-level problem to a set of sub-problems and take advantage of its special structure of a multi-commodity flows problem and solves these sub-problems by available algorithms. This project will use a case study to evaluate the performance of the proposed approach.